Docstoc

Berkshire Encyclopedy

Document Sample
Berkshire Encyclopedy Powered By Docstoc
					                       BERKSHIRE PUBLISHING GROUP




   Berkshire
   Encyclopedia of

Human—Computer
Interaction                 When science fiction
                            becomes science fact




         William Sims Bainbridge
 Edited by
 National Science Foundation
Berkshire Encyclopedia of
Human-Computer
  Interaction
Berkshire Encyclopedia of
Human-Computer
  Interaction
            VOLUME               1
   William Sims Bainbridge
                    Editor




       Great Barrington, Massachusetts U.S.A.
          www.berkshirepublishing.com
Copyright © 2004 by Berkshire Publishing Group LLC

All rights reserved. No part of this book may be reproduced or utilized in any form or by any means, elec-
tronic or mechanical, including photocopying, recording, or by any information storage and retrieval sys-
tem, without permission in writing from the publisher.

Cover photo: Thad Starner sporting a wearable computer.
   Photo courtesy of Georgia Institute of Technology.
Cover background image: Courtesy of Getty Images.

For information:
     Berkshire Publishing Group LLC
     314 Main Street
     Great Barrington, Massachusetts 01230
     www.berkshirepublishing.com

Printed in the United States of America

Library of Congress Cataloging-in-Publishing Data
Berkshire encyclopedia of human-computer interaction / William Sims Bainbridge, editor.
       p. cm.
   “A Berkshire reference work.”
   Includes bibliographical references and index.
   ISBN 0-9743091-2-5 (hardcover : alk. paper)
  1. Human-computer interaction--Encyclopedias. I. Bainbridge, William Sims. II. Title.

  QA76.9.H85B46 2004
  004'.01'9--dc22
                                                                                              2004017920
  BERKSHIRE PUBLISHING STAFF
                        Project Director
                       Karen Christensen

                     Project Coordinators
            Courtney Linehan and George Woodward

                        Associate Editor
                          Marcy Ross

                          Copyeditors
Francesca Forrest, Mike Nichols, Carol Parikh, and Daniel Spinella

          Information Management and Programming
               Deborah Dillon and Trevor Young

                      Editorial Assistance
                       Emily Colangelo

                          Designer
                       Monica Cleveland

                    Production Coordinator
                         Janet Lowry

                      Composition Artists
        Steve Tiano, Brad Walrod, and Linda Weidemann

                    Composition Assistance
                        Pam Glaven

                         Proofreaders
   Mary Bagg, Sheila Bodell, Eileen Clawson, and Cassie Lynch

                     Production Consultant
                          Jeff Potter

                            Indexer
                         Peggy Holloway
                                                           CONTENTS




                         List of Entries, ix
                        Reader’s Guide, xv
                       List of Sidebars, xix
                        Contributors, xxiii
                       Introduction, xxxiii
                       Publisher’s Note, xli
                      About the Editor, xliii

                            Entries
                        Volume I: A–L
                            1–440
                         Vol II: M–W
                           441–826
                     Appendix 1: Glossary, 827
Appendix 2: Master Bibliography of Human-Computer Interaction, 831
                    HCI in Popular Culture, 893
                             Index, 931
                                  •
                Index repeated in this volume, I-1

                                                                     vii
                                                         LIST OF ENTRIES




Adaptive Help Systems       Animation                    Artificial Intelligence
  Peter Brusilovsky           Abdennour El Rhalibi         Robert A. St. Amant
                              Yuanyuan Shen
Adaptive Interfaces                                      Asian Script Input
  Alfred Kobsa              Anthropology and HCI           William Sims Bainbridge
                              Allen W. Batteau             Erika Bainbridge
Affective Computing
  Ira Cohen                 Anthropometry                Atanasoff-Berry Computer
  Thomas S. Huang             Victor L. Paquet             John Gustafson
  Lawrence S. Chen            David Feathers
                                                         Attentive User Interface
Altair                      Application Use Strategies     Ted Selker
  William Sims Bainbridge     Suresh K. Bhavnani
                                                         Augmented Cognition
Alto                        Arpanet                        Amy Kruse
  William Sims Bainbridge     Amy Kruse                    Dylan Schmorrow
                              Dylan Schmorrow
                              Allen J. Sears

                                                                                     ix
X ❘❙❚ BERKSHIRE ENCYCLOPEDIA OF HUMAN-COMPUTER INTERACTION



Augmented Reality              Compilers                     Digital Divide
  Rajeev Sharma                  Woojin Paik                   Linda A. Jackson
  Kuntal Sengupta
                               Computer-Supported            Digital Government
Avatars                          Cooperative Work              Jane E. Fountain
  Jeremy Bailenson               John M. Carroll               Robin A. McKinnon
  James J. Blascovich            Mary Beth Rosson
                                                             Digital Libraries
Beta Testing                   Constraint Satisfaction         Jose-Marie Griffiths
  Gina Neff                      Berthe Y. Choueiry
                                                             Drawing and Design
Braille                        Converging Technologies         Mark D. Gross
  Oleg Tretiakoff                William Sims Bainbridge
                                                             E-business
Brain-Computer Interfaces      Cybercommunities                Norhayati Zakaria
  Melody M. Moore                Lori Kendall
  Adriane D. Davis                                           Education in HCI
  Brendan Z. Allison           Cybersex                        Jan Stage
                                 David L. Delmonico
Browsers                         Elizabeth Griffin           Electronic Journals
  Andy Cockburn                                                 Carol Tenopir
                               Cyborgs
Cathode Ray Tubes                William Sims Bainbridge     Electronic Paper Technology
  Gregory P. Crawford                                           Gregory P. Crawford
                               Data Mining
CAVE                             Mohammad Zaki               Eliza
  Thomas DeFanti                                                William H. Sterner
  Dan Sandin                   Data Visualization
                                 Kwan-Liu Ma                 E-mail
Chatrooms                                                      Nathan Bos
  Amanda B. Lenhart            Deep Blue
                                 Murray Campbell             Embedded Systems
Children and the Web                                           Ronald D. Williams
  Dania Bilal                  Denial-of-Service Attack
                                 Adrian Perrig               ENIAC
Classrooms                       Abraham Yaar                  William Sims Bainbridge
  Chris Quintana
                               Desktop Metaphor              Ergonomics
Client-Server Architecture       Jee-In Kim                    Ann M. Bisantz
  Mark Laff
                               Dialog Systems                Errors in Interactive Behavior
Cognitive Walkthrough            Susan W. McRoy                Wayne D. Gray
  Marilyn Hughes Blackmon
                               Digital Cash                  Ethics
Collaboratories                  J. D. Tygar                   Helen Nissenbaum
  Gary M. Olson
                                                                         LIST OF ENTRIES ❚❙❘ XI



Ethnography                      Handwriting Recognition and   Information Theory
  David Hakken                     Retrieval                      Ronald R. Kline
                                   R. Manmatha
Evolutionary Engineering           V. Govindaraju              Instruction Manuals
  William Sims Bainbridge                                         David K. Farkas
                                 Haptics
Expert Systems                     Ralph L. Hollis             Internet—Worldwide Diffusion
  Jay E. Aronson                                                  Barry Wellman
                                 History of Human-Computer        Phuoc Tran
Eye Tracking                       Interaction                    Wenhong Chen
  Andrew T. Duchowski              Jonathan Grudin
                                                               Internet in Everyday Life
Facial Expressions               Hollerith Card                   Barry Wellman
  Irfan Essa                       William Sims Bainbridge        Bernie Hogan

Fly-by-Wire                      Human-Robot Interaction       Iterative Design
   C. M. Krishna                   Erika Rogers                   Richard Baskerville
                                                                  Jan Stage
Fonts                            Hypertext and Hypermedia
  Thomas Detrie                    David K. Farkas             Keyboard
  Arnold Holland                                                 Alan Hedge
                                 Icons
Games                               Stephanie Ludi             Language Generation
  Abdennour El Rhalibi                                           Regina Barzilay
                                 Identity Authentication
Gender and Computing               Ashutosh P. Deshpande       Laser Printer
  Linda A. Jackson                 Parag Sewalkar                Gary Starkweather

Geographic Information Systems   Impacts                       Law and HCI
  Michael F. Goodchild             Chuck Huff                    Sonia E. Miller

Gesture Recognition              Information Filtering         Law Enforcement
  Francis Quek                      Luz M. Quiroga               Roslin V. Hauck
                                    Martha E. Crosby
Graphical User Interface                                       Lexicon Building
  David England                  Information Organization        Charles J. Fillmore
                                    Dagobert Soergel
Grid Computing                                                 Liquid Crystal Displays
  Cavinda T. Caldera             Information Overload            Gregory P. Crawford
                                    Ruth Guthrie
Groupware                                                      Literary Representations
  Timothy J. Hickey              Information Retrieval            William Sims Bainbridge
  Alexander C. Feinman              Dagobert Soergel
                                                               Machine Translation
Hackers                          Information Spaces             Katrin Kirchhoff
  Douglas Thomas                    Fionn Murtagh
XII ❘❙❚ BERKSHIRE ENCYCLOPEDIA OF HUMAN-COMPUTER INTERACTION



Markup Languages                 Online Education                Programming Languages
 Hong-Gee Kim                      Robert S. Stephenson            David MacQueen
                                   Glenn Collyer
Mobile Computing                                                 Prototyping
 Dharma P. Agrawal               Online Questionnaires             Richard Baskerville
                                   James Witte                     Jan Stage
Mosaic                             Roy Pargas
 William Sims Bainbridge                                         Psychology and HCI
                                 Online Voting                     Judith S. Olson
Motion Capture and Recognition     R. Michael Alvarez
 Jezekiel Ben-Arie                 Thad E. Hall                  Recommender and Reputation
                                                                   Systems
Mouse                            Ontology                          Cliff Lampe
 Shumin Zhai                       Christopher A. Welty            Paul Resnick

Movies                           Open Source Software            Repetitive Strain Injury
 William Sims Bainbridge           Gregory R. Madey                Jack Tigh Dennerlein

MUDs                             Optical Character Recognition   Scenario-Based Design
 Richard Allan Bartle              V. Govindaraju                  John M. Carroll
                                   Swapnil Khedekar
Multiagent systems                                               Search and Rescue
 Gal A. Kaminka                  Peer-to-Peer Architecture         Howie Choset
                                   Julita Vassileva
Multimodal Interfaces                                            Search Engines
 Rajeev Sharma                   Pen and Stylus Input              Shannon Bradshaw
 Sanshzar Kettebekov               Alan Hedge
 Guoray Cai                                                      Security
                                 Personality Capture               Bhavani Thuraisingham
Multiuser Interfaces               William Sims Bainbridge
 Prasun Dewan                                                    Semantic Web
                                 Physiology                        Bhavani Thuraisingham
Musical Interaction                Jennifer Allanson
 Christopher S. Raphael                                          Smart Homes
 Judy A. Franklin                Planning                          Diane J. Cook
                                   Sven Koenig                     Michael Youngblood
Natural-Language Processing        Michail G. Lagoudakis
  James H. Martin                                                Sociable Media
                                 Pocket Computer                   Judith Donath
Navigation                         William Sims Bainbridge
  John J. Rieser                                                 Social Informatics
                                 Political Science and HCI         Howard Rosenbaum
N-grams                            James N. Danziger
  James H. Martin                  Michael J. Jensen             Social Proxies
                                                                   Thomas Erickson
Olfactory Interaction            Privacy                           Wendy A. Kellogg
  Ricardo Gutierrez-Osuna          Jeffrey M. Stanton
                                                                        LIST OF ENTRIES ❚❙❘ XIII



Social Psychology and HCI       Task Analysis                     Jenny Preece
  Susan R. Fussell                Erik Hollnagel                  Diane Maloney-Krichmar

Sociology and HCI               Telecommuting                   Value Sensitive Design
  William Sims Bainbridge         Ralph David Westfall            Batya Friedman

Socio-Technical System Design   Telepresence                    Video
  Walt Scacchi                    John V. Draper                  Immanuel Freedman

Software Cultures               Text Summarization              Video Summarization
  Vaclav Rajlich                  Judith L. Klavans               A. Murat Tekalp

Software Engineering            Theory                          Virtual Reality
  Richard Kazman                  Jon May                         Larry F. Hodges
                                                                  Benjamin C. Lok
Sonification                    Three-Dimensional Graphics
  David M. Lane                   Benjamin C. Lok               Viruses
  Aniko Sandor                                                    J. D. Tygar
  S. Camille Peres              Three-Dimensional Printing
                                  William Sims Bainbridge       Visual Programming
Spamming                                                          Margaret M. Burnett
  J. D. Tygar                   Touchscreen                       Joseph R. Ruthruff
                                  Andrew L. Sears
Speech Recognition                Rich Goldman                  Wearable Computer
  Mary P. Harper                                                 Thad Starner
  V. Paul Harper                Ubiquitous Computing             Bradley Rhodes
                                  Olufisayo Omojokun
Speech Synthesis                  Prasun Dewan                  Website Design
  Jan P.H. van Santen                                            Barbara S. Chaparro
                                Unicode                          Michael L. Bernard
Speechreading                     Unicode Editorial Committee
  Marcus Hennecke                                               Work
                                Universal Access                 Christine A. Halverson
Spell Checker                     Gregg Vanderheiden
  Woojin Paik                                                   Workforce
                                Usability Evaluation             Brandon DuPont
Sphinx                            Jean Scholtz                   Joshua L. Rosenbloom
  Rita Singh
                                User Modeling                   World Wide Web
Statistical Analysis Support      Richard C. Simpson             Michael Wilson
  Robert A. St. Amant
                                User Support                    WYSIWYG
Supercomputers                    Indira R. Guzman               David M. Lane
  Jack Dongarra
                                User-Centered Design
Tablet Computer                   Chadia Abras
  William Sims Bainbridge
                                                                            READER’S GUIDE




This list is provided to assist readers in locating en-    Navigation
tries on related topics. It classifies articles into ten   Online Education
general categories: Applications; Approaches;              Online Voting
Breakthroughs; Challenges; Components; Disciplines;        Planning
Historical Development; Interfaces; Methods; and           Recommender and Reputation Systems
Social Implications. Some entries appear in more           Search and Rescue
than one category.                                         Statistical Analysis Support
                                                           Supercomputers
Applications                                               Telecommuting
Classrooms                                                 Ubiquitous Computing
Digital Government                                         Video
Digital Libraries
E-business                                                 Approaches
Games                                                      Application Use Strategies
Geographic Information Systems                             Beta Testing
Grid Computing                                             Cognitive Walkthrough
Law Enforcement                                            Constraint Satisfaction
Mobile Computing                                           Ethics

                                                                                                xv
XVI ❘❙❚ BERKSHIRE ENCYCLOPEDIA OF HUMAN-COMPUTER INTERACTION



Ethnography                                    Components
Evolutionary Engineering                       Adaptive Help Systems
Information Theory                             Animation
Iterative Design                               Braille
Ontology                                       Cathode Ray Tubes
Open Source Software                           Client-Server Architecture
Prototyping                                    Desktop Metaphor
Scenario-Based Design                          Electronic Paper Technology
Social Informatics                             Fonts
Socio-Technical System Design                  Keyboard
Task Analysis                                  Laser Printer
Theory                                         Liquid Crystal Displays
Universal Access                               Mouse
Usability Evaluation                           N-grams
User Modeling                                  Peer-to-Peer Architecture
User-Centered Design                           Social Proxies
Value Sensitive Design                         Spell Checker
Website Design                                 Touchscreen
                                               Unicode
Breakthroughs                                  WYSIWYG
Altair
Alto                                           Disciplines
Arpanet                                        Anthropology and HCI
Atanasoff-Berry Computer                       Artificial Intelligence
CAVE                                           Ergonomics
Converging Technologies                        Law and HCI
Deep Blue                                      Political Science and HCI
Eliza                                          Psychology and HCI
ENIAC                                          Social Psychology and HCI
Hollerith Card                                 Sociology and HCI
Mosaic
Sphinx                                         Historical Development
                                               Altair
Challenges                                     Alto
Denial-of-Service Attack                       ENIAC
Digital Divide                                 History of HCI
Errors in Interactive Behavior
Hackers                                        Interfaces
Identity Authentication                        Adaptive Interfaces
Information Filtering                          Affective Computing
Information Overload                           Anthropometry
Privacy                                        Asian Script Input
Repetitive Strain Injury                       Attentive User Interface
Security                                       Augmented Cognition
Spamming                                       Augmented Reality
Viruses                                        Brain-Computer Interfaces
                                                              READER’S GUIDE ❚❙❘ XVII



Compilers                               Markup Languages
Data Visualization                      Motion Capture and Recognition
Dialog Systems                          Natural-Language Processing
Drawing and Design                      Optical Character Recognition
Eye Tracking                            Personality Capture
Facial Expressions                      Programming Languages
Fly-by-Wire                             Search Engines
Graphical User Interface                Semantic Web
Haptics                                 Software Engineering
Multimodal Interfaces                   Sonification
Multiuser Interfaces                    Speech Recognition
Musical Interaction                     Speech Synthesis
Olfactory Interaction                   Speechreading
Online Questionnaires                   Text Summarization
Pen and Stylus Input                    User Support
Physiology                              Video Summarization
Pocket Computer                         Visual Programming
Smart Homes                             World Wide Web
Tablet Computer                         Social Implications
Telepresence                            Chatrooms
Three-Dimensional Graphics              Children and the Web
Three-Dimensional Printing              Collaboratories
Virtual Reality                         Computer-Supported Cooperative Work
Wearable Computer                       Cybercommunities
                                        Cybersex
Methods                                 Cyborgs
Avatars                                 Education in HCI
Browsers                                Electronic Journals
Data Mining                             E-mail
Digital Cash                            Gender and Computing
Embedded Systems                        Groupware
Expert Systems                          Human-Robot Interaction
Gesture Recognition                     Impacts
Handwriting Recognition and Retrieval   Internet—Worldwide Diffusion
Hypertext and Hypermedia                Internet in Everyday Life
Icons                                   Literary Representations
Information Organization                Movies
Information Retrieval                   MUDs
Information Spaces                      Multiagent systems
Instruction Manuals                     Sociable Media
Language Generation                     Software Cultures
Lexicon Building                        Work
Machine Translation                     Workforce
                                                                    LIST OF SIDEBARS




Adaptive Help Systems                                 Chatrooms
Farewell “Clippy”                                     Life Online

Adaptive Interfaces                                   Classrooms
Keeping Disabled People in the Technology Loop        History Comes Alive in Cyberspace
                                                      Learning through Multimedia
Anthropology and HCI
Digital Technology Helps Preserve Tribal Language     Computer-Supported Cooperative Work
                                                      Internet Singing Lessons
Anthropology and HCI                                  Social Context in Computer-Supported
Eastern vs. Western Cultural Values                     Cooperative Work

Augmented Cognition                                   Cybercommunities
Putting Humans First in Systems Design                Welcome to LamdaMOO

Braille                                               Cybersex
Enhancing Access to Braille Instructional Materials   Cybersex Addiction


                                                                                             xix
XX ❘❙❚ BERKSHIRE ENCYCLOPEDIA OF HUMAN-COMPUTER INTERACTION



Digital Divide                                  Internet in Everyday Life
HomeNetToo Tries to Bridge Digital Divide       Finding Work Online
                                                Information Technology and Competitive
Digital Libraries                                 Academic Debate
Vannevar Bush on the Memex
                                                Law Enforcement
Education in HCI                                Fighting Computer Crime
Bringing HCI Into the Real World
                                                Literary Representations
Eliza                                           Excerpt from Isaac Asimov’s I, Robot
Talking with ELIZA                              Excerpt from “The Sand-Man” (1817) by
                                                   E. T. A. Hoffman
E-mail
The Generation Gap                              Machine Translation
                                                Warren Weaver on Machine Translation
Errors in Interactive Behavior
To Err Is Technological                         Movies
                                                HAL’s Birthday Celebration
Fonts
Our Most Memorable Nightmare                    MUDs
                                                The Wide World of a MUD
Gender and Computing
“Computer Girl” Site Offers Support for         Online Education
  Young Women                                   An Online Dig for Archeology Students
Narrowing the Gap                               Virtual Classes Help Rural Nurses

Geographic Information Systems                  Political Science and HCI
Geographic Information Systems Aid Land         Washington Tales of the Internet
  Conservation
                                                Psychology and HCI
Groupware                                       Human Factors Come into the Forefront
Away Messages                                   Virtual Flight for White-Knuckled Travelers
The Wide World of Wikis
                                                Repetitive Strain Injury
History of HCI                                  The Complexities of Repetitive Strain
Highlights from My Forty Years of HCI History
                                                Scenario-Based Design
Human-Robot Interaction                         The Value of a Devil’s Advocate
Carbo-Powered Robots
                                                Social Psychology and HCI
Hypertext and Hypermedia                        Love and HCI
Ted Nelson on Hypertext and the Web
                                                Sociology and HCI
Impacts                                         “Who’s on First” for the Twenty-First
Therac-25 Safety Is a System Property             Century
                                                                  LIST OF SIDEBARS ❚❙❘ XXI



Spell Checker                             Work
Check the Spell Checker                   Software Prescribes Break Time for Enhanced
                                            Productivity
Task Analysis
Excerpt from Cheaper by the Dozen         Workforce
                                          Cultural Differences
Unicode                                   Employee Resistance to Technology
History and Development of Unicode
Relationship of the Unicode Standard to   World Wide Web
  ISO_IEC 10646                           “Inventing” the World Wide Web
                                          Tim Berners-Lee on the Web as Metaphor
Usability Evaluation
Global Usability                          WYSIWYG
Is Usability Still a Problem?             The Future of HCI
                                                    CONTRIBUTORS




Abras, Chadia                 Alvarez, R. Michael
Goucher College               Caltech-MIT Voting Technology Project
  User-Centered Design          Online Voting

Agrawal, Dharma P.            Aronson, Jay E.
University of Cincinnati      University of Georgia
  Mobile Computing              Expert Systems

Allanson, Jennifer            Bailenson, Jeremy
Lancaster University          Stanford University
  Physiology                     Avatars

Allison, Brendan Z.           Bainbridge, Erika
Georgia State University      Harvard University, Center for Hellenic Studies
  Brain-Computer Interfaces     Asian Script Input




                                                                            xxiii
XXIV ❘❙❚ BERKSHIRE ENCYCLOPEDIA OF HUMAN-COMPUTER INTERACTION



Bainbridge, William Sims                       Bilal, Dania
National Science Foundation                    University of Tennessee
  Altair                                          Children and the Web
  Alto
  Asian Script Input                           Bisantz, Ann M.
  Converging Technologies                      State University of New York, Buffalo
  Cyborgs                                         Ergonomics
  ENIAC
  Evolutionary Engineering                     Blackmon, Marilyn Hughes
  Hollerith Card                               University of Colorado, Boulder
  Literary Representations                       Cognitive Walkthrough
  Mosaic
  Movies                                       Blascovich, James J.
  Personality Capture                          University of California, Santa Barbara
  Pocket Computer                                Avatars
  Sociology and HCI
  Tablet Computer                              Bos, Nathan
  Three-Dimensional Printing                   University of Michigan
                                                 E-mail
Bartle, Richard Allan
Multi-User Entertainment Limited               Bradshaw, Shannon
  MUDs                                         University of Iowa
                                                 Search Engines
Barzilay, Regina
Massachusetts Institute of Technology          Brusilovsky, Peter
  Language Generation                          University of Pittsburgh
                                                 Adaptive Help Systems
Baskerville, Richard
Georgia State University                       Burnett, Margaret M.
  Iterative Design                             Oregon State University
  Prototyping                                    Visual Programming

Batteau, Allen W.                              Cai, Guoray
Wayne State University                         Pennsylvania State University
  Anthropology and HCI                           Multimodal Interfaces

Ben-Arie, Jezekiel                             Caldera, Cavinda T.
University of Illinois, Chicago                Syracuse University
  Motion Capture and Recognition                 Grid Computing

Bernard, Michael L.                            Campbell, Murray
Wichita State University                       IBM T.J. Watson Research Center
  Website Design                                 Deep Blue

Bhavnani, Suresh K.
University of Michigan
  Application Use Strategies
                                                                       CONTRIBUTORS ❚❙❘ XXV



Carroll, John M.                           Crosby, Martha E.
Pennsylvania State University              University of Hawaii
  Computer-Supported Cooperative Work        Information Filtering
  Scenario-Based Design
                                           Danziger, James N.
Chaparro, Barbara S.                       University of California, Irvine
Wichita State University                     Political Science and HCI
  Website Design
                                           Davis, Adriane D.
Chen, Lawrence                             Georgia State University
Eastman Kodak Research Labs                  Brain-Computer Interfaces
  Affective Computing
                                           DeFanti, Thomas
Chen, Wenhong                              University of Illinois, Chicago
University of Toronto                        Cave
  Internet – Worldwide Diffusion
                                           Delmonico, David L.
Choset, Howie                              Duquesne University
Carnegie Mellon University                   Cybersex
  Search and Rescue
                                           Dennerlien, Jack Tigh
Choueiry, Berthe Y.                        Harvard School of Public Health
University of Nebraska, Lincoln              Repetitive Strain Injury
  Constraint Satisfaction
                                           Deshpande, Ashutosh P.
Cockburn, Andy                             Syracuse University
University of Canterbury                     Identity Authentication
  Browsers
                                           Detrie, Thomas
Cohen, Ira                                 Arizona State University
Hewlett-Packard Research Labs,               Fonts
University of Illinois, Urbana-Champaign
  Affective Computing                      Dewan, Prasun
                                           Microsoft Corporation
Collyer, Glenn                               Multiuser Interfaces
iDacta, Inc.                                 Ubiquitous Computing
  Online Education
                                           Donath, Judith
Cook, Diane J.                             Massachusetts Institute of Technology
University of Texas, Arlington               Sociable Media
  Smart Homes
                                           Dongarra, Jack
Crawford, Gregory P.                       University of Tennessee
Brown University                             Supercomputers
  Cathode Ray Tubes
  Electronic Paper Technology
  Liquid Crystal Displays
XXVI ❘❙❚ BERKSHIRE ENCYCLOPEDIA OF HUMAN-COMPUTER INTERACTION



Draper, John V.                                Fountain, Jane E.
Raven Research                                 Harvard University
  Telepresence                                   Digital Government

Duchowski, Andrew T.                           Franklin, Judy A.
Clemson University                             Smith College
  Eye Tracking                                   Musical Interaction

DuPont, Brandon                                Freedman, Immanuel
Policy Research Institute                      Dr. Immanuel Freedman, Inc.
  Workforce                                      Video

El Rhalibi, Abdennour                          Friedman, Batya
Liverpool John Moores University               University of Washington
   Animation                                     Value Sensitive Design
   Games
                                               Fussell, Susan R.
England, David                                 Carnegie Mellon University
Liverpool John Moores University                 Social Psychology and HCI
  Graphical User Interface
                                               Goldman, Rich
Erickson, Thomas                               University of Maryland, Baltimore
IBM T. J. Watson Research Center                 Touchscreen
  Social Proxies
                                               Goodchild, Michael F.
Essa, Irfan                                    University of California, Santa Barbara
Georgia Institute of Technology                  Geographic Information Systems
  Facial Expressions
                                               Govindaraju, V.
Farkas, David K.                               University at Buffalo
University of Washington                         Handwriting Recognition and Retrieval
  Hypertext and Hypermedia                       Optical Character Recognition
  Instruction Manuals
                                               Gray, Wayne D.
Feathers, David                                Rensselaer Polytechnic Institute
State University of New York, Buffalo            Errors in Interactive Behavior
   Anthropometry
                                               Griffin, Elizabeth J.
Feinman, Alexander C.                          Internet Behavior Consulting
Brandeis University                              Cybersex
  Groupware
                                               Griffiths, Jose-Marie
Fillmore, Charles J.                           University of Pittsburgh
International Computer Science Institute         Digital Libraries
   Lexicon Building
                                                                          CONTRIBUTORS ❚❙❘ XXVII



Gross, Mark D.                                Hauck, Roslin V.
University of Washington                      Illinois State University
  Drawing and Design                              Law Enforcement

Grudin, Jonathan                              Hedge, Alan
Microsoft Research                            Cornell University
  Computer Science                              Keyboard
  History of HCI                                Pen and Stylus Input

Gustafson, John                               Hennecke, Marcus
Sun Microsystems                              TEMIC Telefunken Microelectronic GmbH
  Atanasoff-Berry Computer                      Speechreading

Guthrie, Ruth                                 Hickey, Timothy J.
California Polytechnic University of Pomona   Brandeis University
  Information Overload                          Groupware

Gutierrez-Osuna, Ricardo                      Hodges, Larry F.
Texas A&M University                          University of North Carolina, Charlotte
  Olfactory Interaction                         Virtual Reality

Guzman, Indira R.                             Hogan, Bernie
Syracuse University                           University of Toronto
  User Support                                  Internet in Everyday Life

Hakken, David                                 Holland, Arnold
State University of New York Institute of     California State University, Fullerton
  Technology                                    Fonts
   Ethnography
                                              Hollis, Ralph L.
Hall, Thad E.                                 Carnegie Mellon University
Century Foundation                              Haptics
  Online Voting
                                              Hollnagel, Erik
Halverson, Christine                          University of Linköping
IBM T. J. Watson Research Center                Task Analysis
  Work
                                              Huang, Thomas S.
Harper, Mary P.                               University of Illinois, Urbana-Champaign
Purdue University                               Affective Computing
  Speech Recognition
                                              Huff, Chuck
Harper, V. Paul                               Saint Olaf College
United States Patent and Trademark Office       Impacts
  Speech Recognition
XXVIII ❘❙❚ BERKSHIRE ENCYCLOPEDIA OF HUMAN-COMPUTER INTERACTION



Jackson, Linda A.                                Klavans, Judith L.
Michigan State University                        Columbia University
   Digital Divide                                  Text Summarization
   Gender and Computing
                                                 Kline, Ronald R.
Jensen, Michael J.                               Cornell University
University of California, Irvine                   Information Theory
   Political Science and HCI
                                                 Kobsa, Alfred
Kaminka, Gal                                     University of California, Irvine
Bar Ilan University                                Adaptive Interfaces
  Multiagent systems
                                                 Koenig, Sven
Kazman, Richard                                  Georgia Institute of Technology
Carnegie Mellon University                         Planning
  Software Engineering
                                                 Krishna, C. M.
Kellogg, Wendy A.                                University of Massachusetts, Amherst
IBM T. J. Watson Research Center                   Fly-by-Wire
  Social Proxies
                                                 Kruse, Amy
Kendall, Lori                                    Strategic Analysis, Inc.
State University of New York, Purchase College      Arpanet
   Cybercommunities                                 Augmented Cognition

Kettebekov, Sanshzar                             Laff, Mark
Oregon Health and Science University             IBM T.J. Watson Research Center
  Multimodal Interfaces                            Client-Server Architecture

Khedekar, Swapnil                                Lagoudakis, Michail G.
University at Buffalo                            Georgia Institute of Technology
  Optical Character Recognition                    Planning

Kim, Hong-Gee                                    Lampe, Cliff
Dankook University                               University of Michigan
  Markup Languages                                 Recommender and Reputation Systems

Kim, Jee-In                                      Lane, David M.
Konkuk University                                Rice University
  Desktop Metaphor                                 Sonification
                                                   WYSIWYG
Kirchhoff, Katrin
University of Washington                         Lenhart, Amanda B.
  Machine Translation                            Pew Internet & American Life Project
                                                   Chatrooms
                                                                    CONTRIBUTORS ❚❙❘ XXIX



Lok, Benjamin C.                          Miller, Sonia E.
University of Florida                     S. E. Miller Law Firm
  Three-Dimensional Graphics                 Law and HCI
  Virtual Reality
                                          Moore, Melody M.
Ludi, Stephanie                           Georgia State University
Rochester Institute of Technology           Brain-Computer Interfaces
  Icons
                                          Murtagh, Fionn
Ma, Kwan-Liu                              Queen’s University, Belfast
University of California, Davis            Information Spaces
  Data Visualization
                                          Neff, Gina
MacQueen, David                           University of California, Los Angeles
University of Chicago                       Beta Testing
  Programming Languages
                                          Nissenbaum, Helen
Madey, Gregory R.                         New York University
University of Notre Dame                    Ethics
  Open Source Software
                                          Olson, Gary M.
Maloney-Krichmar, Diane                   University of Michigan
Bowie State University                      Collaboratories
  User-Centered Design
                                          Olson, Judith S.
Manmatha, R.                              University of Michigan
University of Massachusetts, Amherst        Psychology and HCI
  Handwriting Recognition and Retrieval
                                          Omojokun, Olufisayo
Martin, James H.                          University of North Carolina, Chapel Hill
University of Colorado, Boulder             Ubiquitous Computing
  Natural-Language Processing
  N-grams                                 Paik, Woojin
                                          University of Massachusetts, Boston
May, Jon                                    Compilers
University of Sheffield                     Spell Checker
  Theory
                                          Paquet, Victor L.
McKinnon, Robin A.                        State University of New York, Buffalo
Harvard University                           Anthropometry
  Digital Government
                                          Pargas, Roy
McRoy, Susan W.                           Clemson University
University of Wisconsin, Milwaukee          Online Questionnaires
  Dialog Systems
XXX ❘❙❚ BERKSHIRE ENCYCLOPEDIA OF HUMAN-COMPUTER INTERACTION



Peres, S. Camille                              Rogers, Erika
Rice University                                California Polytechnic State University
  Sonification                                   Human-Robot Interaction

Perrig, Adrian                                 Rosenbaum, Howard
Carnegie Mellon University                     Indiana University
  Denial-of-Service Attack                       Social Informatics

Preece, Jenny                                  Rosenbloom, Joshua L.
University of Maryland, Baltimore County       University of Kansas
  User-Centered Design                           Workforce

Quek, Francis                                  Rosson, Mary Beth
Wright State University                        Pennsylvania State University
 Gesture Recognition                             Computer-Supported Cooperative Work

Quintana, Chris                                Ruthruff, Joseph R.
University of Michigan                         Oregon State University
  Classrooms                                     Visual Programming

Quiroga, Luz M.                                Sandin, Dan
University of Hawaii                           University of Illinois, Chicago
  Information Filtering                          CAVE

Rajlich, Vaclav                                Sandor, Aniko
Wayne State University                         Rice University
  Software Cultures                              Sonification

Raphael, Christopher S.                        Scacchi, Walt
University of Massachusetts, Amherst           University of California, Irvine
  Musical Interaction                            Socio-Technical System Design

Resnick, Paul                                  Schmorrow, Dylan
University of Michigan                         Defense Advanced Projects Agency
  Recommender and Reputation Systems             Arpanet
                                                 Augmented Cognition
Rhodes, Bradley
Ricoh Innovations                              Scholtz, Jean
  Wearable Computer                            National Institute of Standards and Technology
                                                 Usability Evaluation
Rieser, John J.
Vanderbilt University                          Sears, Andrew L.
  Navigation                                   University of Maryland, Baltimore County
                                                 Touchscreen
                                                                          CONTRIBUTORS ❚❙❘ XXXI



Sears, J. Allen                                 Stage, Jan
Corporation for National Research Initiatives   Aalborg University
  Arpanet                                         Education in HCI
                                                  Iterative Design
Selker, Ted                                       Prototyping
Massachusetts Institute of Technology
  Attentive User Interface                      Stanton, Jeffrey M.
                                                Syracuse University
Sewalkar, Parag                                    Privacy
Syracuse University
  Identity Authentication                       Starkweather, Gary
                                                Microsoft Corporation
Sengupta, Kuntal                                  Laser Printer
Advanced Interfaces
  Augmented Reality                             Starner, Thad
                                                Georgia Institute of Technology
Sharma, Rajeev                                    Wearable Computers
Advanced Interfaces
  Augmented Reality                             Stephenson, Robert S.
  Multimodal Interfaces                         Wayne State University
                                                   Online Education
Shen, Yuan Yuan
Liverpool John Moores University                Sterner, William H.
  Animation                                     University of Chicago
                                                   Eliza
Simpson, Richard C.
University of Pittsburgh                        Tekalp, A. Murat
  User Modeling                                 University of Rochester
                                                  Video Summarization
Singh, Rita
Carnegie Mellon University                      Tenopir, Carol
  Sphinx                                        University of Tennessee
                                                  Electronic Journals
Soergel, Dagobert
University of Maryland                          Thomas, Douglas
  Information Organization                      University of Southern California
  Information Retrieval                           Hackers

St. Amant, Robert A.                            Thuraisingham, Bhavani
North Carolina State University                 National Science Foundation
   Artificial Intelligence                        Security
   Statistical Analysis Support                   Semantic Web

                                                Tran, Phuoc
                                                University of Toronto
                                                  Internet — Worldwide Diffusion
XXXII ❘❙❚ BERKSHIRE ENCYCLOPEDIA OF HUMAN-COMPUTER INTERACTION



Tretiakoff, Oleg                               Westfall, Ralph David
C.A. Technology, Inc.                          California State Polytechnic University, Pomona
  Braille                                        Telecommuting

Tygar, J. D.                                   Williams, Ronald D.
University of California, Berkeley             University of Virginia
  Digital Cash                                   Embedded Systems
  Spamming
  Viruses                                      Wilson, Michael
                                               CCLRC Rutherford Appleton Laboratory
Unicode Editorial Committee                      World Wide Web
  Unicode
                                               Witte, James
van Santen, Jan P.H.                           Clemson University
Oregon Health and Science University             Online Questionnaires
  Speech Synthesis
                                               Yaar, Abraham
Vanderheiden, Gregg                            Carnegie Mellon University
University of Wisconsin, Madison                 Denial of Service Attack
  Universal Access
                                               Youngblood, Michael
Vassileva, Julita                              University of Texas, Arlington
University of Saskatchewan                       Smart Homes
  Peer-to-Peer Architecture
                                               Zakaria, Norhayati
Wellman, Barry                                 Syracuse University
University of Toronto                            E-business
  Internet - Worldwide Diffusion
  Internet in Everyday Life                    Zaki, Mohammad
                                               Rensselaer Polytechnic Institute
Welty, Christopher A.                            Data Mining
IBM T.J. Watson Research Center
  Ontology                                     Zhai, Shumin
                                               IBM Almaden Research Center
                                                 Mouse
                                                                             INTRODUCTION

                                                                 By William Sims Bainbridge




In hardly more than half a century, computers have      hitherto existed only in the pages of science fic-
become integral parts of everyday life, at home,        tion. For a sense of the wide reach of HCI, consider
work, and play. Today, computers affect almost          the following vignettes:
every aspect of modern life, in areas as diverse as
car design, filmmaking, disability services, and sex     ■ Gloria, who owns a small fitness training busi-
education. Human-computer interaction (HCI) is             ness, is currently trying out a new system in which
a vital new field that examines the ways in which          she and a client dance on sensor pads on the floor,
people communicate with computers, robots, in-             while the computer plays rhythms and scores
formation systems, and the Internet. It draws upon         how quickly they are placing their feet on the
several branches of social, behavioral, and infor-         designated squares.
mation science, as well as on computer science and       ■ Elizabeth has made friends through chatrooms
electrical engineering. The traditional heart of HCI       connected to French and British music groups
has been user interface design, but in recent              that are not well known in the United States. She
years the field has expanded to include any science        occasionally shares music files with these friends
and technology related to the ways that humans             before buying CDs from foreign online distrib-
use or are affected by computing technology.               utors, and she has helped one of the French bands
HCI brings to the fore social and ethical issues that      translate its website into English.


                                                                                                        xxxiii
XXXIV ❘❙❚ BERKSHIRE ENCYCLOPEDIA OF HUMAN-COMPUTER INTERACTION



  ■ Carl’s work team develops drivers for new color         puter recognition of pen or stylus movements on
    printers far more quickly and effectively than be-      tablet or pocket computers.
    fore, because the team comprises expert design-              All of these have been very active areas of research
    ers and programmers who live in different               or development since he wrote, and several are fun-
    time zones around the world, from India to              damental to commercial products that have already
    California, collectively working 24 hours a day,        appeared. For example, many companies now use
    7 days a week, by means of an Internet-based col-       speech recognition to automate their telephone in-
    laboration system.                                      formation services, and hundreds of thousands of
  ■ Bella is blind, but her wearable computer uses          people use stylus-controlled pocket computers every
    Internet and the Global Positioning System not          day. Many articles in the encyclopedia describe
    only to find her way through the city safely but        new approaches that may be of tremendous impor-
    also to find any product or service she needs at        tance in the future.
    the best price and to be constantly aware of her             Our entire perspective on HCI has been evolving
    surroundings.                                           rapidly in recent years. In 1997, the National Research
  ■ Anderson, whose Internet moniker is Neo, dis-           Council—a private, nonprofit institution that pro-
    covers that his entire life is an illusion, main-       vides science, technology, and health policy advice
    tained by a vast computer plugged directly into         under a congressional charter—issued a major re-
    his nervous system.                                     port, More Than Screen Deep, “to evaluate and sug-
                                                            gest fruitful directions for progress in user interfaces
     The first three stories are real, although the names   to computing and communications systems.” This
are pseudonyms, and the scenarios are duplicated            high-level study, sponsored by the National Science
millions of times in the modern world of personal           Foundation (NSF), concluded with three recom-
computers, office automation, and the World Wide            mendations to the federal government and univer-
Web. The fourth example could be realized with to-          sity researchers.
day’s technology, simply given a sufficient investment
                                                              1. Break away from 1960s technologies and para-
in infrastructure. Not only would it revolutionize the
                                                                 digms. Major attempts should be made to find
lives of blind people like Bella, it would benefit the
                                                                 new paradigms for human-machine interac-
sighted public too, so we can predict that it will in
                                                                 tion that employ new modes and media for in-
fact become true over the next decade or two. The
                                                                 p u t a n d o u t p u t a n d t h a t i nv o l v e n e w
story about Mr. Anderson is pure fiction, no doubt
                                                                 conceptualizations of application interfaces.
recognizable to many as the premise of the 1999 film
                                                                 (192)
The Matrix. It is doubtful that HCI ever could (or
                                                              2. Invest in the research required to provide the com-
should) become indistinguishable from real life.
                                                                 ponent subsystems needed for every-citizen in-
                                                                 terfaces. Research is needed that is aimed at both
                                                                 making technological advances and gaining
Background on HCI                                                understanding of the human and organizational
In a brief history of HCI technology published in
                                                                 capabilities these advances would support. (195)
1996, the computer scientist Brad Myers noted
                                                              3. Encourage research on systems-level design and
that most computer interface technology began as
                                                                 development of human-machine interfaces that
government-supported research projects in univer-
                                                                 support multiperson, multimachine groups
sities and only years later was developed by corpo-
                                                                 as well as individuals. (196)
rations and transformed into commercial products.
He then listed six up-and-coming research areas:                In 2002, John M. Carroll looked back on the his-
natural language and speech, computer-supported             tory of HCI and noted how difficult it was at first to
cooperative work, virtual and augmented reality,            get computer science and engineering to pay atten-
three-dimensional graphics, multimedia, and com-            tion to issues of hardware and software usability. He
                                                                                         INTRODUCTION ❚❙❘ XXXV



argued that HCI was born as the fusion of four fields       technological revolution is likely to give computer
(software engineering, software human factors, com-         technology an additional powerful boost: nanotech-
puter graphics, and cognitive science) and that it con-     nology. The word comes from a unit for measuring
tinues to be an emerging area in computer science.          tiny distances, the nanometer, which is one billionth
The field is expanding in both scope and importance.        of a meter (one millionth of a millimeter, or one mil-
For example, HCI incorporates more and more from            lionth the thickness of a U.S. dime). The very
the social sciences as computing becomes increas-           largest single atoms are just under a nanometer in
ingly deeply rooted in cooperative work and human           size, and much of the action in chemistry (including
communication.                                              fundamental biological processes) occurs in the range
     Many universities now have research groups             between 1 nanometer and 100–200 nanometers. The
and training programs in HCI. In addition to the de-        smallest transistors in experimental computer
signers and engineers who create computer interfaces        chips are about 50 nanometers across.
and the researchers in industry and academia who are             Experts working at the interface between nano-
developing the fundamental principles for success           technology and computing believe that nanoelec-
in such work, a very large number of workers in many        tronics can support continued rapid improvements
industries contribute indirectly to progress in HCI.        in computer speed, memory, and cost for twenty
The nature of computing is constantly changing. The         to thirty years, with the possibility of further progress
first digital electronic computers, such as ENIAC (com-     after then by means of integrated design approaches
pleted in 1946), were built to solve military problems,     and investment in information infrastructure. Two
such as calculating ballistic trajectories. The 1950s and   decades of improvement in computer chips would
1960s saw a great expansion in military uses and ex-        mean that a desktop personal computer bought in
tensive application of digital computers in commerce        2024 might have eight thousand times the power
and industry. In the late 1970s, personal computers         of one bought in 2004 for the same price—or could
entered the home, and in the 1980s they developed           have the same power but cost only twenty cents and
more user-friendly interfaces. The 1990s saw the trans-     fit inside a shirt button. Already, nanotechnology
formation of Internet into a major medium of com-           is being used to create networks of sensors that can
munications, culminating in the expansion of the            detect and identify chemical pollutants or biologi-
World Wide Web to reach a billion people.                   cal agents almost instantly. While this technology
     In the first decade of the twenty-first century,       will first be applied to military defense, it can be
two trends are rushing rapidly forward. One is the          adapted to medical or personal uses in just a few years.
extension of networking to mobile computers and                  The average person’s wristwatch in 2024 could
embedded devices literally everywhere. The other is         be their mobile computer, telling them everything
the convergence of all mass media with computing,           they might want to know about their environment—
such that people listen to music, watch movies, take        where the nearest Thai restaurant can be found, when
pictures, make videos, carry on telephone conversa-         the next bus will arrive at the corner up the road,
tions, and conduct many kinds of business on com-           whether there is anything in the air the person hap-
puters or on networks of which computers are central        pens to be allergic to, and, of course, providing any
components. To people who are uncomfortable with            information from the world’s entire database that
these trends, it may seem that cyberspace is swal-          the person might want to know. If advances in nat-
lowing real life. To enthusiasts of the technology, it      ural-language processing continue at the rate they
seems that human consciousness is expanding to en-          are progressing today, then the wristwatch could also
compass everything.                                         be a universal translator that allows the person to
     The computer revolution is almost certainly            speak with anyone in any language spoken on the
going to continue for decades, and specialists in           face of the planet. Of course, predictions are al-
human-computer interaction will face many new               ways perilous, and it may be that progress will slow
challenges in the years to come. At least one other         down. Progress does not simply happen of its own
XXXVI ❘❙❚ BERKSHIRE ENCYCLOPEDIA OF HUMAN-COMPUTER INTERACTION



accord, and the field of human-computer interac-              and data. After a while I realized I was going deaf
tion must continue to grow and flourish if comput-            from the noise and took to wearing earplugs. Later,
ers are to bring the marvelous benefits to human life         back at Harvard in a faculty position, I began writ-
that they have the potential to bring.                        ing my own statistical analysis programs for my first
                                                              personal computer, an Apple II. I remember that one
                                                              kind of analysis would take a 36 hours to run, with
My Own Experience with Computers                              the computer humming away in a corner as I went
Computer and information technologies have pro-               about my daily life. For a decade beginning in 1983,
gressed amazingly over the past fifty years, and they         I programmed educational software packages in so-
may continue to do so for the next half century. My first     ciology and psychology, and after a series of com-
computer, if it deserves that word, was a Geniac I re-        puter-related projects found myself running the
ceived for my sixteenth birthday in 1956. Costing only        sociology program at the National Science Founda-
$20, it consisted of masonite disks, wires, light bulbs       tion and representing the social and behavioral
and a vast collection of nuts, bolts, and clips. From these   sciences on the major computing initiatives of
parts I could assemble six rotary switches that could be      NSF and the federal government more generally.
programmed (by hardwiring them) to solve simple               After eight years of that experience, I moved to the
logic problems such as playing tick-tack-toe. I devel-        NSF Directorate for Computer and Information
oped a great affection for the Geniac, as I did for the       Science and Engineering to run the NSF’s programs
foot-long slide rule I lugged to my high school               in human-computer interaction, universal access,
classes, but each was a very far cry from the pocket com-     and artificial intelligence and cognitive science be-
puter or even the programmable calculator my sixteen-         fore becoming deputy director of the Division of
year-old daughter carries in her backpack today.              Information and Intelligent Systems, which contains
     Geniac was not really an electronic computer be-         these programs.
cause it lacked active components—which in 1956                    My daughters, aged sixteen and thirteen, have
meant relays or vacuum tubes, because transistors             used their considerable computer expertise to cre-
were still very new and integrated circuits had not           ate the Center for Glitch Studies, a research project
yet been invented. The first real computer I saw, in          to discover and analyze programming errors in com-
the early 1960s, was the massive machine used by my           mercial video games. So far they have documented
father’s company, Equitable Life Insurance, to keep           on their website more than 230 programming errors
its records. Only decades later did I learn that my           in popular video games. The hundreds of people who
uncle, Angus McIntosh, had been part of a team in             visit the website are not a passive audience, but send
World War II that seized the German computer that             e-mail messages describing errors they themselves
was cracking Soviet codes, and that the secret Colossus       discovered, and they link their own websites into a
computer at Bletchley Park where he worked had                growing network of knowledge and virtual social
been cracking German codes. In the middle of the              relationships.
twentieth century, computers were huge, rare, and
isolated from the general public, whereas at the be-
ginning of the twenty-first century they are essen-           A Personal Story—NSF’s FastLane
tial parts of everyday life.                                  Computers have become vastly more important at
     My first experience programming computers                work over recent decades, and they have come to play
came in 1974, when I was a graduate student in the            increasingly more complex roles. For example,
sociology department at Harvard University, and I             NSF has created an entire online system for re-
began using the machines for statistical analysis of          viewing grant proposals, called FastLane, and thou-
data. Starting the next year at the University of             sands of scientists and educators have become
Washington, where I was a beginning assistant                 familiar with it through serving as reviewers or prin-
professor, I would sit for hours at a noisy keypunch          cipal investigators.
machine, making the punch cards to enter programs
                                                                                      INTRODUCTION ❚❙❘ XXXVII



     A researcher prepares a description of the proj-      the institution, and the abstract is posted on the web
ect he or she hopes to do and assembles ancillary          for anyone to see. Each year, the researcher submits
information such as a bibliography and brief biog-         a report, electronically of course, and the full record
raphies of the team members. The researcher sub-           of the grant accumulates in the NSF computer sys-
mits this material, along with data such as the dollar     tem until the work has been completed.
requests on the different lines of the formal budget.          Electronic systems connect the people—
The only software required is a word processor             researcher, program director, and reviewers—into a
and a web browser. As soon as the head of the in-          system of information flow that is also a social sys-
stitution’s grants office clicks the submit button, the    tem in which each person plays a specific role. Be-
full proposal appears at NSF, with the data already        cause the system was designed over a number of years
arranged in the appropriate data fields, so nobody         to do a particular set of jobs, it works quite well, and
has to key it in.                                          improvements are constantly being incorporated.
     Peer review is the heart of the evaluation process.   This is a prime example of Computer-Supported
As director of the HCI program, I categorize pro-          Cooperative Work, one of the many HCI topics cov-
posals into review panels, then recruit panelists          ered in this encyclopedia.
who were experts in the field with specializations that
matched the scope of the proposals. Each panelist re-
views certain proposals and submits a written review       The Role of the Berkshire Encyclopedia
electronically.
     Once the individual reviews have been submit-
                                                           of Human-Computer Interaction
                                                           Because the field of HCI is new, the Berkshire
ted, the panel meets face-to-face to discuss the
                                                           Encyclopedia of Human-Computer Interaction breaks
proposals and recommend funding for the best ones.
                                                           new ground. It offers readers up-to-date information
The panelists all have computers with Electronic
                                                           about several key aspects of the technology and its
Panel System (EPS) groupware that provides easy ac-
                                                           human dimensions, including
cess to all the proposals and reviews associated with
the particular panel. During the discussion of a par-        ■ applications—major tools that serve human
ticular proposal, one panelist acts as “scribe,” keep-         needs in particular ways, with distinctive usability
ing a summary of what was said in the EPS. Other               issues.
panelists can read the summary, send written com-            ■ approaches—techniques through which scien-
ments to the scribe, and may be asked to approve the           tists and engineers design and evaluate HCI.
final draft online.                                          ■ breakthroughs—particular projects that marked
     Next the NSF program officer combines all the             a turning point in the history of HCI.
evaluations and writes a recommendation in the elec-         ■ challenges—problems and solutions, both tech-
tronic system, for approval by the director of the di-         nical and human, especially in controversial
vision in which the program is located. More often             areas.
than not, unfortunately, the decision is to decline to       ■ components—key parts of a software or hard-
fund the proposal. In that case, the program officer           ware system that are central to how people use it.
and division director processes the action quickly on        ■ disciplines—the contributions that various sci-
their networked computers, and an electronic no-               ences and academic fields make to HCI.
tification goes immediately to the principal inves-          ■ interfaces—hardware or software systems that
tigator, who can access FastLane to read the reviews           mediate between people and machines.
and summary of the panel discussion.                         ■ methods—general computer and information
     In those rarer and happier situations when a              science solutions to wide classes of technical
grant is awarded, the principal investigator and pro-          problems.
gram officer negotiate the last details and craft an         ■ social implications—technological impacts on so-
abstract, describing the research. The instant the             ciety and policy issues, and the potential of multi-
award is made, the money goes electronically to                user HCI systems to bring about social change.
XXXVIII ❘❙❚ BERKSHIRE ENCYCLOPEDIA OF HUMAN-COMPUTER INTERACTION



    These categories are not mutually exclusive; many     of HCI. I have written occasional encyclopedia arti-
articles fit in two or more of them. For example, the     cles since the early 1990s, when I was one of sev-
short article on laser printers concerns an output in-    eral subject matter editors of The Encyclopedia of
terface and explains how a laser printer puts words       Language and Linguistics. Often, an editor working
and pictures on paper. But this article also concerns     on a specialized encyclopedia for one publisher or
a breakthrough, the actual invention of the laser         another would send me an e-mail message asking
printer, and it was written by the inventor himself,      if I would write a particular essay, and I would
Gary Starkweather.                                        send it in, also by e-mail. I had a very good experi-
                                                          ence contributing to the Encyclopedia of Community,
Contributors                                              edited by Karen Christensen and David Levinson
The 175 contributors to the encyclopedia possess the      of Berkshire Publishing. I suggested to Karen that
full range and depth of expertise covered by HCI,         Berkshire might want to do an encyclopedia of
and more. They include not only computer scien-           human-computer interaction and that I could re-
tists and electrical engineers, but also social and       cruit excellent authors for such a project. Berkshire
behavioral scientists, plus practicing engineers, sci-    has extensive experience developing high-quality ref-
entists, scholars, and other experts in a wide range      erence works, both in partnership with other pub-
of other fields. The oldest authors were born around      lishing houses and on its own.
the time that the very first experimental digital elec-        Almost all the communication to create the
tronic computer was built, and the entire history         encyclopedia was carried out online. Although I know
of computing has taken place during their lives.          many people in the field personally, it was a great
    Among the influential and widely respected con-       help to have access to the public databases placed on
tributors is Jose-Marie Griffiths, who contributed        the Web by NSF, including abstracts of all grants
the article on digital libraries. As a member of the      made in the past fifteen years, and to the online pub-
U.S. President’s Information Technology Advisory          lications of organizations such as the Association for
Committee, Griffiths understands the full scope and       Computing Machinery and to the websites of all of
social value of this new kind of public resource.         the authors, which often provide copies of their pub-
Contributors Judith S. Olson, Gary M. Olson, and          lications. Berkshire created a special password-
John M. Carroll are among the very few leaders who        protected website with information for authors and
have been elected to the Academy of the Special           a section where I could review all the essays as they
Interest Group on Computer-Human Interaction of           were submitted.
the Association for Computing Machinery (SIGCHI).
In 2003 Carroll received the organization’s Life-         For the Reader
time Achievement Award for his extensive accomplish-      There are many challenges ahead for HCI, and many
ments, including his contributions to the Blacksburg      are described in this encyclopedia. Difficult prob-
Electronic Village, the most significant experiment       lems tend to have both technical and human aspects.
on community participation in computer-mediated           For the benefit of the reader, the articles identify stan-
communication. Jack Dongarra, who wrote the con-          dard solutions and their ramifications, both positive
tribution on supercomputers, developed the                and negative, and may also cover social or political
LINPACK Benchmark, which is used to test the speed        controversies surrounding the problem and its pos-
of these upper-end machines and which is the              sible solutions. Many of the articles describe how a
basis of the annual list of the five hundred fastest      particular scientific discipline or branch of engi-
computers in the world.                                   neering approaches HCI, and what it contributes to
                                                          the multidisciplinary understanding of and im-
Building the Encyclopedia:                                provement in how computers, robots, and informa-
Computer-Supported Cooperative Work                       tion systems can serve human needs. Other articles
The creation of this encyclopedia is an example of        focus on a particular interface, modality, or medium
computer-supported cooperative work, a main area          in which people receive information and control the
                                                                                                 INTRODUCTION ❚❙❘ XXXIX



computer or system of which it is a part. These articles         ■ Some seventy-five diverse illustrations, which range
explain the technical features of the hardware or soft-            from “antique” photos of the ENIAC computer (c.
ware; they also explain the way humans perceive,                   1940s) to cutting-edge computerized images.
learn, and behave in the particular context. Still other         ■ A bibliography of HCI books and journal
articles concern how computer and information sci-                 articles.
ence has developed to solve a wide class of problems,            ■ A popular culture appendix that includes more
using vivid examples to explain the philosophy of                  than 300 annotated entries on books, plays,
the method, paying some attention as well to the hu-               movies, television shows, and songs that have
man side of the equation.                                          connections to HCI.
      Many articles—sometimes as their central focus
and sometimes incidentally—examine the social im-                                                 William Sims Bainbridge
plications of HCI, such as the impact of a particular
kind of technology, the way that the technology                The views expressed are those of the author and do not
fits into societal institutions, or a social issue involving   necessarily reflect the position of the National Science
computing. The technology can strengthen either                Foundation
cooperation or conflict between human beings, and
the mutual relations between technological change
and social change are often quite complex.                     FURTHER READING
      For information technology workers, this ency-
clopedia provides insight into specialties other than          Asher, R. E., & Simpson, J. M. Y. (Eds.). (1994). The encyclopedia of
the one they work in and offers useful perspectives on            language and linguistics. Oxford, UK: Pergamon.
the broad field. For policy makers, it provides a basis        Bainbridge, W. S. (1989). Survey research: A computer-assisted intro-
                                                                  duction. Belmont, CA: Wadsworth.
for thinking about the decisions we face in exploit-           Bainbridge, W. S. (1992). Social research methods and statistics: A
ing technological possibilities for maximum human                 computer-assisted introduction. Belmont, CA: Wadsworth.
benefit. For students, this encyclopedia lays out how          Carroll, J. M. (Ed.). (2002). Human-computer interaction in the new
                                                                  millennium. Boston: Addison-Wesley.
to use the technology to make a better world and of-           Christensen, K., & Levinson, D. (2003). Encyclopedia of community:
fers a glimpse of the rapidly changing computer-as-               From the village to the virtual world. Thousand Oaks, CA: Sage.
sisted human world in which they are living their lives.       Myers, B. A. (1996). A brief history of human computer interaction
      To illuminate and expand on the articles them-              technology. ACM Interactions, 5(2), 44–54.
                                                               National Research Council. (1997). More than screen deep. Washington,
selves, the encyclopedia includes the following spe-              DC: National Academy Press.
cial features:                                                 Roco, M. C., & Bainbridge, W. S. (2001). Societal implications of
                                                                  nanoscience and nanotechnology. Dordrecht, Netherlands: Kluwer.
  ■   Approximately eighty sidebars with key primary           Roco, M. C., & Bainbridge, W. S. (2003). Converging technologies for
      text, glossary terms, quotes, and personal stories          improving human performance. Dordrecht, Netherlands: Kluwer.
      about how HCI has had an impact on the work
      and lives of professionals in the field.
                                                                       PUBLISHER’S NOTE

                                                                              By Karen Christensen




The Berkshire Encyclopedia of Human-Computer              relationships than in binary code; but it was books—
Interaction (HCI) is our first independent title. We’ve   and a career in publishing—that at last brought home
done many other award-winning encyclopedias but           to me that computers can support and expand hu-
HCI will always have a unique place in our hearts         man connections and improve our lives in myriad
and in our history.                                       ways. Berkshire Publishing Group, based in a tiny
    Even though most of our work has been in the          New England town, depends on human-computer
social sciences, when William Bainbridge at the           interaction to maintain working relationships, and
National Science Foundation wrote to suggest the          friendships too, with many thousands of experts
topic of HCI, I knew instantly that it was the right      around the world. We are convinced, in fact, that this
topic for our “knowledge and technology” company.         topic is central to our development as a twenty-first
I grew up with the computer industry. My father, a        century publishing company,
computer engineer in the Silicon Valley, tried very           The Berkshire Encyclopedia of Human-Computer
hard to explain the fundamentals of computing, and        Interaction takes computing into new realms, intro-
even built a machine out of plywood and blinking          ducing us to topics that are intriguing both in their
lights to show my sixth-grade class that information      technical complexity and because they present us—
can be captured and communicated with nothing             human beings—with a set of challenging questions
more than a combination of on-off switches. I was         about our relationship with “thinking”machines. There
a reader, much more interested in human stories and       are opportunities and risks in any new technology, and

                                                                                                             xli
XLII ❘❙❚ BERKSHIRE ENCYCLOPEDIA OF HUMAN-COMPUTER INTERACTION



HCI has intrigued writers for many decades because                 The Berkshire Encyclopedia of Human-Computer
it leads us to a central philosophical, religious, and even   Interaction provides us with an essential grounding
historical question: What does it mean to be human?           in the most relevant and intimate form of technol-
We’ll be exploring this topic and related ones in fur-        ogy, making scientific and technological research
ther works about technology and society.                      available to a wide audience. This topic and other as-
     Bill Bainbridge was an exceptional editor: or-           pects of what Bill Bainbridge likes to refer to as “con-
ganized, focused, and responsive. Working with him            verging technologies” will continue to be a core part
has been deeply rewarding, and it’s no surprise               of our print and online publishing program. And, as
that the hundreds of computer scientists and engi-            befits a project so closely tied to electronic tech-
neers he helped us recruit to contribute to the en-           n o l o g y, a n o n l i n e ve r s i o n o f t h e B e r k s h i re
cyclopedia were similarly enthusiastic and gracious.          Encyclopedia of Human-Computer Interaction will be
All these experts—computer scientists and engineers           available through xrefplus. For more information,
as well as people working in other aspects of HCI—            visit www.xreferplus.com.
truly wanted to work with us to ensure that their
work would be accessible and understandable.                                                        Karen Christensen
     To add even greater interest and richness to the                                 CEO, Berkshire Publishing Group
work, we’ve added dozens of photographs, personal                                      karen@berkshirepublishing.com
stories, glossary terms, and other sidebars. In addi-
tion to article bibliographies, there is a master bib-
liography at the end, containing all 2,590 entries in         Editor’s Acknowledgements
the entire encyclopedia listed together for easy ref-         Karen Christensen, cofounder of the Berkshire
erence. And we’ve added a characteristic Berkshire            Publishing Group, deserves both thanks and
touch, an appendix designed to appeal to even the             praise for recognizing that the time had come when
most resolute Luddite: “HCI in Popular Culture,” a            a comprehensive reference work about human re-
database compilation listing with 300 sci-fi novels,          lations with computing systems was both possible
nonfiction titles, television programs and films from         and sorely needed. Courtney Linehan at Berkshire
The Six-Million Dollar Man to The Matrix (per-                was both skilled and tireless in working with the au-
haps the quintessential HCI story), and even a hand-          thors, editor, and copyeditors to complete a mar-
ful of plays and songs about computers and                    velous collection of articles that are technically
technology.                                                   accurate while communicating clearly to a broad
     The encyclopedia has enabled us to develop a             public. At various stages in the process of develop-
network of experts as well as a cutting-edge resource         ing the encyclopedia, Marcy Ross and George
that will help us to meet the needs of students,              Woodward at Berkshire made their own indispen-
professionals, and scholars in many disciplines. Many         sable contributions. Among the authors, Mary
articles will be of considerable interest and value to        Harper, Bhavani Thuraisingham, and Barry Wellman
librarians—Digital Libraries, Information Filtering,          were unstinting in their insightful advice. I would
Information Retrieval, Lexicon Building, and much             particularly like to thank Michael Lesk who, as di-
more—and even to publishers. For example, we have             rector of the Division of Information and Intelligent
an article on “Text Summarization” written by Judith          Systems of the National Science Foundation, gave
Klavans, Director of Research at the Center for               me the opportunity to gain invaluable experience
Advanced Study of Language, University of                     managing the grant programs in Universal Access
Maryland. “Summarization is a technique for                   and Human-Computer Interaction.
identifying the key points of a document or set of
related documents, and presenting these selected                                          William Sims Bainbridge
points as a brief, integrated independent represen-                                               Deputy Director,
tation” and is essential to electronic publishing, a key            Division of Information and Intelligent Systems
aspect of publishing today and in the future.                                         National Science Foundation
                                                                  ABOUT THE EDITOR




William Sims Bainbridge is deputy director of the    resented the social and behavioral sciences on five ad-
Division of Information and Intelligent Systems of   vanced technology initiatives: High Performance
the National Science Foundation, after having di-    Computing and Communications, Knowledge and
rected the division’s Human-Computer Interaction,    Distributed Intelligence, Digital Libraries, Information
Universal Access, and Knowledge and Cognitive        Technology Research, and Nanotechnology.
Systems programs. He coedited Converging Tech-           Bill Bainbr idge is also the author of ten
nologies to Improve Human Performance, which         books, four textbook-software packages, and some
explores the combination of nanotechnology, bio-     150 shorter publications in information science,
technology, information technology, and cognitive    social science of technology, and the sociology of
science (National Science Foundation, 2002;          culture. He earned his doctorate from Harvard
www.wtec.org/ConvergingTechnologies). He has rep-    University.




                                                                                                        xliv
                                                                                   ADAPTIVE HELP SYSTEMS

                                                                                     ADAPTIVE INTERFACES

                                                                                     AFFECTIVE COMPUTING

                                                                                                   ALTAIR




            A
                                                                                                     ALTO

                                                                                               ANIMATION

                                                                                   ANTHROPOLOGY AND HCI

                                                                                         ANTHROPOMETRY

                                                                               APPLICATION USE STRATEGIES

                                                                                                 ARPANET

                                                                                   ARTIFICIAL INTELLIGENCE

                                                                                       ASIAN SCRIPT INPUT

                                                                           THE ATANASOFF-BERRY COMPUTER

                                                                                 ATTENTIVE USER INTERFACE

                                                                                    AUGMENTED COGNITION

                                                                                      AUGMENTED REALITY

                                                                                                 AVATARS




                                                                 The first wave of research on adaptive help emerged
          ADAPTIVE HELP                                     in early 1980 when the UNIX system—due to its low
                                                            cost and efficiency—reached many universities whose
            SYSTEMS                                         users lacked the advanced technical training (such as
                                                            knowledge of complicated commands) needed to op-
Adaptive help systems (AHSs; also called intelligent        erate UNIX. Early work on adaptive and intelligent
help systems) are a specific kind of help system and a       help systems focused almost exclusively on UNIX and
recognized area of research in the fields of artificial in-   its utilities, such as text editors and e-mail. From 1980
telligence and human-computer interaction. The goal         to 1995 this research direction involved more than a
of an adaptive help system is to provide personalized       hundred researchers working on at least two-dozen
help to users working with complex interfaces, from         projects. The most representative projects of this gen-
operating systems (such as UNIX) to popular appli-          eration were UNIX Consultant and EUROHELP. The
cations (such as Microsoft Excel). Unlike traditional       widespread use of graphical user interfaces (GUIs)
static help systems that serve by request the same in-      in early 1990 caused a pause in AHS research, because
formation to different users, AHSs attempt to adapt to      GUIs resolved a number of the problems that the early
the knowledge and goals of individual users, offering       generation of AHS sought to address. In just a few years,
the most relevant information in the most relevant          however, GUIs reached the level of complexity
way.                                                        where adaptive help again became important, giving

                                                                                                                   1
2 ❘❙❚ BERKSHIRE ENCYCLOPEDIA OF HUMAN-COMPUTER INTERACTION



                                                                 Many AHSs use two classic approaches to model
  INTERFACE Interconnections between a device, program,      the user. First, they track the user’s actions to under-
  or person that facilitate interaction.                     stand which commands and concepts the user
                                                             knows and which are not known, and second, they use
                                                             task models to deduce the user’s current goal and miss-
rise to a second wave of research on AHSs. Lumière,          ing knowledge. The first technology is reasonably sim-
the most well known project of this wave, intro-             ple: The system just records all used commands and
duced the idea of intelligent help to millions of users      parameters, assuming that if a command is used, it
of Microsoft applications.                                   must be known. The second is based on plan recog-
                                                             nition and advanced domain knowledge representa-
                                                             tion in such forms as a goal-plan-action tree. To identify
Active and Passive AHSs                                      the current goal and missing pieces of knowledge, the
Adaptive help systems are traditionally divided into         system first infers the user’s goal from an observed se-
two classes: active and passive. In a passive AHS, the
user initiates the help session by asking for help. An
active help system initiates the help session itself. Both                         Farewell Clippy
kinds of AHSs have to solve three challenging prob-
lems: They must build a model of user goals and knowl-

                                                               M
                                                                         any PC users through the years quickly learned
edge, they must decide what to present in the next help                  how to turn off “Clippy,” the Microsoft Office
message, and they must decide how to present it. In                      helper who appeared out of nowhere eagerly hop-
addition, active AHSs also need to decide when to              ing to offer advice to the baffled. The Microsoft press re-
intervene with adaptive help.                                  lease below was Clippy’s swan song.

User Modeling                                                          REDMOND, Wash., April 11, 2001—Whether
                                                                       you love him or you hate him, say farewell to
To be useful, a help message has to present informa-                   Clippy automatically popping up on your screen.
tion that is new to the user and relevant to the user‘s                     Clippy is the little paperclip with the soulful
current goal. To determine what is new and relevant,                   eyes and the Groucho eyebrows. The electronic
AHSs track the user’s goals and the user’s knowledge                   ham who politely offers hints for using Microsoft
about the interface and maintain a user model. Two                     Office software.
major approaches to user modeling in AHSs are “ask                          But, after four years on-screen, Clippy will
                                                                       lose his starring role when Microsoft Office XP
the user” and “observe the user.” Most passive AHSs                    debuts on May 31. Clippy, the Office Assistant in-
have exploited the first of these approaches. UNIX                      troduced in Office 97, has been demoted in Office
Consultant demonstrates that a passive AHS can be                      XP. The wiry little assistant is turned off by de-
fairly advanced: It involves users in a natural-language               fault in Office XP, but diehard supporters can
dialogue to discover their goals and degree of knowl-                  turn Clippy back on if they miss him.
edge and then provides the most relevant information.                       “Office XP is so easy to use that Clippy is
                                                                       no longer necessary, or useful,” explained Lisa
In contrast, active AHSs, introduced by the com-                       Gurry, a Microsoft product manager. “With new
puter scientist Gerhard Fischer in 1985, strive to de-                 features like smart tags and Task Panes, Office XP
duce a user’s goals by observing the user at work; they                enables people to get more out of the product
then strive to identify the lack of knowledge by de-                   than ever before. These new simplicity and ease-
tecting errors and suboptimal behavior. EURO-                          of-use improvements really make Clippy obso-
HELP provides a good example of an active help system                  lete,” she said.
                                                                            “He’s quite down in the dumps,” Gurry joked.
capable of identifying a knowledge gap and filling it                   “He has even started his own campaign to try
provocatively. In practical AHSs the two approaches                    to get his old job back, or find a new one.”
often coexist: The user model is initiated through a           Source: Microsoft. Retrieved March 10, 2004, from
short interview with the user and then kept updated            http://www.microsoft.com/presspass/features/2001/apr01/04-11clippy.asp

through observation.
                                                                                              ADAPTIVE INTERFACES ❚❙❘ 3



quence of commands. It then tries to find a more ef-         intelligence and HCI and has helped to establish re-
ficient (or simply correct) sequence of commands to          search on intelligent interfaces and user modeling. A
achieve this goal. Next, it identifies the aspects of the    treasury of knowledge accumulated by various AHS
interface that the user needs to know to build this se-     projects over the last thirty years is being used now
quence. These aspects are suspected to be unknown           to develop practical adaptive help and adaptive per-
and become the candidates to be presented in help           formance support systems.
messages.
                                                                                                              Peter Brusilovsky
Providing Adaptive Help:
Deciding What to Present and How                            See also Artificial Intelligence; Task Analysis; User
Deciding what should be the focus of the next help          Modeling
message is the most challenging job of an adaptive help
system. A number of passive AHSs simply avoid this
problem, allowing the users to determine what they          FURTHER READING
need and focusing on adaptive presentation only.
Classic AHSs, which use plan recognition, can deter-        Brusilovsky, P., Kobsa, A., & Vassileva, J. (Eds.). (1998). Adaptive hy-
                                                               pertext and hypermedia. Dordrecht, Netherlands: Kluwer.
mine quite precisely what the user needs, but this func-    Encarnação, L. M., & Stoev, S. L. (1999). An application-independent
tionality requires elaborate knowledge representation.         intelligent user support system exploiting action-sequence based
To bypass the knowledge representation barrier, mod-           user modeling. In J. Kay (Ed.), Proceedings of 7th International
ern practical AHSs use a range of alternative                  Conference on User Modeling, UM99, June 20–24, 1999 (pp.
                                                               245–254). Vienna: Springer.
(though less precise) technologies that are either sta-     Fischer, G. (2001). User modeling in human-computer interaction.
tistically or socially based. For example, Lumière used        User Modeling and User-Adapted Interaction, 11(1–2), 65–86.
a complex probabilistic network to connect ob-              Goodman, B. A., & Litman, D. J. (1992). On the interaction between
                                                               plan recognition and intelligent interfaces. User Modeling and User-
served user actions with available help interventions,         Adapted Interaction, 2(1), 83–115.
while the system developed by MITRE researchers             Hegner, S. J., Mc Kevitt, P., Norvig, P., & Wilensky, R. L. (Eds.). (2001).
Linton and Schaefer compared the skills of individual          Intelligent help systems for UNIX. Dordrecht, Netherlands: Kluwer.
users with a typical set of interface skills assembled by   Horvitz, E., Breese, J., Heckerman, D., Hovel, D., & Rommelse, K.
                                                               (1998). The Lumière project: Bayesian user modeling for inferring
observing multiple users.                                      the goals and needs of software users. In Proceedings of Fourteenth
     As soon as the focus of the next help message is          Conference on Uncertainty in Artificial Intelligence (pp. 256–265).
determined, the AHS has to decide how to present the           San Francisco: Morgan Kaufmann.
target content. While some AHSs ignore this part and        Linton, F., & Schaefer, H.-P. (2000). Recommender systems for learn-
                                                               ing: Building user and expert models through long-term obser-
focus solely on the selection part, it has been shown          vation of application use. User Modeling and User-Adapted
that adaptive presentation of help information can in-         Interaction, 10(2–3), 181–208.
crease the user’s comprehension speed and decrease          Oppermann, R. (Ed.). (1994). Adaptive user support: Ergonomic de-
                                                               sign of manually and automatically adaptable software. Hillsdale,
errors. Most often the content presentation is adapted         NJ: Lawrence Erlbaum Associates.
to the user’s knowledge, with, for example, expert users    Wilensky, R., Chin, D., Luria, M., Martin, J., Mayfield, J., & Wu, D.
receiving more specific details and novice users re-            (1988). The Berkeley UNIX Consultant project. Computational
ceiving more explanations. To present the adaptive             Linguistics, 14(4), 35–84.
                                                            Winkels, R. (1992). Explorations in intelligent tutoring systems and help.
content, classic AHSs that operated in a line-based            Amsterdam: IOS Press.
UNIX interface relied mostly on a natural language
generation approach. Modern AHSs operating in the
context of Graphical User Interfaces exploit adaptive
hypermedia techniques to present the content and links
to further information that is most suitable for the           ADAPTIVE INTERFACES
given user.
     Research into adaptive help systems has contributed    Computer interfaces are becoming ever richer in
to progress in a number of subfields within artificial        functionality, software systems are becoming more
4 ❘❙❚ BERKSHIRE ENCYCLOPEDIA OF HUMAN-COMPUTER INTERACTION



complex, and online information spaces are be-              perform certain indicative actions (such as entering
coming larger in size. On the other hand, the num-          certain keywords in search machines).
ber and diversity of people who use computer systems            User adaptability and adaptivity recently gained
are increasing as well. The vast majority of new users      strong popularity on the World Wide Web under the
are thereby not computer experts, but rather layper-        notion of “personalization.” This popularity is due
sons such as professionals in nontechnical areas, eld-      to the fact that the audiences of websites are often
erly people, and children. These users vary with            even less homogeneous than the user populations of
respect not only to their computer skills, but also         commercial software. Moreover, personalization has
to their fields of expertise, their tasks and goals, their   been recognized as an important instrument for on-
mood and motivation, and their intellectual and             line customer relationship management.
physical capabilities.
     The traditional strategy for enabling heteroge-
neous user groups to master the complexity and rich-        Acquiring Information about Users
ness of computers was to render computer interaction        To acquire the information about users that is needed
as simple as possible and thereby to cater to the low-      to cater to them, people can use several methods. A
est common denominator of all users. Increasingly,          simple way is to ask users directly, usually through an
though, developers are creating computer applica-           initial questionnaire. However, this questionnaire must
tions that can be “manually” customized to users’           be kept extremely short (usually to less than five ques-
needs by the users themselves or by an available ex-        tions) because users are generally reluctant to spend
pert. Other applications go beyond this capability.         efforts on work that is not directly related to their cur-
They are able within certain limits to recognize            rent tasks, even if this work would save them time in
user needs and to cater to them automatically.              the long run. In certain kinds of systems, specifically
Following the terminology of Reinhard Oppermann,            tutoring systems, user interviews can be clad in the
we will use the term adaptable for the manual type          form of quizzes or games. In the future, basic infor-
of application and adaptive for the automatic type.         mation about users may be available on smartcards,
                                                            that is, machine-readable plastic cards that users swipe
                                                            through a reading device before the beginning of a
Adaptable and Adaptive Systems                              computer session or that can even be read from a dis-
Adaptable systems are abundant. Most commercial             tance as users approach a computer terminal.
software allows users to modify system parameters                Various methods draw assumptions about
and to indicate individual preferences. Web portals         users based on their interaction behavior. These
permit users to specify the information they want to        methods include simple rules that predict user
see (such as stock quotes or news types) and the form       characteristics or assign users to predetermined
in which it should be displayed by their web browsers.      user groups with known characteristics when cer-
Web shops can store basic information about their           tain user actions are being observed (the latter
customers, such as payment and shipping data, past          method is generally known as the “stereotype
purchases, wish lists for future purchases, and birth-      approach” to user modeling). Probabilistic reason-
dates of friends and family to facilitate transac-          ing methods take uncertainty and evidences from
tions online. In contrast, adaptive systems are still       different sources into account. Plan recogni-
quite rare. Some shopping websites give purchase            tion methods aim at linking individual actions of
recommendations to customers that take into ac-             users to presumable underlying plans and goals.
count what these customers bought in the past.              Machine-learning methods try to detect regu-
Commercial learning software for high school math-          larities in users’ actions (and to use the learned
ematics adapts its teaching strategies to the presumed      patterns as a basis for predicting future actions).
level of expertise of each student. Advertisements on       Clique-based (collaborative) filtering methods
mobile devices are already being targeted to users in       determine those users who are closest to the cur-
certain geographical locations only or to users who         rent user in an n-dimensional attribute space and
                                                                                                          ADAPTIVE INTERFACES ❚❙❘ 5




                            Keeping Disabled People in the Technology Loop


         AUSTIN, Texas (ANS)—If communications technology is           been defined in recent years. While the proliferation of
         fueling the economy and social culture of the 21st century,   computers, home gadgets and gizmos is on the rise, it’s
         why should 18 percent of the population be left behind?       workers like Berger who make sure the disabled aren’t left
             Stephen Berger, a specialist in retrofitting the latest    out of the loop.
         computer and phone technology for the disabled, is                Other workers in the field, according to Berger, are
         trying to make sure they’re not.                              coming from educational institutions. For example, Neil
             From an office in Austin, Berger works to make sure        Scott and Charlie Robinson, from Stanford University
         that those with hearing and vision impairments or other       and Louisiana Tech University respectively, are working
         disabilities can benefit from the latest in Internet, cell    on the things the Hollywood movies are made of.
         phone and other technologies.                                     […]
             As a project manager at Siemens Information and               “Guys like this are breaking the barrier between the
         Communication Mobile, where he’s responsible for stan-        blind and computers,” he said.“(The blind) will soon have
         dards and regulatory management, Berger works to un-          an interface with no visual, just audio computer controls
         ravel such problems as why those who use hearing aids         with no touch, just head position and voice controls.”
         couldn’t use many brands of cell phones.                          Other devices, like the Home RF systems—that’s home
             “Some new cell phones make a buzz in hearing aids,”       radio frequency—link all the major appliances and
         Berger explained. “The Federal Communications                 electronics of the home together. That means tele-
         Commission took note and said it needed to be resolved.”      phone, Dolby sound, Internet, entertainment electronics
             But what was needed was either better technology or       and other devices are all connected into one wireless net-
         protocols that both the hearing impaired and the cell         work with voice control for those who aren’t mobile.
         phone companies could agree on. Berger helped deter-              “It’s microphones implanted in wallpaper, security
         mine what types of hearing aids work with certain             systems by voice, household appliances that work on a
         types of phones. The intelligence was passed around the       vocal command,” Berger said. “It’s what the movies are
         industry, and the problem is now minimal.                     made of and it’s here today.”
             Berger is one of the many technology specialists in           Source: Innovations keep disabled in the technology loop. American
                                                                           News Services, October 12, 2000.
         huge communications companies whose niche has



use them as predictors for unknown attributes of                       ■ Usage data, such as selections (e.g., of webpages
the current user. Clustering methods allow one to                        or help texts with certain content), temporal
generalize groups of users with similar behaviors                        viewing behavior (particularly “skipping” of
or characteristics and to generate user stereotypes.                     webpages or streaming media), user ratings (e.g.,
                                                                         regarding the usefulness of products or the rel-
                                                                         evance of information), purchases and related
Types of Information about the User                                      actions (e.g., in shopping carts, wish lists), and
Researchers have considered numerous kinds of
                                                                         usage regularities (such as usage frequencies,
user-related data for personalization purposes,
                                                                         high correlations between situations and spe-
including the following:
                                                                         cific actions, and frequently occurring sequences
 ■   Data about the user, such as demographic data,                      of actions)
     and information or assumptions about the                          ■ Environmental data, such as data about the
     user’s knowledge, skills, capabilities, interests,                  user’s software and hardware environments and
     preferences, goals, and plans                                       information about the user’s current location
6 ❘❙❚ BERKSHIRE ENCYCLOPEDIA OF HUMAN-COMPUTER INTERACTION



    (where the granularity ranges from country level      dents’ learning” (Conati et al. 2000, 404). Corbett and
    to the precise coordinates) and personalization-      Trask showed that a certain tutoring strategy (namely
    relevant data of this location.                       subgoal scaffolding based on a continuous knowl-
                                                          edge trace of the user) decreases the average number
                                                          of problems required to reach cognitive mastery of
Privacy                                                   Lisp concepts. In studies reviewed by Specht and
Storing information about users for personaliza-          Kobsa, students’ learning time and retention of learn-
tion is highly privacy relevant. Numerous con-            ing material improved significantly if learners with
sumer surveys show consistently that users are            low prior knowledge received “strict” recommen-
concerned about their privacy online, which also          dations on what to study next (which amounted to
affects personalized systems on the Web. Some             the blocking of all other learning material), while stu-
popular personalization methods also seem in con-         dents with high prior knowledge received noncom-
flict with privacy laws that protect the data of iden-     pulsory recommendations only. Strachan and
tified or identifiable individuals in more than thirty      colleagues found significantly higher user ratings for
countries. Such laws usually call for parsimony, pur-     the personalized version of a help system in a com-
pose specificity, and user awareness or even user con-     mercial tax advisor system than for its nonperson-
sent in the collecting and processing of personal data.   alized version.
The privacy laws of many countries also restrict the           Personalization for e-commerce on the Web
transborder flow of personal data or even extend their     has also been positively evaluated to some extent,
coverage beyond the national boundaries. Such laws        both from a business and a user point of view. Jupiter
then also affect personalized websites abroad that        Communications reports that personalization at
serve users in these regulated countries, even if there   twenty-five consumer e-commerce sites boosted the
is no privacy law in place in the country where the       number of new customers by 47 percent and revenues
websites are located. Well-designed user interaction      by 52 percent in the first year. Nielsen NetRatings re-
will be needed in personalized systems to commu-          ports that registered visitors to portal sites (who
nicate to users at any point the prospective benefits      obtain the privilege of adapting the displayed in-
of personalization and the resulting privacy conse-       formation to their interests) spend more than three
quences to enable users to make educated choices.         times longer at their home portal than other users
A flexible architecture, moreover, will be needed to       and view three to four times more pages. Nielsen
allow for optimal personalization within the con-         NetRatings also reports that e-commerce sites offer-
straints set by users’ privacy preferences and the le-    ing personalized services convert approximately twice
gal environment. Alternatively, anonymous yet             as many visitors into buyers than do e-commerce sites
personalized interaction can be offered.                  that do not offer personalized services. In design stud-
                                                          ies on beneficial personalized elements in a Web-based
                                                          procurement system, participants, however,“expressed
Empirical Evaluation                                      their strong desire to have full and explicit control of
A number of empirical studies demonstrate in sev-         data and interaction” and “to readily be able to make
eral application areas that well-designed adaptive user   sense of site behavior, that is, to understand a site’s
interfaces may give users considerable benefits. Boyle     rationale for displaying particular content” (Alpert
and Encarnacion showed that the automatic ad-             et al. 2003, 373).
justment of the wording of a hypertext document to             User-adaptable and user-adaptive interfaces have
users’ presumed familiarity with technical vocabu-        shown their promise in several application areas. The
lary improved text comprehension and search               increase in the number and variety of computer users
times significantly in comparison with static hyper-       is likely to increase their promise in the future. The
text. Conati and colleagues presented evidence that       observation of Browne still holds true, however:
“adaptive prompts based on the student model ef-          “Worthwhile adaptation is system specific. It is de-
fectively elicited self-explanations that improved stu-   pendent on the users of that system and requirements
                                                                                                               AFFECTIVE COMPUTING ❚❙❘ 7



to be met by that system” (Browne 1993, 69). Careful                           Specht, M., & Kobsa, A. (1999). Interaction of domain expertise and inter-
user studies with a focus on expected user benefits                                 face design in adaptive educational hypermedia. Retrieved March 24,
                                                                                   2004, from http://wwwis.win.tue.nl/asum99/specht/specht.html
through personalization are, therefore, indispensa-                            Strachan, L., Anderson, J., Sneesby, M., & Evans, M. (2000). Minimalist
ble for all practical deployments.                                                 user modeling in a complex commercial software system. User
                                                                                   Modeling and User-Adapted Interaction, 10(2–3), 109–146.
                                                         Alfred Kobsa          Teltzrow, M., & Kobsa, A. (2004). Impacts of user privacy preferences
                                                                                   on personalized systems—A comparative study. In C.-M. Karat,
See also Artificial Intelligence and HCI; Privacy; User                             J. Blom, & J. Karat (Eds.), Designing personalized user experiences
Modeling                                                                           for e-commerce (pp. 315–332). Dordrecht, Netherlands: Kluwer
                                                                                   Academic Publishers.



FURTHER READING
Alpert, S., Karat, J., Karat, C.-M., Brodie, C., & Vergo, J. G. (2003). User     AFFECTIVE COMPUTING
   attitudes regarding a user-adaptive e-commerce web site. User
   Modeling and User-Adapted Interaction, 13(4), 373–396.                      Computations that machines make that relate to hu-
Boyle, C., & Encarnacion, A. O. (1994). MetaDoc: An adaptive hy-
   pertext reading system. User Modeling and User-Adapted Interaction,
                                                                               man emotions are called affective computations.
   4(1), 1–19.                                                                 Such computations include but are not limited to
Browne, D. (1993). Experiences from the AID Project. In M. Schneider-          the recognition of human emotion, the expression
   Hufschmidt, T. Kühme, & U. Malinowski (Eds.), Adaptive user                 of emotions by machines, and direct manipulation
   interfaces: Principles and practice (pp. 69–78). Amsterdam: Elsevier.
Carroll, J., & Rosson, M. B. (1989). The paradox of the active user. In        of the human user’s emotions. The motivation for
   J. Carroll (Ed.), Interfacing thought: Cognitive aspects of human-          the development of affective computing is derived
   computer interaction (pp. 80–111). Cambridge, MA: MIT Press.                from evidence showing that the ability of humans
Conati, C., Gertner, A., & VanLehn, K. (2002). Using Bayesian networks
   to manage uncertainty in student modeling. User Modeling and
                                                                               to feel and display emotions is an integral part of hu-
   User-Adapted Interaction, 12(4), 371–417.                                   man intelligence. Emotions help humans in areas
Corbett, A. T., & Trask, H. (2000). Instructional interventions in com-        such as decisionmaking and human-to-human com-
   puter-based tutoring: Differential impact on learning time and ac-          munications. Therefore, it is argued that in order
   curacy. Proceedings of ACM CHI’ 2000 Conference on Human Factors
   in Computing Systems (pp. 97–104).
                                                                               to create intelligent machines that can interact ef-
Hof, R., Green, H., & Himmelstein, L. (1998, October 5). Now it’s YOUR         fectively with humans, one must give the machines
   WEB. Business Week (pp. 68–75).                                             affective capabilities.
ICONOCAST. (1999). More concentrated than the leading brand.                        Although humans interact mainly through
   Retrieved August 29, 2003, from http://www.iconocast.com/is-
   sue/1999102102.html                                                         speech, we also use body gestures to emphasize
Kobsa, A. (2002). Personalized hypermedia and international privacy.           certain parts of the speech and as one way to display
   Communications of the ACM, 45(5), 64–67. Retrieved August 29,               emotions. Scientific evidence shows that emo-
   2003, from http://www.ics.uci.edu/~kobsa/papers/2002-CACM-
   kobsa.pdf
                                                                               tional skills are part of what is called intelligence.
Kobsa, A., Koenemann, J., & Pohl, W. (2001). Personalized hyperme-             A simple example is the ability to know when some-
   dia presentation techniques for improving customer relationships.           thing a person says to another is annoying or
   The Knowledge Engineering Review, 16(2), 111–155. Retrieved August          pleasing to the other, and be able to adapt accord-
   29, 2003, from http://www.ics.uci.edu/~kobsa/papers/2001-KER-
   kobsa.pdf
                                                                               ingly. Emotional skills also help in learning to dis-
Kobsa, A., & Schreck, J. (2003). Privacy through pseudonymity in user-         tinguish between important and unimportant things,
   adaptive systems. ACM Transactions on Internet Technology, 3(2),            an integral part of intelligent decision-making. For
   149–183. Retrieved August 29, 2003, from http://www.ics.uci.edu/            computers to be able to interact intelligently with
   ~kobsa/papers/2003-TOIT-kobsa.pdf
Oppermann, R. (Ed.). (1994). Adaptive user support: Ergonomic design           humans, they will need to have such emotional skills
   of manually and automatically adaptable software. Hillsdale, NJ:            as the ability to display emotions (for example,
   Lawrence Erlbaum.                                                           through animated agents) and the ability to recog-
Rich, E. (1979). User modeling via stereotypes. Cognitive Science, 3,
   329–354.
                                                                               nize the user’s emotions. The ability to recognize
Rich, E. (1983). Users are individuals: Individualizing user models.           emotions would be useful in day-to-day interaction,
   International Journal of Man-Machine Studies, 18, 199–214.                  for example, when the user is Web browsing or
8 ❘❙❚ BERKSHIRE ENCYCLOPEDIA OF HUMAN-COMPUTER INTERACTION



                                                              lem with this approach is that the humans often feel
  FUNCTIONALITY The capabilities of a given program or        blended emotions. In addition, the choice of words
  parts of a program.                                         may be too restrictive or culturally dependent. Another
                                                              way to describe emotions is to have multiple dimen-
                                                              sions or scales. Instead of choosing discrete labels, emo-
searching: If the computer can recognize emo-                 tions are describe on several continuous scales, for
tions, it will know if the user is bored or dissatis-         example from pleasant to unpleasant or from simple
fied with the search results. Affective skills might also      to complicated. Two common scales are valence and
be used in education: A computer acting as a virtual          arousal.Valence describes the pleasantness of the stim-
tutor would be more effective if it could tell by stu-        uli, with positive (or pleasant) on one end and nega-
dents’ emotional responses that they were having dif-         tive (or unpleasant) on the other. For example,
ficulties or were bored—or pleased.                            happiness has a positive valence, while disgust has a
    It would not, however, be necessary for com-              negative valence. The other dimension is arousal, or
puters to recognize emotions in every application.            activation, which describes the degree to which the
Airplane control and banking systems, for example,            emotion stimulates the person experiencing it. For ex-
do not require any affective skills. However, in ap-          ample, sadness has low arousal, whereas surprise has
plications in which computers take on a social role           high arousal. The different emotional labels could be
(as a tutor, assistant, or even companion), it may en-        plotted at various positions on a two-dimensional plane
hance their functionality if they can recognize users’        spanned by these two axes to construct a two-di-
emotions. Computer agents could learn users’                  mensional emotion model. In 1954 the psychologist
preferences through the users’ emotions. Computers            Harold Schlosberg suggested a three-dimensional
with affective capabilities could also help human             model in which he added an axis for attention-rejec-
users monitor their stress levels. In clinical set-           tion to the above two. This was reflected by facial ex-
tings, recognizing a person’s inability to interpret          pressions as the degree of attention given to a person
certain facial expressions may help diagnose early            or object. For example, attention is expressed by
psychological disorders. In addition to recognizing           wide open eyes and an open mouth. Rejection shows
emotions, the affective computer would also have              contraction of eyes, lips, and nostrils.
the ability to display emotions. For example, syn-                 Although psychologists and others argue about
thetic speech with emotions in the voice would sound          what exactly emotions are and how to describe them,
more pleasing than a monotonous voice and                     everyone agrees that a lack of emotions or the
would enhance communication between the user                  presence of emotional disorders can be so dis-
and the computer. For computers to be affective, they         abling that people affected are no longer able to lead
must recognize emotions, be capable of measuring              normal lives or make rational decisions.
signals that represent emotions, and be able to
synthesize emotions.
                                                              Technology for Recognizing Emotions
                                                              Technologies for recognizing human emotions be-
General Description of Emotions                               gan to develop in the early 1990s. Three main modal-
Human beings possess and express emotions in every-           ities have been targeted as being relevant for this task:
day interactions with others. Emotions are often              visual, auditory, and physiological signals. The vi-
reflected on the face, in hand and body gestures, and          sual modality includes both static images and videos
in the voice. The fact that humans understand                 containing information such as facial expressions
emotions and know how to react to other people’s              and body motion. The audio modality uses prima-
expressions greatly enriches human interaction.               rily human voice signal as input, while the physio-
     There is no clear definition of emotions. One way         logical signals measure changes in the human body,
to handle emotions is to give them discrete labels, such      such as changes in temperature, blood pressure, heart
as joy, fear, love, surprise, sadness, and so on. One prob-   rate, and skin conductivity.
                                                                                       AFFECTIVE COMPUTING ❚❙❘ 9



Facial Expressions                                           of the universality of how emotions are expressed
One of the most common ways for humans to display            vocally, unlike the case for facial expressions. Research
emotions is through facial expressions. The best-known       that began in the late 1990s concentrated on com-
study of facial expressions was done by the psychol-         bining voice and video to enhance the recognition
ogist Paul Ekman and his colleagues. Since the               capabilities of voice-only systems.
1970s, Ekman has argued that emotions are manifested
directly in facial expressions, and that there are six ba-
sic universal facial expressions corresponding to            Multimodal Input
happiness, surprise, sadness, fear, anger, and disgust.      Many researchers believe that combining different
Ekman and his collaborator, the researcher Wallace           modalities enables more accurate recognition of a
Friesen, designed a model linking facial motions to ex-      user’s emotion than relying on any single modality
pression of emotions; this model is known as the Facial      alone. Combining different modalities presents both
Action Coding System (FACS). The facial action               technological and conceptual challenges, however.
coding system codes facial expressions as a combina-         On the technological side, the different signals
tion of facial movements known as action units. The          have different sampling rates (that is, it may take
action units have some relation to facial muscular mo-       longer to register signals in one modality than in an-
tion and were defined based on anatomical knowledge           other), and the existence of one signal can reduce the
and by studying videotapes of how the face changes           reliability of another (for example, when a person is
its appearance.                                              speaking, facial expression recognition is not as re-
     Ekman’s work inspired many other researchers to         liable). On the conceptual side, emotions are not al-
analyze facial expressions by means of image and video       ways aligned in time for different signals. For
processing. Although the FACS is designed to be              example, happiness might be evident visually before
performed by human observers viewing a video frame           it became evident physiologically.
by frame, there have been attempts to automate it in
some fashion, using the notion that a change in fa-
cial appearance can be described in terms of a set of        Computers Displaying Emotions
facial expressions that are linked to certain emo-           For affective computing, it is as important that com-
tions. Work on automatic facial-expression recogni-          puters display emotions as it is that they recognize them.
tion started in the early 1990s. In all the research, some   There are a number of potential ways in which com-
method to extract measurements of the facial features        puters could evince an emotion.A computer might de-
from facial images or videos was used and a classifier        pend on facial expressions of animated agents or on
was constructed to categorize the facial expressions.        synthesized speech, or emotion could be conveyed to
Comparison of facial expression recognition methods          a user through wearable devices and text messages. The
shows that recognition rates can, on limited data sets       method would be determined by the application do-
and applications, be very high. The generality of these      main and preset goals. For example, interaction in an
results has yet to be determined.                            office environment requires emotion to be expressed
                                                             differently from the way it would be expressed during
                                                             pursuit of leisure activities, such as video games; sim-
Voice                                                        ilarly, in computer-assisted tutoring, the computer’s
Quantitative studies of emotions expressed in the            goal is to teach the human user a concept, and the
voice have had a longer history than quantitative            display of emotions facilitates this goal, while in a game
studies of facial expressions, starting in the 1930s.        of poker, the computer’s goal is to hide its intention and
Studies of the emotional content of speech have              deceive its human adversary.
examined the pitch, duration, and intensity of the                A computer could also synthesize emotions in
utterance. Automatic recognition systems of emo-             order to make intelligent decisions with regard to prob-
tions from voice have so far not achieved high ac-           lems whose attributes cannot all be quantified ex-
curacy. In addition, there is no agreed-upon theory          actly, or when the search space for the best solution is
10 ❘❙❚ BERKSHIRE ENCYCLOPEDIA OF HUMAN-COMPUTER INTERACTION



large. By assigning valence to different choices based
on different emotional criteria, many choices in a large                                           ALTAIR
space can be eliminated quickly, resulting in a quick
and good decision.                                                              People have called the legendary Altair the first true
                                                                                personal computer. However, although it played
                                                                                an important role in the development of per-
Research Directions                                                             sonal computers, we would be more correct to say
Affective computing is still in its infancy. Some com-                          that Altair was the last hobby computer that set
puter systems can perform limited recognition of hu-                            many of the ambitions for the social movement
man emotion with limited responses in limited                                   that produced real personal computers during the
application domains. As the demand for intelligent                              1970s. Thus, from the standpoint of human-com-
computing systems increases, however, so does the need                          puter interaction, Altair is worth remembering be-
for affective computing.                                                        cause it marked a crucial transition between two
    Various moral issues have been brought up as                                eras of amateur computing: an experimental era
relevant in the design of affective computers. Among                            lasting from about 1950 until the mid-1970s, when
them are privacy issues: If a computer can recognize                            home computers were among the more esoteric
human emotions, a user may want assurances that                                 projects attempted by electronics hobbyists, and
information on his or her emotional state will not be                           the true personal computer era, beginning with
abused. There are also issues related to computers’ ma-                         such computers as the Apple II in 1977.
nipulation of people’s emotions: Users should have as-                               Altair was announced to the world late in 1974
surance that computers will not physically or                                   in the issue of Popular Electronics magazine
emotionally harm them. There are also questions re-                             dated January 1975. Some controversy exists about
garding who will have responsibility for computer                               how active a role the magazine played in launch-
actions. As affective technology advances, these issues                         ing Altair, but clearly Altair was actually designed
will have increasing relevance.                                                 and manufactured by the small company Micro
                                                                                Instrumentation and Telemetry Systems (MITS)
         Ira Cohen, Thomas S. Huang, Lawrence S. Chen                           in Albuquerque, New Mexico, headed by H. Edward
                                                                                “Ed” Roberts. Altair was a kit, costing $397, that
                                                                                required much skill from the builder and did not
FURTHER READING                                                                 include sufficient memory or input-output devices
                                                                                to perform any real tasks. The central processing
Darwin, C. (1890). The expression of the emotions in man and animals            unit was the new Intel 8080 microprocessor chip.
   (2nd ed.). London: John Murray.                                              Altair came with only 256 bytes of memory, and
Ekman, P. (Ed.). (1982). Emotion in the human face (2nd ed.). New
   York: Cambridge University Press.
                                                                                a notoriously unreliable 4-kilobyte memory ex-
Ekman, P., & Friesen,W.V. (1978). Facial action coding system: Investigator’s   pansion board kit cost an additional $264. After
   guide. Palo Alto, CA: Consulting Psychologists Press.                        a while MITS offered data input and output by
James, W. (1890). The principles of psychology. New York: Henry Holt.           means of an audio cassette recorder, but initially
Jenkins, J. M., Oatley, K., and Stein, N. L. (Eds.). (1998). Human emotions:
   A reader. Malden, MA: Blackwell.
                                                                                only a dedicated amateur was in a practical posi-
Lang, P. (1995). The emotion probe: Studies of motivation and attention.        tion to add a keyboard (perhaps a used teletype
   American Psychologist, 50(5), 372–385.                                       machine) or punched paper tape reader. Input-
Pantic, M., & Rothkrantz, L. J. M. (2000). Automatic analysis of facial ex-     output for the original computer was accomplished
   pressions: The state of the art. IEEE Transactions on Pattern Analysis
   and Machine Intelligence, 22(12), 1424–1445.                                 by switches and lights on the front of the cabinet.
Picard, R. W. (1997). Affective computing. Cambridge, MA: MIT Press.                 Popular Electronics hyped the Altair as if it were
Picard, R. W., Vyzas, E., & Healey, J. (2001). Toward machine emotional         a fully developed “minicomputer” and suggested
   intelligence: Analysis of affective physiological state. IEEE Transactions
   Pattern Analysis and Machine Intelligence, 23(10), 1175–1191.
                                                                                some excessively demanding applications: an au-
Schlosberg, H. (1954). Three dimensions of emotion. Psychological Review,       topilot for airplanes or boats, a high-speed in-
   61(2), 81–88.                                                                put-output device for a mainframe computer, a
                                                                                                     ALTAIR ❚❙❘ 11



brain for a robot, an automatic controller for an
air-conditioning system, and a text-to-Braille con-       “Personal computers are notorious for having a half-life
verter to allow blind people to read ordinary             of about two years. In scientific terms, this means that
printed matter. Altair rescued MITS from the verge        two years after you buy the computer, half of your friends
of bankruptcy, but the company could never                will sneer at you for having an outdated machine.”
fully deliver on the promise of the computer and                                                  —Peter H. Lewis
was absorbed by another company in 1977.
Hundreds of amateurs built Altairs on the way to
careers in the future personal computer industry,
its subcomponent interface bus became the widely        soldering each wire, capacitor, and resistor into
used S100 standard, and the computer contributed        place manually. Any of these parts might burn out
greatly to the revolution in human-computer in-         in use, so repair shops flourished, and companies
teraction that occurred during its decade. Notably,     such as Allied Radio and Lafayette Electronics sold
the mighty Microsoft corporation began life as a        individual parts to hobbyists and to anyone else
tiny partner of MITS, producing a BASIC inter-          who was willing to buy. For the novice, these dis-
preter for programming the Altair.                      tributors sold kits that provided all the parts needed
     Altair was far from being the first hobby          to build a project, and more advanced amateurs
computer, however. That honor probably be-              followed instructions in a number of magazines to
longs to Edmund C. Berkeley’s Simon relay-              build projects from parts they bought separately
based computer produced in 1950 and publicized          from distributors.
among hobbyists in the pages of Radio-Electronics           In purely financial terms building a stereo sys-
magazine. The most widely owned hobby digital           tem from a kit during the 1960s, as tens of thou-
computer before Altair was probably Berkeley’s          sands of people did, made little sense, but the result
GENIAC (Genius Almost-Automatic Computer),              was fully as good as the best ready-made system
which cost less than twenty dollars in 1955. Lacking    that could be bought in stores, and in some cases
vacuum tubes, relays, or transistors, this assem-       the designs were identical. The introduction of in-
bly of Masonite board, rotary switches, lights, and     tegrated circuits gradually reduced the role of re-
wires instructed students in the rudiments of logic     pairpersons, and by the dawn of the twenty-first
programming (programming the steps of logical           century much electronic equipment really could
deductions). Immediately prior to Altair, two less      not be repaired effectively and was simply replaced
influential hobby computers were also based on          when it broke down. Already by the late 1970s
Intel chips: the Scelbi 8H and the Titus Mark-8.        the electronics hobby was in decline, and the home-
The difference is that Altair was expandable and        built computer craze during that decade was prac-
intended to evolve into a full-featured personal        tically a last fling.
computer.                                                   For a decade after the introduction of Altair,
     The 1970s marked a turning point in the his-       a vibrant software hobbyist subculture prevailed
tory of hobby electronics, and innovative proj-         as people manually copied programs from a host
ects such as Altair could be seen as desperate          of amateur computer magazines, and many peo-
measures in the attempt to keep the field alive.        ple “brewed” their own personally designed word
Today some enthusiasts build electronic equipment       processors and fantasy games. This subculture de-
from kits or from scratch, just as others build their   clined after the introduction of complicated graph-
own harpsichords, but they no longer have the same      ical user interface operating systems by Apple and
relationship to the electronics industry that they      Microsoft, but it revived during the mid-1990s as
enjoyed during the middle decades of the twenti-        vast numbers of people created their own web-
eth century. Prior to the development of integrated     sites in the initially simple HTML (hypertext
circuits, factories constructed radios, televisions,    markup language). During its heyday this sub-
and audio amplifiers largely by hand, laboriously       culture was a great training ground of personnel
12 ❘❙❚ BERKSHIRE ENCYCLOPEDIA OF HUMAN-COMPUTER INTERACTION



for the electronics and computer industries because                          Data Systems” (SDS), and it established Xerox PARC
amateurs worked with the same technology that pro-                           near Stanford University in the area that would soon
fessionals worked with. Altair was a watershed per-                          be nicknamed “Silicon Valley.” Xerox proclaimed the
sonal computer in the sense that amateurs assembled                          grand goal of developing the general architecture of
it personally and that it transformed them person-                           information rather than merely producing a num-
ally into computer professionals.                                            ber of unconnected, small-scale inventions.
                                                                                  The Alto was part of a larger system of software
                                      William Sims Bainbridge                and hardware incorporating such innovations as ob-
                                                                             ject-oriented programming, which assembles pro-
See also Alto                                                                grams from many separately created, reusable
                                                                             “objects,” the ethernet LAN, and laser printers. At the
                                                                             time computers were large and expensive, and a com-
FURTHER READING                                                              mon framework for human-computer interaction
                                                                             was time sharing: Several users would log onto a
Freiberger, P., & Swaine, M. (1999). Fire in the valley: The making of the   mainframe or minicomputer simultaneously from
   personal computer (2nd ed.). New York: McGraw-Hill.
Roberts, H. E., & Yates, W. (1975). Altair minicomputer. Popular
                                                                             dumb terminals, and it would juggle the work from
   Electronics, 7(1), 33–38.                                                 all of the users simultaneously. Time sharing was
Roberts, H. E., & Yates, W. (1975). Altair minicomputer. Popular             an innovation because it allowed users to interact
   Electronics, 7(2), 56–58.                                                 with the computer in real time; however, because the
Mims, F. M. (1985, January). The tenth anniversary of the Altair 8800.
   Computers & Electronics, 23(1), 58–60, 81–82.                             computer was handling many users it could not de-
                                                                             vote resources to the HCI experience of each user. In
                                                                             contrast, Alto emphasized the interface between the
                                                                             user and the machine, giving each user his or her own
                             ALTO                                            computer.
                                                                                  In April 1973 the first test demonstration of an
The Alto computer, developed at the Xerox                                    Alto showed how different using it would be from us-
Corporation’s Palo Alto Research Center (Xerox                               ing the text-only computer terminals that people were
PARC) in the 1970s, was the prototype of the late                            used to when it began by painting on its screen a pic-
twentieth-century personal computer. Input was by                            ture of the Cookie Monster from the television
means of both keyboard and mouse; the display                                program Sesame Street. The Alto’s display em-
screen integrated text and graphics in a system of win-                      ployed bitmapping (controlling each pixel on the
dows, and each computer could communicate with                               screen separately) to draw any kind of diagram,
others over a local area network (LAN). The Alto was                         picture, or text font, including animation and pull-
significant for human-computer interaction (HCI)                              down menus. This capability was a great leap forward
in at least three ways. First, it established a new dom-                     for displaying information to human beings, but it
inant framework for how humans would interact with                           required substantial hardware resources, both in terms
computers. Second, it underscored the importance                             of memory size and processing speed, as well as rad-
of theory and research in HCI. Third, the failure of                         ically new software approaches.
Xerox to exploit Alto technology by gaining a dom-                                During the 1970s the typical computer display
inant position in the personal computer industry is                          consisted of letters, numbers, and common punctu-
a classic case study of the relationship between in-                         ation marks in a single crude font displayed on a black
novators and the technology they create.                                     background in one color: white or green or amber.
    During the late 1960s the Xerox Corporation was                          In contrast, the default Alto display was black on
aware that it might gradually lose its dominant posi-                        white, like printed paper. As originally designed,
tion in the office copier business, so it sought ways of                      the screen was 606 pixels wide by 808 pixels high, and
expanding into computers. In 1969 it paid $920 mil-                          each of those 489,648 pixels could be separately con-
lion to buy a computer company named “Scientific                              trolled. The Xerox PARC researchers developed sys-
                                                                                                      ANIMATION ❚❙❘ 13



tems for managing many font sizes and styles si-            computing. In contrast, the model that flourished
multaneously and for ensuring that the display screen       during the 1980s was autonomous personal com-
and a paper document printed from it could look             puting based on stand-alone computers such as the
the same. All this performance placed a heavy bur-          Apple II and original IBM PC, with networking de-
den on the computer’s electronics, so an Alto often         veloping fully only later. The slow speed and lim-
ran painfully slow and, had it been commercialized,         ited capacity of the Alto-like Lisa and original
would have cost on the order of $15,000 each.               128-kilobyte Macintosh computers introduced by
    People have described the Alto as a “time ma-           Apple in 1983 and 1984 suggest that Alto would re-
chine,” a computer that transported the user into           ally not have been commercially viable until 1985,
the office of the future, but it might have been too         a dozen years after it was first built.
costly or too slow to be a viable personal com-                  One lesson that we can draw from Alto’s history
puter for the average office or home user of the            is that corporate-funded research can play a deci-
period in which it was developed. Human-computer            sive role in technological progress but that it can-
interaction research of the early twenty-first cen-         not effectively look very far into the future. That role
tury sometimes studies users who are living in the          may better be played by university-based labora-
future. This means going to great effort to create an       tories that get their primary funding from govern-
innovation, such as a computer system or an envi-           ment agencies free from the need to show immediate
ronment such as a smart home (a computer-con-               profits. On the other hand, Xerox PARC was so spec-
trolled living environment) or a multimedia                 tacularly innovative that we can draw the opposite
classroom, that would not be practical outside the          lesson—that revolutions in human-computer in-
laboratory. The innovation then becomes a test bed          teraction can indeed occur inside the research lab-
for developing future systems that will be practical,       oratories of huge corporations, given the right
either because the research itself will overcome some       personnel and historical circumstances.
of the technical hurdles or because the inexorable
progress in microelectronics will bring the costs                                              William Sims Bainbridge
down substantially in just a few years.
    Alto was a remarkable case study in HCI with            See also Altair; Graphical User Interface
respect to not only its potential users but also its cre-
ators. For example, the object-oriented program-
ming pioneered at Xerox PARC on the Alto and                FURTHER READING
other projects changed significantly the work of pro-
grammers. Such programming facilitated the sepa-            Hiltzik, M. (1999). Dealers of lightning: Xerox PARC and the dawn of
                                                               the computer age. New York: HarperBusiness.
ration between two professions: software engineering        Lavendel, G. (1980). A decade of research: Xerox Palo Alto Research
(which designs the large-scale structure and func-             Center. New York: Bowker.
tioning of software) and programming (which writes          Smith, D. C., & Alexander, R. C. (1988). Fumbling the future: How
the detailed code), and it increased the feasibility of        Xerox invented, then ignored the first personal computer. New
                                                               York: William Morrow.
dividing the work of creating complex software              Waldrop, M. M. (2001). The dream machine: J. C. R. Licklider and the
among many individuals and teams.                              revolution that made computing personal. New York: Viking.
    People often have presented Alto as a case study
of how short sighted management of a major cor-
poration can fail to develop valuable new technol-
ogy. On the other hand, Alto may have been both
too premature and too ambitious. When Xerox fi-                                ANIMATION
nally marketed the Alto-based Star in 1981, it was a
system of many small but expensive computers, con-          Animation, the creation of simulated images in mo-
nected to each other and to shared resources such           tion, is commonly linked with the creation of car-
as laser printers—a model of distributed personal           toons, where drawn characters are brought into play
14 ❘❙❚ BERKSHIRE ENCYCLOPEDIA OF HUMAN-COMPUTER INTERACTION



to entertain. More recently, it has also become a            side and a cycle of still images on the inside that
significant addition to the rich multimedia mate-            show an image in graduating stages of motion.
rial that is found in modern software applications           Whenever the barrel spins rapidly, the dark frames
such as the Web, computer games, and electronic              of the still pictures disappear and the picture appears
encyclopedias.                                               to move. Another, even simpler example is the flipbook,
                                                             a tablet of paper with a single drawing on each page.
                                                             When the book is flicked through rapidly, the draw-
Brief History                                                ings appear to move.
Animations are formed by showing a series of still pic-           Once the basic principles of animation were dis-
tures rapidly (at least twelve images per second) so that    covered, a large number of applications and techniques
the eye is tricked into viewing them as a continuous         emerged. The invention of these simple animation de-
motion. The sequence of still images is perceived as         vices had a significant influence on the development
motion because of two phenomena, one optical                 of films, cartoons, computer-generated motion graph-
(persistence of vision) and one psychological (phi prin-     ics and pictures, and more recently, of multimedia.
ciple). Persistence of vision can be explained as the pre-
disposition of the brain and eye to keep on seeing a
picture even after it has moved out of the field of vi-       Walt Disney and Traditional
sion. In 1824 British scientist, physician, and lexicog-
rapher Peter Mark Roget (1779–1869) explained this
                                                             Animation Techniques
                                                                 During the early to mid-1930s, animators at
phenomenon as the ability of the retina to retain the
                                                             Walt Disney Studios created the “twelve animation
image of an object for 1/20 to 1/5 second after its re-
                                                             principles” that became the basics of hand-drawn car-
moval; it was demonstrated two years later using a
                                                             toon character animation. While some of these prin-
thaumatrope, which is a disk with images drawn on
                                                             ciples are limited to the hand-drawn cartoon animation
both sides that, when twirled rapidly, gives the illusion
                                                             genre, many can be adapted for computer animation
that the two images are combined together to form
                                                             production techniques. Here are the twelve principles:
one image.
     The other principle is the phi phenomenon or stro-        1. Squash and stretch—Use shape distortion to em-
boscopic effect. It was first studied by German psy-               phasize movement.
chologist Max Wertheimer (1880–1943) and                       2. Anticipation—Apply reverse movement to pre-
German-American psycho-physiologist Hugo                          pare for and bring out a forward movement.
Munsterberg (1863–1916) during the period from 1912            3. Staging—Use the camera viewpoint that best
to 1916. They demonstrated that film or animation                  shows an action.
watchers form a mental connection that completes the           4. Straight-ahead vs. pose-to-pose action—Apply
action frame-to-frame, allowing them to perceive a se-            the right procedure.
quence of motionless images as an uninterrupted                5. Follow-through and overlapping action—Avoid
movement. This mental bridging means that even if                 stopping movement abruptly.
there are small discontinuities in the series of               6. Slow-in and slow-out—Allow smooth starts and
frames, the brain is able to interpolate the missing de-          stops by spacing frames appropriately.
tails and thus allow a viewer to see a steady movement.        7. Arcs—Allow curved motion in paths of action.
     In the nineteenth century, many animation de-             8. Secondary actions—Animate secondary actions
vices, such as the zoetrope invented by William George            to bring out even more life.
Horner (1786–1837), the phenakistiscope (1832), the            9. Timing—Apply time relations within actions
praxinoscope (1877), the flipbook, and the thaumatrope             to create the illusion of movement.
were direct applications of the persistence of vision.        10. Exaggeration—Apply caricature to actions and
For example, the zoetrope is a cylindrical device                 timing.
through which one can see an image in action. The ro-         11. Solid drawing—Learn and use good drawing
tating barrel has evenly spaced peepholes on the out-             techniques.
                                                                                               ANIMATION ❚❙❘ 15



 12. Appeal—Create and animate appealing                    or a running character. In sprite-based animation, a
     characters.                                            single image or sequence of images can be at-
                                                            tached to a sprite. The sprite can animate in one place
     Traditional animation techniques use cel anima-        or move along a path. Many techniques—for exam-
tion in which images are painted on clear acetate sheets    ple, tiling, scrolling, and parallax— have been devel-
called cels. Animation cels commonly use a layering         oped to process the background layer more efficiently
technique to produce a particular animation frame.          and to animate it as well. Sometimes sprite-based
The frame background layer is drawn in a separate           animation is called path-based animation. In path-
cel, and there is a cel for each character or object that   based animation, a sprite is affixed to a curve drawn
moves separately over the background. Layering              through the positions of the sprite in consecutive
enables the animator to isolate and redraw only the         frames, called a motion path. The sprite follows
parts of the image that change between consecutive          this curve during the course of the animation. The
frames. There is usually a chief animator who               sprite can be a single rigid bitmap (an array of pix-
draws the key-frames, the ultimate moments in the           els, in a data file or structure, which correspond bit
series of images, while in-between frames are drawn         for bit with an image) that does not change or a se-
by others, the in-betweeners. Many of the processes         ries of bitmaps that form an animation loop. The
and lingo of traditional cel-based animation, such as       animation techniques used by computers can be
layering, key-frames, and tweening (generating im-          frame-by-frame, where each frame is individually cre-
mediate frames between two images to give the ap-           ated, or real-time, where the animator produces
pearance that the first image evolves smoothly into          the key-frames and the computer generates the
the next), have carried over into two-dimensional           frames in between when the animation is dis-
and three-dimensional computer animation.                   played at run time.
                                                                 Two-dimensional computer animation tech-
Two-Dimensional                                             niques are widely used in modern software and
                                                            can be seen in arcade games, on the Web, and even
Computer Animation                                          in word processors. The software used to design two-
In recent years, computer programs have been de-            dimensional animations are animation studios
veloped to automate the drawing of individual               that allow animators to draw and paint cels, provide
frames, the process of tweening frames between key-         key-frames with moving backgrounds, use multiple
frames, and also the animation of a series of frames.       layers for layering, support linking to fast video disk
Some animation techniques commonly used in two-             recorders for storage and playback, and allow scans
dimensional (2D) computer animation are either              to be imported directly. Examples of this software
frame-based or sprite-based.                                include Adobe Photoshop (to create animated GIFs),
    Frame-based animation is the simplest type of           Macromedia Director (multimedia authoring tool
animation. It is based on the same principle as the         that includes sophisticated functions for animation),
flipbook, where a collection of graphic files, each con-      Macromedia Flash (vector-based authoring tool to
taining a single image, is displayed in sequence and        produce real-time animation for the Web). Even some
performs like a flipbook. Here again, to produce            programming languages such as Java are used to pro-
the illusion of motion, graphic images, with each im-       duce good quality animation (frame-by-frame and
age slightly different from the one before it in the se-    real-time) for the Web.
quence, are displayed at a high frame-rate (the number
of frames of an animation displayed every second).          Three-Dimensional
    Sprite-based animation uses a technique that is
similar to the traditional animation technique in           Computer Animation
which an object is animated on top of a static graphic      Three-dimensional computer animations are based
background. A sprite is any element of an animation         on a three-dimensional (3D) coordinate system,
that moves independently, such as a bouncing ball           which is a mathematical system for describing
16 ❘❙❚ BERKSHIRE ENCYCLOPEDIA OF HUMAN-COMPUTER INTERACTION



three-dimensional space. Space is measured along         Skeletal Structure Animation
three coordinates, the X direction, Y direction, and     Skeletal structures are bones-based. Widely used to
Z direction. These coordinates correspond to the         control three-dimensional creatures, they appear in
width, length, and depth of objects or space. The        practically all modern three-dimensional model-
X, Y, Z coordinates of points in space are used to de-   ing software studios. They enable the artist to preset
fine polygons, and collections of polygons make          and control the rotation points of a three-dimen-
up the definition of three-dimensional objects.           sional creature, facilitating its animation. The ani-
     The process of 3D animation involves at least the   mator can then model a geometric skin (representing
following stages: modeling, rendering, and anima-        how the creature would appear) and link it to the
tion. Modeling is the process of creating 3D ob-         bones structure. Skeletal structures software with
jects from simple 2D objects by lofting (the process     graphical and powerful interfaces provide rich en-
of transforming a two-dimensional cross section ob-      vironments in which artists can control the complex
ject into a complete three-dimensional object) or        algorithms involved in creating animated three-
from other simple 3D objects called “primitives”         dimensional creatures (human, animal, or imaginary).
(spheres, cubes, cylinders, and so on). Primitives can   The availability of a skeletal animation environment
be combined using a variety of Boolean operations        characteristically brings another advantage—the
(union, subtraction, intersection, and so on). They      exploitation of inverse kinematics (IK) to bring a char-
can also be distorted in different ways. The resulting   acter to life.
model is called a mesh, which is a collection of
faces that represent an object. Rendering is used to     Inverse Kinematics
create an image from data that represents objects        IK is a common technique for positioning multi-
as meshes, and to apply colors, shading, textures, and   linked objects, such as virtual creatures. When using
lights to them. In its simplest form, the process of     an animation system capable of IK, a designer can
three-dimensional computer animation is very sim-        position a hand in space by grabbing the hand and
ilar to the two-dimensional process of key-frames        leading it to a position in that space. The con-
and tweening. The main differences are that three-       nected joints rotate and remain connected so that,
dimensional animations are always vector-based and       for example, the body parts all stay connected. IK
real-time.                                               provides a goal-directed method for animating a 3D
                                                         creature. It allows the animator to control a three-
Spline-Based Animation                                   dimensional creature’s limbs by treating them as a
Motion paths are more believable if they are curved,     kinematics chains. The points of control are attached
so animation programs enable designers to create         to the ends of these chains and provide a single han-
spline-based motion paths. (Splines are algebraic        dle that can be used to control a complete chain.
representations of a family of curves.) To define        IK enables the animator to design a skeleton system
spline-based curves, a series of control points is de-   that can also be controlled from data sets gener-
fined and then the spline is passed through the con-      ated by a motion capture application.
trol points. The control points define the beginning
and end points of different parts of the curve. Each     Motion Capture
point has control handles that enable designers to       Motion capture is the digital recording of a creature’s
change the shape of the curve between two control        movement for immediate or postponed analysis and
points. The curves and the control points are de-        playback. Motion capture for computer character an-
fined in 3D space. Most computer animation sys-          imation involves the mapping of human action onto
tems enable users to change the rate of motion along     the motion of a computer character. The digital data
a path. Some systems also provide very sophisti-         recorded can be as simple as the position and ori-
cated control of the velocity of an object along         entation of the body in space, or as intricate as the
paths.                                                   deformations of the expression of the visage.
                                                                                         ANTHROPOLOGY AND HCI ❚❙❘ 17



Advances in                                                  work in animation lies in physic-based modeling in
                                                             which objects or natural phenomena are animated
Three-Dimensional Animation                                  according to their real physical properties, in real-
With the support of powerful computers, three-
                                                             time motion capture, and in goal-orientated ani-
dimensional animation allows the production and
                                                             mation. Considering the numerous applications of
rendering of a photo-realistic animated virtual world.
                                                             animation, from multimedia to archeology and
Three-dimensional scenes are complex virtual envi-
                                                             chemistry, the future possibilities seem endless.
ronments composed of many elements and effects,
such as cameras, lights, textures, shading, and envi-
                                                                          Abdennour El Rhalibi and Yuanyuan Shen
ronment effects, and all these elements can be ani-
mated. Although cel animation is traditionally
                                                             See also Data Visualization; Graphic Display;
two-dimensional, advances in three-dimensional
                                                             Graphical User Interface
rendering techniques and in camera animation have
made it possible to apply three-dimensional tech-
niques to make two-dimensional painted images ap-
                                                             FURTHER READINGS
pear visually three-dimensional. The 3D animation
techniques described in this section are supported           CoCo, D. (1995). Real-time 3D games take off. Computer Graphics
by modern 3D animation studios that are software                World, 8(12), 22–33.
programs such as Maya (alias|wavefront), Softimage           Corra, W. T., Jensen, R. J., Thayer, C. E., & Finkelstein, A. (1998).
                                                                Texture mapping for cel animation. In Proceedings of SIGGRAPH’ 98,
(Softimage), 3D Studio Max (Discreet), or Rhino3D               Computer Graphics Proceedings, Annual Conference Series (pp. 435–446).
(Robert McNeel & Associates).                                Kerlow, I. V. (2000). The art of 3-D computer animation and imaging
     Examples of environment effects include rain,              (2nd ed.). New York: Wiley.
fire, fog, or dying stars. A technique widely used in         Lassiter, J. (1987). Principles of traditional animation applied to 3D
                                                                computer animation. SIGGRAPH 87 (pp. 35–44).
real-time applications involving an environmental            Maltin, L. (1987). Of mice and magic—A history of American animated
effect is called a particle system. A particle system is        cartoons. New York: Penguin Books.
a method of graphically producing the appearance             O’Rourke, M. (1995). Principles of three-dimensional computer ani-
                                                                mation. New York: W. W. Norton.
of amorphous substances, such as clouds, smoke,
                                                             Parent, R. (2001). Computer animation: Algorithms and techniques.
fire, or sparkles. The substance is described as a col-         San Francisco: Morgan-Kaufmann.
lection of particles that can be manipulated dy-             Potter, C. D. (1995). Anatomy of an animation. Computer Graphics
namically for animation effects. Some even more                 World, 18(3). 36–43.
                                                             Solomon, C. (1994). The history of animation: Enchanted drawings.
recent techniques include physics-based behavior such           New York: Wings Books.
as a realistic animation of cloth, hair, or grass affected   Thomas, F., & Johnson, O. (1981). The illusion of life. New York:
by the wind.                                                    Abbeville Press.
                                                             Watt, A., & Policarpo, F. (2001). 3D games—real-time rendering and
                                                                software technology. New York: Addison-Wesley.
                                                             Watt, A. H., & Watt, M. (1992). Advanced animation and rendering.
Endless Possibilities                                           New York: Addison-Wesley.
                                                             Williams, R. (2001). The animator’s survival kit. New York: Faber &
Animation has become an ubiquitous component                    Faber.
of human-computer interfaces. It has evolved
from prehistoric paintings in Altamira caves to re-
alistic virtual worlds in sophisticated multimedia
computers. The technologies supporting animation             ANTHROPOLOGY AND HCI
are still emerging and will soon support even more
complex worlds, more realistic character animation,          As a social science that brings together social anthro-
considerably easier 3D animation development, bet-           pology, linguistics, archaeology, and human biology,
ter quality animations on the Web, and better inter-         anthropology clearly has a major contribution to
actions with virtual reality interfaces. The current         make to the study of human-computer interaction
18 ❘❙❚ BERKSHIRE ENCYCLOPEDIA OF HUMAN-COMPUTER INTERACTION



(HCI). However, bringing that contribution into fo-             A third concept, actually a method, employed by
cus is at times a challenge, not only because of the       anthropologists in the study of HCI is actor-network
extreme interdisciplinarity but also because of collab-    theory. This theory views artifacts and social roles as
orations between anthropologists and computer sci-         coevolving nodes in a common network. Insofar as
entists and the sometimes-blurred boundaries               each node encodes information about the entire net-
between anthropology and related disciplines, includ-      work (for example, in any country, electrical appli-
ing sociology and psychology. Despite these challenges,    ances are tailored to the specific power system of the
anthropology has created distinctive methods and a         country and the expectations of the users) and is ca-
distinctive epistemology, and has offered new insights     pable of state changes based on network inputs, both
for understanding human-computer interaction.              artifacts and social roles can be considered to have
Anthropology also poses three profound questions.          agency within the network. This concept, originally
                                                           developed by the sociologist Michel Callon, in his study
                                                           of the French government’s involvement in techno-
Methods                                                    logical projects, and elaborated by the sociologist John
Anthropology’s development of ethnographic meth-           Law in a study of Portuguese sailing vessels in the
ods is a notable contribution to research in HCI. More     sixteenth century, is very pertinent to rapidly chang-
than simple naturalistic observation, ethnography is       ing technologies such as computers. Indeed, observ-
a structured process informed by theoretical models        ing the shifting topology of the Internet and Internet
through which researchers attempt to elucidate the co-     computing makes it clear that user roles are anticipated
herence of a context. For example, anthropologist          and complemented by machine behavior (for instance,
Bonnie Nardi, in her study of end-user computing used      collaborative filtering), and machine states enable or
concepts of formalisms and communication to inter-         constrain users’ agency within the network (for ex-
pret how users developed their own programs; an-           ample, the structures of search engines). Although sil-
thropologist Lucy Suchman used a mechanistic concept       icon and carbon units are distinct, for now, the
of cognition as a foil to understand how users inter-      image of the cyborg (cybernetic organism), and the
acted with an expert-system-based help facility em-        emergence of integrated biological/computational sys-
bedded in a copying machine. In both these cases           tems, suggests other possibilities.
researchers combined intensive naturalistic observa-            This hints at the final, and perhaps most impor-
tion with conceptual insights to develop new HCI           tant anthropological contribution to HCI, the evolu-
models.                                                    tionary perspective. All branches of anthropology have
     A frequently employed variation on ethnographic       been concerned with the evolution of human societies,
methods is called ethnomethodology. As originally          languages, and even genotypes. Although there is room
developed by sociologist Harold Garfinkel, eth-            for debate over the telos or chaos of evolutionary
nomethodology stipulates that individuals make             processes, understanding humans and their artifacts
sense out of a context in an ad hoc, almost indeter-       as goal-seeking objects who learn is fundamental to
minate manner. In place of social order, the actors        any anthropological viewpoint. Using the archaeo-
in a given context are synthesizing what appears to        logical record and anthropological knowledge of so-
be order, accepting or rejecting information as it fits     cieties with simpler toolkits, the anthropologist David
with their synthesis. The mutual intelligibility of        Hakken has questioned the extent to which the
an interaction is thus an ongoing achievement be-          widespread use of computers in society justifies being
tween the actors, a result rather than a starting point.   called a “revolution”; he concludes that due to their
Thus, two users can construct two quite different          failure to transform the character of labor, computers
meanings out of similar interactions with comput-          are “just one more technology” in the implementation
ers, depending on the experiences they bring to the        of an automated, massified Fordist model of pro-
interaction. This suggests some obvious limita-            duction—a model inspired by Henry Ford in which
tions on the abilities of computers to constrain or        large quantities of products are produced through the
reproduce human actions.                                   repetitive motions of unskilled workers.
                                                                                            ANTHROPOLOGY AND HCI ❚❙❘ 19



Epistemology                                                      of passage for most anthropologists. When re-
What distinguishes anthropology from other disciplines            searchers have lived for an extended period of time
such as psychology and sociology that use similar                 in an unfamiliar village, cut off from their normal
methods is in many ways a matter of epistemology—                 social moorings, when cultural disorientation be-
that is, the stance it takes toward the subject mat-              comes embedded in their daily routine, they acquire
ter. Central to this stance is the orthogonal view, that          a profound conviction that all social forms are
is, the ability to analyze a situation from a fresh               conventional, that otherness is not alien, and that
and original yet plausible perspective. It is the or-             belonging and familiarity are rare and fragile flow-
thogonal view that enables anthropologist Constance               ers. It is this experience and this conviction more
Perin to see office automation as a panopticon,                   than any methodological or conceptual apparatus
that suggests to linguistic anthropologist Charlotte              that defines anthropology and that enables the or-
Linde that failed communication can improve per-                  thogonal view.
formance, or that led Edwin Hutchins, a cognitive                     It is this implicitly critical stance that has con-
anthropologist, to understand a cockpit as a cogni-               strained anthropology’s contribution to the study
tive device. Orthogonal viewpoints originate from                 of automation human factors. “Human factors” is
the experience of fieldwork, or rather, field immer-                an engineering discipline using engineering meth-
sion, preferably in a remote setting, which is the rite           ods of analytic decomposition to solve engineering



                   A Personal Story—Eastern vs. Western Cultural Values

  My understanding of human communication using mediated technologies is primarily based on cultural assumptions.
  Cultural values could influence the way a human chooses its medium of communication. On the other hand, with the
  advancement of computer-mediated communication (CMC) technologies (e.g., e-mail, e-commerce sites, weblogs, bul-
  letin boards, newsgroups) people could also change their communication patterns to suit the different forms of a medium.
  Whichever way, apparently, people will not adopt CMC unless and until it fits with their cultural values. Based on my in-
  terviews with a number of informants from different cultural backgrounds, I have observed some disparate yet interesting
  views on communication patterns and preferences, i.e., why and when people use CMC. Let me briefly illustrate one case
  of contrasting communication preferences and patterns.
      When I asked the informants from Eastern cultures why they would use CMC, one of the key responses was that they
  can express themselves better over mediated technologies than to voice their opinions in face-to-face. Public self-expres-
  sion is avoided due to the value of “saving face.” Also, using asynchronous medium such as e-mail, does not require
  spontaneous response. People could first think, reflect, and then express. On the contrary, the informants from Western
  cultures felt that using e-mail is best for complex and detailed information, as they require very explicit forms of instruc-
  tions. Additionally, people send messages via CMC in order to get quick response so that tasks can get completed. Also,
  based on a written format, the text becomes an evidence or “proof of say” for a job accomplished. Getting a job or as-
  signment done is perceived as a priority and building a relationship is thus secondary.
      Cultural values could present a new lens to understand why and how certain a communication medium offers differ-
  ent functions or purposes. What is more important is the uniqueness of human beings with a set of cultural assump-
  tions and values, and not the technological features. Anthropologist Edward T. Hall postulates that “communication is
  culture and culture is communication.” Hence, organizations need to understand fully the myriad cultural preferences be-
  fore making a substantial investment in CMC technology. Without such understanding, technology will simply be another
  gadget that gets rusty and dusty!
                                                                                                            Norhayati Zakaria
20 ❘❙❚ BERKSHIRE ENCYCLOPEDIA OF HUMAN-COMPUTER INTERACTION




                        Digital Technology Helps Preserve Tribal Language


      (ANS)—The American Indian language of Comanche                  of American Indian Studies at the University of South
      was once taught through conversation—a vocabulary               Dakota in Vermillion. Students at that university, located
      passed on and polished as it moved from one genera-             in the midst of a large Sioux community, are increasingly
      tion to the next. But as fluency among Comanches de-            interested in learning indigenous languages, he said.
      clines, the tribe has turned to cutting-edge technology to          Under a federal policy of discouraging use of American
      preserve this indigenous language. By next winter, mem-         Indian languages by allowing only English to be spoken
      bers hope to produce an interactive CD-ROM that will            by American Indian children at schools run by the Bureau
      create a digital record of the language and help tribe mem-     of Indian Affairs, Comanche began faltering about 50
      bers learn it.                                                  years ago.
           “You can’t say you’re Comanche without knowing                 Without preservation efforts, researchers predict that
      your own language. That’s the way I feel,” said Billie Kreger   90 percent of the world’s languages, including those of
      of Cache, Okla., vice president of the Comanche Language        the 554 American Indian tribes, will disappear in the next
      and Cultural Preservation Committee. Kreger, 47, did-           century, said Peg Thomas, executive director of The Grotto
      n’t learn much Comanche as a child but has begun study-         Foundation, a nonprofit organization in St. Paul,
      ing it in the past few years.                                   Minn., that provides funding to American Indian organ-
           Of the 10,000 Comanches that still remain in the           izations. Each year about five languages fall into “ex-
      United States, roughly 200 are fluent, according to Karen        tinction,” meaning that they have no youthful speakers,
      Buller, president and chief executive officer of the Santa       she said.
      Fe, N.M.-based organization that is paying for the CD-              According to some estimates, between 300 and 400
      ROM project, the first of its kind in the United States.        American Indian languages have become extinct since
      Tribe members are anxious to record the language                European settlers first arrived in North America.
      while the fluent speakers, who are in their 70s and 80s,             The point of preserving the languages is partly to
      are still living, she said.                                     maintain a connection to the past and learn the history
           Buller’s group, the National Indian Telecom-               of a culture, said Buller. Students of the Comanche lan-
      munications Institute, is paying for the project with           guage discover, for instance, that the words for food prepa-
      $15,000 in grant money from the Fund for the Four               ration are based on the root word for “meat”—because
      Directions.                                                     meat was a key part of the Comanche diet. She and oth-
           The CD-ROM will teach about 1,500 vocabulary               ers say that American Indian children who learn indige-
      words. Students will see Comanche elders pronouncing            nous languages in addition to English appear to perform
      the words and hear the words used in conversations.             better in school.
      Buller’s group is recording conversations on videotape.             But language programs are targeting adults, too.
           Other indigenous language revitalization efforts are       Kreger, of the Comanche Language and Cultural
      under way around the country, too, including language           Preservation Committee, says she is looking forward to
      immersion programs in Alaskan and Hawaiian schools.             using the CD-ROM for her own language studies. “I
      The institute provided teacher training for those projects.     can hardly wait,” she said.
           “All the tribes are saying, ‘We’ve got to save the lan-                                                  Nicole Cusano
      guage,’” said Leonard Bruguier, who heads the Institute            Source: Digital technology helps preserve tribal language. American
                                                                         News Service, June 15, 2000.
                                                                              ANTHROPOLOGY AND HCI ❚❙❘ 21



problems—in other words, the improved perform-           that is, construct—the context of information, a
ance of artifacts according to some preestablished       structure of irreducible complexity. The context is
set of specifications. Anthropology, by contrast,        far more than simply a compilation of information.
would begin by questioning the specifications, adopt-     Computers and other information technologies,
ing a holistic point of view toward the entire proj-     by contrast, focus on the processing of information,
ect. Holism is the intellectual strategy of grasping     stripping information of its contextual properties
the entire configuration rather than breaking it down     and thus of the attributes that humans use to turn
into separate elements. From an anthropological          information into (warranted, usable, and meaning-
viewpoint, specifications are not a given, but open       ful) knowledge.
to interrogation. A holistic viewpoint requires               John Seely Brown, the former director of Xerox
that the researcher adopt multiple disciplinary tools,   Palo Alto Research Center, and researcher Paul
including (but certainly not limited to) direct ob-      Duguid, for example, describe the importance of
servation, interviewing, conversation analysis, en-      context for using information. “The news,” for in-
gineering description, survey research, documentary      stance, is not simply unfiltered information from a
study, and focus groups. For many, anthropology is       distant place; it is information that has been selected,
highly interdisciplinary, assembling research tools      aggregated, evaluated, interpreted, and warranted by
as the contextualized problem requires.                  human journalists, trained in face-to-face classrooms
     How far the anthropologist is permitted to go       or mentored by over-the-shoulder coaches.
with this approach is one of the dilemmas of an-         Physicality is an important component of these re-
thropologists working in software design. The emerg-     lationships: Although people can learn technical skills
ing fields of “design ethnography” and “user-centered     online, they learn integrity and morality only in-
design” have employed ethnographers to better            terpersonally. Making a convincing case for the crit-
understand users’ requirements, and to elicit expert     icality of context for human users, Brown and
knowledge in the construction of expert systems.         Duguid describe six of the context-stripping
However, these efforts are at times compromised by       mechanisms that are supposedly inherent in infor-
a substantial disconnect between the anthropolo-         mation technologies: demassification, decentraliza-
gists’ understanding of requirements and knowledge,      tion, denationalization, despacialization,
and the eng ineers’ understanding of them.               disintermediation, and disaggregation. “These are
Anthropologists see human needs (that is, require-       said to represent forces that, unleashed by infor-
ments) as emergent rather than given, and knowl-         mation technology, will break society down into its
edge (even expert knowledge) as embedded in a            fundamental constituents, primarily individuals and
culturally contingent body of assumptions called         information” (Brown and Duguid 2000, 22). The
“common sense.” Many systems designers, as the late      sum of their argument is that such “6D thinking”
medical anthropologist Diana Forsythe put it, view       is both unrealized and unrealizable. Information
common sense as unproblematic and universal. This        technology does not so much eliminate the social
assumption and others will be discussed below.           context of information, for this is either pointless or
                                                         impossible, as it displaces and decomposes that con-
                                                         text, thus posing new difficulties for users who need
Insights                                                 to turn information into knowledge.
The most important anthropological insight to HCI             Contexts can be high (rich, detailed, and full of
is the emphasis on context for understanding hu-         social cues) or low (impoverished and monochro-
man behavior, including human interaction with cy-       matic), they can be familiar or unfamiliar, and they
bernetic devices. The human organism is unique           can include information channels that are broadband
in its ability to integrate information from a vari-     (a face-to-face conversation) or narrowband (read-
ety of sensory inputs and to formulate an infinite ar-    ing tea leaves, for example). From a human perspec-
ray of potential behavioral responses to these inputs.   tive, all computer interaction, even the most
These arrays of inputs and responses constitute—         multimedia-rich, is narrowband: Sitting where I am,
22 ❘❙❚ BERKSHIRE ENCYCLOPEDIA OF HUMAN-COMPUTER INTERACTION



my computer screen and keyboard occupy no more              ble human action, according to Suchman, it must not
than 25 percent of my field of vision, until I turn my       attempt to anticipate every user state and response (for
head. Looking around, the percentage shrinks to un-         it cannot). Alternatively, a strategy of “real-time user
der 5 percent. The other 95 percent is filled with other     modeling” that incorporates (a) continually updated
work and information storage devices (bookshelves           models of user behavior, (b) detection (and adapta-
and filing cabinets), task aids (charts on the wall),        tion to) diagnostic inconsistencies, (c) sensitivity to
and reminders of relationships: a social context. As a      local conditions, and (d) learning from fault states
low-context device, the computer must be supple-            (such as false alarms and misleading instructions) sug-
mented by these other more social artifacts if it is        gests a better approximation of situated action than
to have human usefulness—that is, if it is to be            preconceived “user models.”
used for knowledge work rather than mere infor-                  Suchman’s findings are based on the concept
mation processing.                                          of “distributed cognition” originally developed by
     Applying the concept of context to a specific tech-     Edwin Hutchins. Instead of understanding cogni-
nological problem, the design of intelligent systems,       tion as information processing (searching, aggre-
Suchman developed a concept of situated action as           gating, parsing, and so on), Hutchins saw mental
an alternative explanation for the rationality of           activity as contextually emergent, using contextual
human action. In place of seeing activity as the ex-        resources (including language and artifacts) as
ecution of a plan (or program), or inversely, seeing        part of an interactive process.
a plan as a retrospective rationalization of activity,           These insights are derived from efforts to use an-
Suchman’s concept of situated action sees plans as          thropological methods in the development of expert
only one of several resources for making sense out          systems and other artificial intelligence devices.
of the ongoing flow of activity.                             Expert systems hold out the hope that in class-
     Human action, or more accurately interaction (for      room instruction, in routine bureaucratic problem
all action is by definition social, even if only one ac-     solving, in medical diagnosis, and in other fields, cer-
tor is physically present), is an ongoing flow of mes-       tain low-level mental tasks could be accomplished
sage input and output. Traditionally social studies have    by computers, in much the same manner as repeti-
assumed that actors have a scheme or mental program         tive manual tasks have been automated. Building
which they are enacting: a plan. In contrast to this,       these systems requires a process of “knowledge ac-
Suchman demonstrates that the rationality of an ac-         quisition” that is viewed as linear and unproblem-
tion is an ongoing construction among those involved        atic. An alternative view, suggested by anthropologist
in the action. The default state of this rationality is a   Jean Lave and computer scientist Etienne Wenger, is
transparent spontaneity in which the participants act       that learning is embedded in (and a byproduct of)
rather than think. Only when the ongoing flow breaks         social relationships and identity formation, and that
down does it become necessary to construct a rep-           people learn by becoming a member of a “commu-
resentation (that is, a plan or image) of what is hap-      nity of practice.”
pening. (Breakdowns, while frequent, are usually easily          The concept of “community of practice” is fur-
repaired.) Language, due to its ability to classify, is a   ther developed by Wenger to describe how experts
powerful resource for constructing such representa-         acquire, share, and use their expertise. Communities
tions, although it is only one of several channels          of practice are groups that share relationships, mean-
that humans use for communication. Using language,          ing, and identity around the performance of some
the participants in an action understand what they          set of tasks, whether processing insurance claims
are doing. Rationality (“understanding what they are        or delivering emergency medicine. The knowledge
doing”) is the achievement rather than the configu-          that they share is embedded in these relationships
ration state of interaction.                                and identities, not something that can be abstracted
     The implications of this for constructing intelli-     and stored in a database (or “knowledge base”).
gent devices (such as expert systems) are profound. In           Anthropologist Marietta Baba has applied these
order for an intelligent device to reproduce intelligi-     concepts along with the concept of “sociotechnical
                                                                                  ANTHROPOLOGY AND HCI ❚❙❘ 23



systems” developed by the Tavistock Institute to ex-       are indicative of a propensity to create normative
amine the response of work groups to the introduc-         closure within any ongoing collectivity.
tion of office automation and engineering systems. At           Both these concepts, of work group cultures and
major corporations she found that efforts to intro-        online communities, point up the importance of
duce new automated systems frequently failed because       culture for computing. As anthropology’s signature
they were disruptive of the work processes, social         concept, culture has an important (if sometimes un-
relationships, identities, and values of the work group,   stated) place in anthropological thinking about hu-
considered as a community of practice.                     man-computer interaction.
     Understanding cognitive activity as distributed
among multiple agents is closely related to the is-
sue of man/machine boundaries, an issue clearly            Culture
of interest to anthropologists. “Cyborg anthropol-         For anthropologists, culture is more profound
ogy” has been an ongoing professional interest at          than simply the attitudes and values shared by a popu-
least since the 1991 publication of anthropologist         lation. As a system of shared understandings, culture
Donna Haraway’s Simians, Cyborgs, and Women.               represents the accumulated learning of a people (or a
Although most cyborg anthropology has focused on           group), rooted in their history, their identity, and their
medical technology (such as imaging systems and            relationship with other groups. Cultures evolve as
artificial organs) rather than on computational tech-       shared projects with other groups. Although they are
nology, the basic concept—of human bodies and              invented and imagined, cultures cannot be conjured
lives becoming increasingly embedded within auto-          up at will, as much of the recent management lit-
mated information (control) circuits—will have in-         erature on corporate culture seems to suggest.
creasing relevance for understanding the adaptation             This is significant, because much of computing
of humans to advanced information technology: As           use is in a corporate or organizational context (even
more and more human faculties, such as memory,             if the organization is virtual). From an anthropo-
skilled manipulation, and interpersonal sensitivity,       logical perspective, it is highly important to note that
are minimalized, disaggregated, and shifted away           much of human-computer interaction is influenced
from the individual organism to automated devices,         either directly, by the regimes of instrumental ra-
the dependence of carbon-based humans on their             tionality in which it takes place, or indirectly, by
artifactual prostheses will increase.                      the fact that it follows protocols established by in-
     Communities also form around technologies.            fluential corporations. Several ethnographies of high-
Technology writer Howard Rheingold has described           tech companies suggest that computerization and
participation in a San Francisco-based usenet as a         the high-tech expectations associated with it are cre-
form of community building. Hakken describes the           ating new corporate cultures: sociologist Gideon
influence of class on the experiences of users with         Kunda and anthropologist Kathleen Gregory-
computing in Sheffield, England. Sociolologist Sherry       Huddleston have described the working atmosphere
Turkle describes the identity experimentation con-         of two high-tech corporations, noting that despite a
ducted by users of multiuser domains. Anthro-              technological aura and emancipatory rhetoric, their
pologist Jon Anderson has examined how Middle              corporate cultures are still mechanisms of control.
Eastern countries have used and adapted the Internet       It should be noted that high tech is less an engineering
with unique methods for unique social goals. These         concept for explaining functionality or performance
include the maintenance of diaspora relationships          than it is an aesthetic conceit for creating auras of
with countrymen scattered around the globe. Online         power and authority.
communities quickly evolve (actually adapt from                 Others have taken note of the fact that com-
surrounding norms) distinctive norms, including            puters create new forms of culture and identity
styles of communication and categories of iden-            and have described numerous microcultures that
tity. Although such collections of norms and val-          have sprung up around such systems as textual data-
ues fall short of full-fledged human cultures, they        banks, engineering design, and online instruction.
24 ❘❙❚ BERKSHIRE ENCYCLOPEDIA OF HUMAN-COMPUTER INTERACTION



The culture of systems developers, as described by         mation society; (2) family, work groups, and personal
Diana Forsythe is particularly notable. Insofar as de-     relationships; (3) public institutions and private cor-
velopers and users have separate and distinctive cul-      porations; (4) communities, both virtual and real;
tural outlooks, there will be a mismatch between           (5) public policy and decision-making; (6) the chang-
their tacit understandings of system functionality         ing shapes of knowledge and culture; and (7) the glob-
and system performance. The frequent experience            alization of the information infrastructure (NSF
of systems not living up to expectations when de-          1996). In many ways this workshop both captured
ployed in the field is less a consequence of poor          and projected forward the anthropological research
engineering than of the fundamental cultural rela-         agenda for understanding the changing social face of
tionships (or disconnects) between developers and          advanced information technology.
users.
    Finally, anthropology’s original interest in the re-
mote and exotic has often taken its attention away         Questions
from the laboratories and highly engineered envi-          Anthropology’s orthogonal viewpoint proposes sev-
ronments in which the most advanced informa-               eral unique questions. Perhaps the first of these is
tion technologies are found. In 2001 Allen Batteau,        the question of control versus freedom. On the
an industrial anthropologist, observed that many           one hand, cybernetic devices exist to create and in-
factories and field installations usually lack the re-      tegrate hierarchies of control, and the fifty-year his-
liable infrastructure of universities or development       tor y of the development of automation has
laboratories. As a consequence, computationally in-        demonstrated the effectiveness of this strategy. On
tensive applications that work so well in the labo-        the other hand, this poses the question of the proper
ratory (or in the movies) crash and burn in the field.      role of a unique node in the control loop, the human
This lack, however, is not simply a matter of these        user: How many degrees of freedom should the user
production environments needing to catch up to the         be allowed? The designer’s answer, “No more than
laboratories: Moore’s Law for nearly forty years           necessary,” can be unsatisfying: Systems that con-
has accurately predicted a doubling of computational       strain the behavior of all their elements limit the
capability every eighteen months, a geometric growth       users’ learning potential.
that outstrips the arithmetic pace of technological            The related concepts of system learning and evo-
diffusion. The dark side of Moore’s Law is that the        lution raise the second outstanding question, which
gap between the technological capabilities of the          has to do with the nature of life. Should systems that
most advanced regions and those of the remote cor-         can evolve, learn from, and reproduce themselves
ners of the human community will continue to grow.         within changing environments be considered “liv-
In 1995 Conrad Kottak, an anthropologist, observed         ing systems”? Studies of artificial life suggest that
that “High technology has the capacity to tear all         they should. The possibility of a self-organizing sys-
of us apart, as it brings some of us closer together”      tem that can replicate itself within a changing en-
(NSF 1996, 29).                                            vironment has been demonstrated by anthropologist
    Many of these observations grew out of a work-         Chris Langston, enlarging our perspective beyond
shop organized by the American Anthropological             the carbon-based naïveté that saw only biological or-
Association and the Computing Research Association         ganisms as living.
called “Culture, Society, and Advanced Information             The final question that this raises, which is the ulti-
Technology.” Held (serendipitously) at the time of the     mate anthropological question, is about the nature
first deployment of graphical Web browsers (an event        or meaning of humanity. Etymologically, anthro-
that as much as any could mark the beginning of the        pology is the “science of man,” a collective term that
popular information revolution), this workshop iden-       embraces both genders, and possibly more. Anthro-
tified seven areas of interest for social research in ad-   pologists always anchor their inquiries on the ques-
vanced information technology: (1) the nature of           tion of “What does it mean to be human?” Otherwise,
privacy, identity, and social roles in the new infor-      their endeavors are difficult to distinguish from com-
                                                                                                           ANTHROPOLOGY AND HCI ❚❙❘ 25



parative psychology, or comparative linguistics, or                          Deal, T., & Kennedy, A. (1999). The new corporate cultures. Reading,
comparative sociology. However, the rise of infor-                              MA: Perseus Books.
                                                                             Emery, F., & Trist, E. (1965). The causal texture of organizational
mation technology has fundamentally challenged some                             environments. Human Relations, 18, 21–31.
received answers to the question of what it means to                         Forsythe, D. (2001). Studying those who study us: An anthropologist in the
be human. What are the human capabilities that                                  world of artificial intelligence. Stanford, CA: Stanford University Press.
computers will never mimic? As Pulitzer-prize-win-                           Garfinkel, H. (1967). Studies in ethnomethodology. Englewood
                                                                                Cliffs, NJ: Prentice-Hall.
ning writer Tracy Kidder asked, Do computers have                            Gregory-Huddleston, K. (1994). Culture conflict with growth:
souls? Will there ever be a computer that meets the                             Cases from Silicon Valley. In T. Hamada & W. Sibley (Eds.),
Turing test—that is, a computer that is indistin-                               Anthropological Perspectives on Organizational Culture. Washington,
                                                                                DC: University Press of America.
guishable from a fully social human individual? More                         Hakken, D. (1999). Cyborgs@Cyberspace: An ethnographer looks to the
specifically, how many generations are required to                               future. New York: Routledge.
evolve a cluster of computers that will (unaided by                          Haraway, D. (1991). Simians, cyborgs, and women—The reinvention of
human tenders) form alliances, reproduce, wor-                                  nature. London: Free Association Books.
                                                                             Hutchins, E. (1994). How a cockpit remembers its speeds. Cognitive
ship a deity, create great works of art, fall into petty                        Science, 19, 265–288.
bickering, and threaten to destroy the planet? As the                        Hutchins, E. (1995). Cognition in the wild. Cambridge, MA: MIT Press.
abilities of silicon-based artifacts to think, feel, learn,                  Kidder, T. (1981). The soul of a new machine. Boston: Little, Brown.
adapt, and reproduce themselves continue to de-                              Kunda, G. (1992). Engineering culture: Control and commitment in a
                                                                                high-tech corporation. Philadelphia: Temple University Press.
velop, the question of the meaning of humanity will                          Langston, C. G. (Ed.). (1989). Artificial life (Santa Fe Institute Studies
probably become the most challenging scientific and                              in the Sciences of Complexity, Proceedings, Volume 6).
philosophical question of the information age.                                  Redwood City, CA: Addison-Wesley.
                                                                             Lave, J., & Wenger, E. (1991). Situated learning: Legitimate peripheral
                                                                                participation. Cambridge, UK: Cambridge University Press.
                                                  Allen W. Batteau           Law, J. (1987). Technology and heterogeneous engineering: The case
                                                                                of Portuguese expansion. In W. E. Bijker, T. P. Hughes, &
See also Ethnography; Sociology and HCI; Social                                 T. Pinch (Eds.), The social construction of technological systems.
Psychology and HCI                                                              (pp. 111–134). Cambridge, MA: MIT Press.
                                                                             Linde, C. (1988). The quantitative study of communicative success:
                                                                                Politeness and accidents in aviation discourse. Language in Society,
                                                                                17, 375–399.
                                                                             Moore, G. (1965, April 19). Cramming more components onto in-
FURTHER READING                                                                 tegrated circuits. Electronics.
                                                                             Nardi, B. A. (1993). A small matter of programming: Perspectives on end
                                                                                user computing. Cambridge, MA: MIT Press.
Anderson, J. (1998). Arabizing the Internet. Emirates Occasional Papers
                                                                             National Science Foundation. (1996). Culture, society, and advanced
   # 30. Abu Dhabi, United Arab Emirates: Emirates Center for
                                                                                information technology (Report of a workshop held on June 1–2,
   Strategic Studies and Research.
                                                                                1995). Washington, DC: U. S. Government Printing Office.
Anderson, J., & Eikelman, D. (Eds.). (2003). New media in the Muslim
                                                                             Perin, C. (1991). Electronic social fields in bureaucracies.
   world: The emerging public sphere (Indiana Series in Middle East
                                                                                Communications of the ACM, 34(12), 74–82.
   Studies). Bloomington: Indiana University Press.
                                                                             Rheingold, H. (1993). The virtual community: Homesteading on the
Baba, M. L. (1995). The cultural ecology of the corporation: Explaining
                                                                                electronic frontier. Reading, MA: Addison-Wesley.
   diversity in work group responses to organizational transforma-
                                                                             Star, S. L. (Ed.). (1995). The cultures of computing. Oxford, UK:
   tion. (1995). Journal of Applied Behavioral Science, 31(2), 202–233.
                                                                                Blackwell Publishers.
Baba, M. L. (1999). Dangerous liaisons: Trust, distrust, and infor-
                                                                             Stone, A. R. (1995). The war of desire and technology at the close of
   mation technology in American work organizations. Human
                                                                                the mechanical age. Cambridge, MA: MIT Press.
   Organization, 58(3), 331–346.
                                                                             Suchman, L. (1987). Plans and situated actions: The problem of human-
Batteau, A. (2000). Negations and ambiguities in the cultures of or-
                                                                                machine communication. Cambridge, UK: Cambridge University
   ganization. American Anthropologist, 102(4), 726–740.
                                                                                Press.
Batteau, A. (2001). A report from the Internet2 ‘Sociotechnical Summit.’
                                                                             Turkle, S. (1995). Life on the screen: Identity in the age of the Internet.
   Social Science Computing Review, 19(1), 100–105.
                                                                                New York: Simon & Schuster.
Borofsky, R. (1994). Introduction. In R. Borofsky (Ed.), Assessing cul-
                                                                             Wenger, E. (1998). Communities of practice: Learning, meaning, and
   tural anthropology. New York: McGraw-Hill.
                                                                                identity. Cambridge, UK: Cambridge University Press.
Brown, J. S., & Duguid, P. (2000). The social life of information. Boston:
   Harvard Business School Press.
Callon, M. (1980). The state and technical innovation: A case study
   of the electrical vehicle in France. Research Policy, 9, 358–376.
26 ❘❙❚ BERKSHIRE ENCYCLOPEDIA OF HUMAN-COMPUTER INTERACTION



                                                            German scholar Rodolpho von Ihering. These pro-
       ANTHROPOMETRY                                        posals called upon German anatomists and anthro-
                                                            pologists to reinvestigate craniometric (relating to
The word anthropometry, which means “the meas-              measurement of the skull) and anthropometric meas-
urement of the physical characteristics and physical        urement methods. The German Anthropological
abilities of people,” is derived from the Greek words       Society convened in Munich and Berlin during the
anthropo meaning “human being” and metry mean-              1870s and early 1880s to establish what anthro-
ing “measure.” Physical characteristics, also called        pometrist J. G. Garson and others have called the
“structural dimensions,” include such aspects as            “Frankfort Agreement” of 1882. This agreement in-
heights, widths, depths, and body segment circum-           troduced craniometric methods distinct from the
ferences. Physical abilities, also called “functional di-   predominant French methods and established a new
mensions,” include such aspects as grip, push and           nomenclature and measurement methods. The ex-
pull strength, reaching capabilities, fields of vision,      istence of the French and German schools only
and functional task performance.                            further cemented the belief that international con-
    Anthropologists, clinicians, and engineers use          sensus on methods, nomenclature, and measure-
anthropometric information in a variety of ways.            ments was needed.
For engineers, in particular, anthropometry provides             During the early twentieth century people at-
information that can be used for the design of oc-          tempted to develop an international consensus on
cupational, pubic, and residential environments. The        the nomenclature of body dimensions and meas-
information can also be used for the design of tools,       urement methods. In 1906, at the Thirteenth
protective head gear, clothing, and workstation equip-      International Congress of Prehistoric Anthropology
ment. Doorway widths, tool handle lengths and cir-          and Archaeology in Monaco, an international agree-
cumferences, ranges of clothing sizes, and the location     ment of anthropometry took form. This congress
of displays and controls on workstations are some           and the Fourteenth International Congress in
of the design applications. Anthropometry also pro-         Geneva, Switzerland, in 1912 began to formalize the
vides information about body segment center of mass         body of anthropometric work. The foundations of
and joint center of rotation characteristics that is        a normative framework and a standardization of an-
used for biomechanical modeling (the study of joint         thropometric measurement had been laid and trans-
forces and torques on the body).                            lated into French, German, and English by 1912. This
                                                            framework standardized anthropometric measure-
                                                            ments on both skeletal and living human subjects.
A Brief History                                             Since 1912 several works by Hrdlicka, Rudolf
Although anthropometry was applied when Greek               Marting, and James Gaven have increased the aware-
and Egyptian artists created standards (canons)             ness of anthropometry and its uses and added to
for the human form centuries ago, not until the             its scientific rigor.
nineteenth century were thought and dialogue on                  After the initial congresses, people attempted
anthropometry organized. Early work in anthro-              to establish consensus throughout the twentieth cen-
pometry focused on the human anatomy, racial char-          tury. Congresses meeting under the name of Hrdlicka
acteristics, skeletal remains, and human growth.            convened on the topic of anthropometry and meas-
Among the noteworthy work documented by phys-               urement methods. Other congresses aimed to create
ical anthropologist Ales Hrdlicka was that of French        standards and databases for general use. During the
anthropologist Paul Pierre Broca and the Belgian sci-       late twentieth century authors such as Bruce
entist Adolphe Quetelet. During the mid-eighteenth          Bradtmiller and K. H. E. Kroemer chronicled these
century Quetelet used statistics to describe anthro-        congresses and offered unique ways to manage
pometric information. Shortly after the Franco-             anthropometric data. During recent years the
Prussian War of 1870, a growing emphasis on                 International Standardization Organization (ISO)
individualism was evident in the proposals of the           technical committee on ergonomics published ISO
                                                                                     ANTHROPOMETRY ❚❙❘ 27



7250: Basic Human Body Measurements for Technical        circumferences, and tape measures are used to meas-
Design (1996) to standardize the language and meas-      ure other circumferences such as the distance
urement methods used in anthropometry and ISO            around the waist. Scales are used to measure body
15535: General Requirements for Establishing an          weight. Photographs and video are used to meas-
Anthropometric Database (2003) to standardize the        ure body dimensions in two dimensions. One
variables and reporting methods of anthropometric        method uses grids that are attached behind and
studies.                                                 to the side of the person measured. Photographs
                                                         are then taken perpendicular to the grids, and the
                                                         space covered by the person in front of the grids
Structural Anthropometric                                can be used to estimate body segment heights,
                                                         widths, and depths. A variant of this method uses
Measurement Methods                                      digital photography for which an anthropometric
Structural anthropometric measurement methods            measurement is obtained by comparing the num-
require a person to be measured while standing or        ber of pixels (small discrete elements that together
sitting. Anatomical landmarks—observable body            constitute an image, as in a television or com-
features such as the tip of the finger, the corner of    puter screen) for a dimension to the number of pix-
the eye, or the bony protrusion of the shoulder          els of a reference object also located in the digital
known as the “acromion process”—standardize              photograph.
the locations on the body from which measurements            Attempts to develop three-dimensional com-
are made. The desire to achieve consistent measure-      puter human models with conventional anthropo-
ments has led to the use of standardized measure-        metric data reveal that limitations exist, such as the
ment postures held by people who are being               uncertainty about three-dimensional definition of
measured. The anthropometric standing posture re-        key points on the body surface, locations of cir-
quires the person to hold the ankles close together,     cumferences, and posture. These limitations have re-
standing erect, arms relaxed and palms facing me-        sulted in the development of more sophisticated
dially (lying or extending toward the median axis of     three-dimensional anthropometric measurement
the body) or anteriorly (situated before or toward       methods.
the front), the head erect and the corners of the eyes       Digital anthropometry is the use of digital and
aligned horizontally with the ears. The anthropo-        computerized technology in the collection of infor-
metric seated posture requires the person to be seated   mation about body size and physical ability. In this
erect on a standard seating surface. The elbows          use, computers are responsible for the actual collec-
and knees are flexed 90 degrees. The palms face me-       tion of anthropometric data and are not relegated
dially with the thumb superior (situated above or        solely to data analysis or storage. Digital anthro-
anterior or dorsal to another and especially a cor-      pometry varies greatly from conventional anthro-
responding part) to the other digits.                    pometry. This variation has changed the nature of
     Structural dimensions include the distances be-     anthropometry itself for both the anthropometrist
tween anatomical landmarks, the vertical distance        and the experimental context in which measurements
from a body landmark to the floor, and the cir-          are taken. Human factors engineer Matthew Reed
cumferences of body segments and are measured            and colleagues have identified some of the potential
with a variety of instruments. Among the most com-       benefits of digital anthropometry:
mon instruments is the anthropometer, which is
a rod and sliding perpendicular arm used to meas-         ■ The capacity to assemble more accurate mod-
ure heights, widths, and depths. A spreading caliper        els of human form, dimensions, and postures
having two curved arms that are hinged together is        ■ The capacity to evaluate multiple body dimen-
sometimes used to measure segment widths and                sions simultaneously
depths defined by the distance between the tips of         ■ The capacity to measure the human and the en-
the arms. Graduated cones are used to measure grip          vironment together
28 ❘❙❚ BERKSHIRE ENCYCLOPEDIA OF HUMAN-COMPUTER INTERACTION



  ■ The improved description of joint centers of ro-       those that employ photography and potentiometry
    tation and movement in three dimensions                as described above, or methods that require an in-
  ■ The capacity to make corrections to dimensions         dividual to mark with a hand-held pen or pencil the
    or create new dimensions after measurements            maximum or comfortable reach locations on a
    have been recorded                                     vertical or horizontal grid.
                                                                Electromagnetic and video-based motion analy-
    Laser scanning is often used in digital anthro-
                                                           sis systems provide new measures of physical abili-
pometry because it allows excellent resolution of
                                                           ties related to the way people move (kinematics) and
the morphological (relating to the form and structure
                                                           can be used with other types of instrumentation,
of an organism or any of its parts) features of the hu-
                                                           such as force plates (hardware that measure the force
man body and can be completed rapidly. Laser
                                                           applied to it), to provide biomechanical (the me-
scanning produces accurate three-dimensional rep-
                                                           chanics of biological and especially muscular ac-
resentations of the complex body surfaces, and
                                                           tivity) information or measures of balance. These
most protocols (detailed plans of a scientific or med-
                                                           systems allow positions of body landmarks to be
ical experiment, treatment or procedure) require
                                                           tracked over time during a physical activity. The data
the placement of surface markers on the body to
                                                           can be evaluated statistically or can serve as an ex-
ensure the proper location of bony protrusions that
                                                           ample of a human task simulation. Such methods of
are used as measurement landmarks beneath the sur-
                                                           data collection allow more lifelike dynamic digital
face of the skin. Other protocols using laser scans have
                                                           human models that can be used to evaluate human
morphological extraction algorithms (procedures for
                                                           performance in virtual environments. However, use
solving a mathematical problem in a finite number of
                                                           of these methods is expensive and time consuming.
steps that frequently involve repetition of an opera-
tion) to estimate landmark locations based on mor-
phological features.
    Potentiometry can also be used to collect digital      Measurement Consistency
anthropometric measurements. Electromechanical             and Variation
potentiometric systems allow the measurer to man-
                                                           Anthropometric measurements are recordings of
ually digitize points in three-dimensional space. The
                                                           body dimensions and physical abilities that are sub-
measurer guides a probe tip manually to render dis-
                                                           ject to variability. No “correct” measurement exists
crete points or body surface contours.
                                                           because a measurement is simply an observation
                                                           or recording of an attribute that is the cumulative
                                                           contribution of many factors.
Functional Performance                                          Anthropometric studies have investigated the
                                                           topic of measurement consistency in relation to
Measurements                                               intrinsic qualities of variability within a given meas-
Conventional functional performance measurements           urement. J. A. Gavan (1950) graded anthropome-
include grip, push, and pull strength, and reaching        try dimensions in terms of consistencies seen through
abilities. For grip strength measurement, an indi-         expert anthropometrists and concluded that “con-
vidual is required to squeeze for several seconds at       sistency increased as: the number of technicians de-
maximum effort a hand dynamometer (a force meas-           creased, the amount of subcutaneous [under the
urement device) set at one or more grip circumfer-         skin] tissue decreased, the experience of the tech-
ences. For the measurement of push and pull                nician increased, and as the landmarks were more
strength, an individual usually holds a static (un-        clearly defined” (Gavan 1950, 425). Claire C. Gordon
changing) posture while either pushing on or pulling       and Bruce Bradtmiller (1992), Charles Clauser and
against a force gauge at a maximum effort over             associates (1998), Gordon and associates (1989), and
several seconds. An individual’s reaching abilities can    others have also studied intra- and interobserver er-
be evaluated with a number of methods, including           ror contributions in anthropometric measurements,
                                                                                                   ANTHROPOMETRY ❚❙❘ 29



including the contributions of different measure-                      modeling methods. The correlation between two di-
ment instruments and the effects of breathing cy-                      mensions provides a measure of how strongly two
c l e s . O t h e r re s e a rch e r s , s u ch a s Ka t h e r i n e   dimensions covary linearly. When two measurements
Brooke-Wavell and colleagues (1994), have evalu-                       are highly correlated the values of one measurement
ated the reliability of digital anthropometric meas-                   can be used to predict the values of another in a
urement systems. These evaluations have brought                        regression analysis, therefore reducing the total num-
about an awareness of anthropometric reliability and                   ber of measurements needed to construct a com-
error as well as acceptable levels of reliability.                     prehensive set of anthropometric tables and human
      Anthropometric data typically are collected for                  models based on partially extrapolated data. When
large samples of populations to capture distribu-                      combinations of anthropometric dimensions are
tional characteristics of a dimension so that it is rep-               considered simultaneously in the evaluation of a
resentative of a target population. Many sources of                    product or environment, mockups and task trialing
anthropometric variability exist within populations.                   involving people or simulation approaches using dig-
Men and women differ greatly in terms of structural                    ital human modeling of people are required.
and functional anthropometric dimensions.
Additionally, the anthropometric dimensions of peo-
ple have changed systematically through time. Today’s                  Important Data Sources
people are generally taller and heavier than those                     The most comprehensive anthropometric studies
of previous generations, perhaps because of improved                   have focused on military personnel, at least in part
availability and nutrition of food in developed coun-                  due to the need for the military to have information
tries. Of course, a person’s body size also changes                    to provide well-designed uniforms, equipment, land
through time, even throughout the course of                            vehicles, and aircraft. Perhaps one of the most
adulthood. As a person ages, for example, his or                       comprehensive studies was the 1988 U.S. Army
her height decreases. Other sources of anthropo-                       Anthropometric Survey (ANSUR), which summa-
metric variability include ethnicity, geography, and                   rized 132 dimensions of approximately nine thou-
occupational status.                                                   sand army personnel.
      The distribution characteristics of an anthropo-                     One of the most inclusive sources of civilian an-
metric dimension are often reported for different                      thropometric data is a U.S. National Aeronautics and
categories of age and gender, and sometime for                         Space Administration (NASA) technical report pro-
different ethnicities or countries. Because the vari-                  duced by the staff of the Anthropology Research
ability of anthropometric dimensional values within                    Project in 1978. This report contains anthropo-
such subgroups often takes the shape of a Gausian                      metric data across a variety of civilian and military
(bell-shaped) distribution, the mean deviation and                     populations for a large number of anthropometric
standard deviation of the sample data are often used                   variables, including information about the mass dis-
to describe the distributional characteristics of a di-                tribution of body segments.
mension. The percentile value—the value of a di-                           More recently, the Civilian American and
mension that is greater than or equal to a certain                     European Surface Anthropometry Resource (CAE-
percentage of a distribution—also provides useful                      SAR) project used laser scanning to collect the body
information. For example, the fifth and ninety-fifth                     surface contours and sizes of approximately twenty-
percentiles of a dimensional value define the outer                     four hundred North American and two thousand
boundaries of the 90 percent midrange of a popu-                       European civilians from 1998 to 2000. Measurements
lation distribution that might enable a designer to                    were recorded with people in standing, standardized
develop an adjustable consumer product or envi-                        seated, and relaxed seated postures. Thousands of
ronment feature that can accommodate 90 percent                        points that define the location of the body’s sur-
or more of the target population.                                      face were collected with each scan, providing ex-
      Multivariate data analysis includes the use of cor-              tremely accurate three-dimensional representations
relation and regression analyses, as well as human                     of the body surface contours for individual human
30 ❘❙❚ BERKSHIRE ENCYCLOPEDIA OF HUMAN-COMPUTER INTERACTION



models that can be used to evaluate the fit of a prod-     that apply anthropometric data to the development
uct or of the person in an environment. Because           of design guidelines. These publications include ISO
markers are also placed over key body landmarks,          14738 Safety of Machinery—Anthropometric
conventional descriptive analysis of dimensions has       Requirements for the Design of Workstations at
also been performed. CAESAR is the largest and most       Machinery (2002), ISO 15534 Ergonomic Design for
valuable anthropometric data source of its kind.          the Safety of Machinery (2000), and ISO 9241
                                                          Documents on the Ergonomic Requirements for Office
                                                          Work with Visual Display Terminals (1992–2001).
Using Anthropometric Data in Design                       The latter publications were developed to improve
Conventional use of anthropometric data in de-            the fit between people and their computers at work.
sign requires determining (1) the population for
which a design in intended, known as the “target
population,” (2) the critical dimension or dimen-         Future Research
sions of the design, (3) appropriate anthropometric       A major challenge of future research is how to sum-
data source, (4) the percentage of the population         marize and interpret the information-rich but com-
to be accommodated by the design, (5) the portion         plex three-dimensional data that accompany the new
of the distribution that will be excluded, usually the    methods of measurement described here. New meth-
largest and/or smallest values of the distribution, and   ods of three-dimensional measurement of body
(6) the appropriate design values through the use of      dimensions such as whole-body scanning provide
univariate or bivariate statistical methods.              new opportunities to move conventional univariate
    Conventional application of anthropometric            anthropometric applications to complete three-di-
data, however, is not able to address the design prob-    mensional static human models that can be used
lems that require the evaluation of many design char-     to evaluate design in new ways. Motion analysis
acteristics simultaneously. Multivariate analysis using   methods in dynamic human modeling also pro-
mockups and task trialing requires recruiting peo-        vide a powerful tool to improve our understanding
ple with the desired range of body size and ability       of the functional abilities of people. The reliability,
and assessing human performance during the sim-           accuracy, and applications of many of these anthro-
ulation, such as judging whether people can reach         pometric measurement methods, however, have yet
a control or easily see a display for a particular de-    to be fully explored.
sign. Static and dynamic digital human modeling                 Perhaps what is most needed is simply more in-
approaches require manipulating models of various         formation about the physical dimensions and abili-
sizes in virtual environments to assess the person-       t i e s i n m o r e d i ve r s e u s e r g r o u p s . L a c k o f
design fit. Analysis methods for dynamic digital hu-       anthropometric information severely limits the use
man modeling approaches are still in their infancy        of anthropometry in the design of living and work-
due to the limited amount of studies recording the        ing spaces that can be used by diverse populations.
needed information and the complicated nature of                U.S. government agencies, particularly the U.S.
the data.                                                 Architectur al and Tr anspor tation Bar r iers
    A variety of fields uses anthropometric data, in-      Compliance Board (Access Board) and the U.S.
cluding anthropology, comparative morphology, hu-         Department of Education’s National Institute on
man factors engineering and ergonomics, medicine,         Disability and Rehabilitation Research (NIDRR), re-
and architectural design. Additionally, digital an-       cently have started to address the information gap
thropometry has been used outside of scientific and        by studying the physical abilities of people with dis-
research endeavors, as seen in the application of a       abilities, such as people who use wheelchairs.
new suit-making technology for Brooks Brothers                  However, much work remains to be done. In par-
(known as “digital tailoring).                            ticular, the need for anthropometric data to in-
    The International Organization for Stan-              form the design of occupational, public, and
dardization has published numerous publications           residential environments of the elderly is expected
                                                                                                                                                     ANTHROPOMETRY ❚❙❘ 31



to increase substantially as the proportion of the eld-                                                 display terminals (VDTs), (ISO Standard 9241). Geneva, Switzerland:
erly in the population continues to increase dra-                                                       International Organization for Standardization.
                                                                                                     International Organization for Standardization. (Ed.). (1996). Basic
matically during the years to come.                                                                     human body measurements for technical design (ISO Standard 7250).
                                                                                                        G e n e v a , Sw i t ze r l a n d : In te r n a t i o n a l O r g a n i z a t i o n f o r
                                  Victor Paquet and David Feathers                                      Standardization.
                                                                                                     International Organization for Standardization. (Ed.). (2000).
                                                                                                        Ergonomic design for the safety of machinery (ISO Standard 15534).
See also Motion Capture                                                                                 G e n e v a , Sw i t ze r l a n d : In te r n a t i o n a l O r g a n i z a t i o n f o r
                                                                                                        Standardization.
                                                                                                     International Organization for Standardization. (Ed.). (2002).
                                                                                                        Safety of machinery—Anthropometric requirements for the design
FURTHER READING                                                                                         of workstations at machinery (ISO Standard 14738). Geneva,
                                                                                                        Switzerland: International Organization for Standardization.
                                                                                                     International Organization for Standardization. (Ed.). (2003). General
Annis, J. F. (1989). An automated device used to develop a new 3-D
                                                                                                        requirements for establishing an anthropometric database (ISO
     database for head and face anthropometry. In A. Mital (Ed.),
                                                                                                        Standard 15535). Geneva, Switzerland: International Organization
     Advances in industrial ergonomics and safety (pp. 181–188). London:
                                                                                                        for Standardization.
     Taylor & Francis.
                                                                                                     Kroemer, K. H. E., Kroemer, H. J., & Kroemer-Elbert, K. E. (1997).
Annis, J. F., Case, H. W., Clauser, C. E., & Bradtmiller, B. (1991).
                                                                                                        Engineering anthropometry. In K. H. E. Kroemer (Ed.), Engineering
     Anthropometry of an aging work force. Experimental Aging
                                                                                                        physiology (pp. 1–60). New York: Van Nostrand Reinhold.
     Research, 17, 157–176.
                                                                                                     Marras, W., & Kim, J. (1993). Anthropometry of industrial popula-
Brooke-Wavell, K., Jones, P. R. M., & West, G. M. (1994). Reliability
                                                                                                        tions. Ergonomics, 36(4), 371–377.
     and repeatability of 3-D body scanner (LASS) measurements com-
                                                                                                     Molenbroek, J. (1987) Anthropometry of elderly people in the
     pared to anthropometry. Annals of Human Biology, 21, 571–577.
                                                                                                        Netherlands: Research and applications. Applied Ergonomics, 18,
Clauser, C., Tebbetts, I., Bradtmiller, B., McConville, J., & Gordon,
                                                                                                        187–194.
     C. (1998). Measurer’s handbook: U.S. Army anthropometric sur-
                                                                                                     Paquet, V. (Ed.). (2004). Anthropometry and disability [Special issue].
     vey (Technical Report No. TR-88/043). Natick, MA: U.S. Army
                                                                                                        International Journal of Industrial Ergonomics, 33(3).
     Natick Research, Development and Engineering Center.
                                                                                                     Paquette, S., Case, H., Annis, J., Mayfield, T., Kristensen, S., & Mountjoy,
Damon, A., & Stout, H. (1963). The functional anthropometry of old
                                                                                                        D. N. (1999). The effects of multilayered military clothing ensembles
     men. Human Factors, 5, 485–491.
                                                                                                        on body size: A pilot study. Natick, MA: U.S. Soldier and Biological
Dempster, W. T., Gabel, W. C., & Felts, W. J. L. (1959). The anthro-
                                                                                                        Chemical Command Soldier Systems Center.
     pometry of the manual work space for the seated subject. American
                                                                                                     Reed, M., Manary, M., Flannagan, C., & Schneider, L. (2000). Effects
     Journal of Physical Anthropometry, 17, 289–317.
                                                                                                        of vehicle interior geometry and anthropometric variables on auto-
Eastman Kodak Company. (2003). Ergonomic design for people at work
                                                                                                        mobile driving posture. Human Factors, 42, 541–552.
     (2nd ed.). New York: Wiley.
                                                                                                     Reed, M., Manary, M., & Schneider, L. (1999). Methods for measur-
Garson, J. (1885). The Frankfort Craniometric Agreement, with
                                                                                                        ing and representing automobile occupant posture (SAE Technical
     critical remarks thereon. Journal of the Anthropological Institute of
                                                                                                        Paper No. 1999-01-0959). Warrendale, PA: Society of Automotive
     Great Britain and Ireland, 14, 64–83.
                                                                                                        Engineers.
Gavan, J. (1950). The consistency of anthropometric measure-
                                                                                                     Robinette, K. (1998). Multivariate methods in engineering anthro-
     ments. American Journal of Physical Anthropometry, 8, 417–426.
                                                                                                        pometry. In Proceedings of the Human Factors and Ergonomics
Gordon, C., & Bradtmiller, B. (1992). Interobserver error in a large
                                                                                                        Society 42nd annual meeting (pp. 719–721). Santa Monica, CA:
     scale anthropometric survey. American Journal of Human Biology,
                                                                                                        Human Factors and Ergonomics Society.
     4, 253–263.
                                                                                                     Robinette, K. (2000). CAESAR measures up. Ergonomics in Design,
Gordon, C., Bradtmiller, B., Clauser, C., Churchill, T., McConville,
                                                                                                        8(3), 17–23.
     J., Tebbetts, I., & Walker, R. (1989). 1987–1988 anthropometric sur-
                                                                                                     Roebuck, J., Kroemer, K. H. E., & Thomson, W. (1975). Engineering
     vey of U.S. Army personnel: Methods and summary statistics
                                                                                                        anthropometry methods. New York: Wiley.
     (Technical Report No. TR-89/027). Natick, MA: U.S. Army Natick
                                                                                                     Steenbekkers, L., & Molenbroek, J. (1990). Anthropometric data of
     Research, Development and Engineering Center.
                                                                                                        children for non-specialized users. Ergonomics, 33(4), 421–429.
Haddon, A. (1934). The history of anthropology. London: Watts & Co.
                                                                                                     Ulijaszek, S., & Mascie-Taylor, C. G. N. (Eds.). (1994). Anthropometry:
Hobson, D., & Molenbroek, J. (1990). Anthropometry and design for
                                                                                                        The individual and the population. Cambridge, UK: Cambridge
     the disabled: Experiences with seating design for cerebral palsy
                                                                                                        University Press.
     population. Applied Ergonomics, 21(1), 43–54.
Hoekstra, P. (1997). On postures, percentiles and 3D surface an-
     thropometry. Contemporary Ergonomics (pp. 130–135). London:
     Taylor & Francis.
Hrdlicka, A. (1918). Physical anthropology; its scope and aims, etc.
     American Journal of Physical Anthropometry, 1, 3–23.
I n t e r n a t i o n a l O r g a n i z a t i o n f o r S t a n d a rd i z a t i o n . ( E d . ) .
     (1992–2003). Ergonomics requirements for office work with visual
32 ❘❙❚ BERKSHIRE ENCYCLOPEDIA OF HUMAN-COMPUTER INTERACTION



                                                           manipulate strategy is also goal-directed because it
        APPLICATION USE                                    can be used to complete the task of drawing three
                                                           arched windows.
          STRATEGIES                                           The definition of a strategy given above subsumes
                                                           more limited strategy definitions used in fields as di-
Strategies for using complex computer applications         verse as business management and cognitive psy-
such as word-processing programs and computer-             chology. These definitions may be stated in terms of
aided drafting (CAD) systems are general and goal-         time (they may define strategy as a long-term plan
directed methods for performing tasks. These               for achieving a goal), the existence of alternate meth-
strategies are important to identify and learn be-         ods (they may consider a strateg y to be any
cause they can make users more efficient and effec-         method that is nonobligatory), or performance out-
tive in completing their tasks, and they are often         comes (they may define a strategy as a method that
difficult to acquire just by knowing commands on            results in a competitive advantage). However, ex-
an interface.                                              cluding these particulars (time, existence of alter-
    To understand strategies for using computer ap-        nate methods, and performance outcomes) from the
plications, consider the task of drawing three iden-       definition of strategy enables us to describe strate-
tical arched windows in a CAD system. As shown in          gies in a more encompassing way, irrespective of
Figure 1A, one way to perform the task is to draw all      whether they are short term or long term, unique or
the arcs across the windows, followed by drawing all       one of many, or efficient or inefficient.
the vertical lines, followed by drawing all the hori-
zontal lines. Another way to perform the same task
(Figure 1B) is to draw all the elements of the first       The Costs and Benefits
window, group the elements and then make three
copies of the grouped elements.                            of Using Strategies
    The first method is called sequence-by-operation        Although the two strategies shown in Figure 1 achieve
because it organizes the drawing task by performing        the same goal, different costs and benefits are asso-
one set of identical operations (in this case draw arc),   ciated with each one’s use. By drawing all the arcs
followed by performing the next set of similar op-         before the lines, the sequence-by-operation strategy
erations (in this case draw line). The second method       reduces the cost of switching between the draw arc,
is called detail-aggregate-manipulate because it or-       and the draw line commands. Furthermore, the strat-
ganizes the task by first detailing all the elements       egy uses simple commands that are useful for per-
of the first object (in this case drawing the parts of      forming a large set of tasks. Therefore, the short-term
the first window), aggregating the elements of the         learning cost of using this strategy is small. However,
first object (in this case grouping all the parts of       because the user is constructing every element in the
the first window), and then manipulating that ag-           drawing, the performance cost (measured in terms
gregate (in this case making two copies of the             of time and effort) can become large when draw-
grouped elements of the first window). Both the            ing repeated elements across many tasks, especially
methods are strategies because they are general            in the long term. In contrast, the detail-aggregate-
and goal-directed. For example, the detail-aggregate-      manipulate strategy requires the user to draw the el-
manipulate strategy is general because it can be used      ements of only one window, and makes the computer
to create multiple copies of sets of objects in a          construct the rest of the windows using the group,
wide range of applications. The above example              and copy commands. For a novice CAD user, the
was for a CAD application, but the same strategy           short-term learning cost for the detail-aggregate-
could be used to create many identical paragraphs          manipulate strategy involves learning the group and
for address labels in a word-processing appli-             copy commands and how to sequence them.
cation, such as Microsoft Word. The detail-aggregate-      However, as is common in the use of any new tool,
                                                                                          APPLICATION USE STRATEGIES ❚❙❘ 33



this short-term learning cost is amortized over the            A. Sequence-by-Operation Strategy
long term because of the efficiency gained over many
invocations of the strategy. This amortization there-
fore lowers the overall performance cost.
    Research has shown that strategies like detail-ag-           1.Draw arcs.                   2. Draw vertical                  3. Draw horizontal
gregate-manipulate can save users between 40 per-                                               lines.                            lines.
cent and 70 percent of the time to perform typical
drawing tasks, in addition to reducing errors.                 B. Detail-Aggregate-Manipulate Strategy
Furthermore, with properly designed strategy-based
training, such strategies can be taught to novice com-
puter users in a short amount of time. For users who
care about saving time and producing accurate draw-            1.Draw arc. 2.Draw lines. 3. Group lines. 4. Copy group.
ings, learning such strategies can therefore make them                      detail                        aggregate                   manipulate
more efficient (save time) and more effective (reduce
errors) with relatively short training.                    Source: Bhavnani, S. K., John, B. E. (1996). Exploring the unrealized potential of computer-aided drafting.
                                                           Proceedings of CHI’96, 337. Copyright 1996 ACM, Inc. Reprinted by permission.



                                                           FIGURE 1. Two strategies to perform the 3-window
                                                           drawing task.
A Framework That Organizes
Strategies for Using Complex
                                                           broad in scope because the powers they exploit are
Computer Applications                                      offered by a large range of computer applications
Given the important role that strategies can play in       such as authoring and information retrieval appli-
improving overall productivity, researchers have at-       cations. Other strategies are narrower in scope and
tempted to identify and organize strategies for com-       applicable to a smaller range of computer applica-
puter application use, such as authoring and               tions such as only to word processors.
information retrieval applications. Frameworks to
organize strategies have suggested the design of:          [Large-Scope Strategies
(1) training that teaches the strategies in a system-      Given the ubiquity of graphical user interfaces (GUIs)
atic way, (2) new systems that provide effective and       across computer applications, most useful computer
efficient strategies to users with little experience, and   applications require some interaction with a visual
(3) evaluation methods to ensure that designers con-       interface. Such computer applications offer the power
sistently offer the commands for using efficient and       of visualization, that is, the power to selectively view
effective strategies.                                      information on the screen. For example, a common
     One useful way to organize strategies is based on     word-processing task is to compare information from
the general capabilities of computer applications that     one part of a document with information in another
the strategies exploit. For example, the detail-aggre-     part of the document. When these two parts of the
gate-manipulate strategy described in Figure 1 ex-         document cannot fit simultaneously on the screen,
ploits the iterative power of computers; it makes the      the user can perform the comparison task in several
computer (instead of the user) perform the repeti-         ways. One way is to scroll back and forth between
tious task of copying the elements multiple times.         the relevant parts of the document. This method is
Strategies have also been identified that exploit other     time-consuming and error-prone because it requires
powers of computers, such as the powers of propa-          the user to remember the information that is not vis-
gation, organization, and visualization.                   ible. Another way to perform the same compari-
     Another way to organize strategies is by the scope    son task is to first bring together on the computer
of their use. For example, some strategies are             screen the two relevant parts of the document, before
34 ❘❙❚ BERKSHIRE ENCYCLOPEDIA OF HUMAN-COMPUTER INTERACTION



  Large Scope                                                                               Computer Applications
 Visualization Strategies



                                                                                                              Other computer
   Medium Scope                             Authoring Applications
                                                                                                             applications (e.g.
   Iteration Strategies                                                                                    information retrieval )
   Propagation Strategies


        Word Processors          Spreadsheets             Drawing Systems          Other authoring
                                                                                  applications (e.g.
     Small Scope            Small Scope                Small Scope                 Web authoring)
      ext
     T Transformation       Formula Decomposition      Graphic Precision
     Strategies             Strategies                 Strategies




F I G U R E 2 . Strategies have been identified to exploit different powers of computers at different scopes levels. Large
scope strategies are useful to many classes of computer applications, such as authoring and information retrieval
applications. Medium scope strategies apply to a single class of computer applications, such as authoring appli-
cations. Small scope strategies apply to a single sub-class of applications, such as only to word processors. The
dotted lines represent how future strategies can be included in the framework.



comparing them. The information can be brought                           Second, there are strategies that exploit the power
together on the screen by different commands, such                   of propagation provided by authoring applications.
as by opening two windows of the same document                       The power of propagation enables users to set up de-
scrolled to the relevant parts of the document, or by                pendencies between objects, such that modifications
using the split window command in Microsoft Word                     automatically ripple through to the dependent ob-
to view two parts of the document simultaneously.                    jects. For example, often users have to change the
    In addition to being useful for word-process-                    font and size of headings in a document to conform
ing tasks, this visualization strategy is also useful                to different publication requirements. One way to
when one is drawing a complex building in a CAD                      perform this task is to make the changes manually.
system, or when one is comparing information from                    This is time-consuming, especially when the docu-
two different webpages when retrieving information                   ment is long, and error-prone, because certain head-
on the Web. Hence strategies that exploit the                        ings may be missed or incorrectly modified. A more
power of visualization have wide scope, spanning                     efficient and effective method of performing the same
many different classes of computer applications.                     task is to first make the headings in a document
                                                                     dependent on a style definition in Microsoft Word.
Medium-Scope Strategies                                              When this style definition is modified, all depend-
While visualization strategies have the widest use                   ent headings are automatically changed. This strat-
across classes of computer applications, there are                   egy is useful across such applications as spreadsheets
three sets of strategies that are limited in scope to                (where different results can be generated by altering
only one class of computer applications:                             a variable such as an interest rate), and CAD systems
    First, there are strategies that exploit the itera-              (where it can be used to generate variations on a re-
tive power of computers, such as the detail-aggre-                   peated window design in a building façade).
gate-manipulate strategy discussed earlier. These are                    Third, there are strategies that exploit the power
useful mainly for authoring applications such as                     of organization provided by authoring applications.
drawing systems and word processors.                                 The power of organization enables users to explic-
                                                                            APPLICATION USE STRATEGIES ❚❙❘ 35



itly structure information in representations (such         ate and manipulate precise graphic objects. For ex-
as in a table). These explicit representations enable       ample, a common precision drawing task is to cre-
users to make rapid changes to the content with-            ate a line that is precisely tangent and touching the
out having to manually update the structure of the          end of an arc (as shown in the arched windows in
representation. For example, one way to represent           Figure 1). One way to perform this task is to visually
tabular information in a word-processing applica-           locate, and then click the end of the arc when draw-
tion is by using tabs between the words or numbers.         ing the line. This is error-prone because the user re-
However, because tabs do not convey to the com-             lies on visual feedback to detect the precise location
puter an explicit tabular representation consisting         of the end of the arc. Another way is to use the snap-
of rows and columns, the tabular structure may              to-object command, which enables the user to click
not be maintained when changes are made to the              a point that is only approximately at the end of the
content. A more efficient and effective way to per-          arc. The computer responds by automatically locat-
form this task is to first make the table explicit to the    ing the precise end of the arc, and therefore en-
computer by using the command insert table, and             ables the user to draw a line that is precisely tangent
then to add content to the table. Because the com-          to the end of the arc.
puter has an internal data structure for representing            Similar small-scope strategies have been iden-
a table, the tabular representation will be maintained      tified for word-processing applications (such as those
during modifications (such as adding more content            that assist in transforming text to generate summaries
to a cell in the table). Organization strategies are also   or translations) and for spreadsheets (such as
useful in other authoring applications. For example,        those that decompose formulas into subformulas to
information can be stored using a set-subset repre-         enable quick debugging).
sentation in a spreadsheet (as when different
sheets are used to organize sets of numbers) and in         Future Extensions of the
a CAD system (as when different layers are used to          Strategy Framework
organize different types of graphic information).           The strategy framework described above focuses on
     As discussed above, strategies that exploit the        authoring applications. However, the framework can
powers of iteration, propagation, and organization          also be extended to organize the large number of
are useful mainly for authoring applications.               search strategies that have been identified for use
However, it is important to note that the powers of         with information retrieval applications such as gen-
iteration, propagation, and organization can also be        eral-purpose search engines like Google. In contrast
offered by other classes of computer applications,          to computer powers that are useful in organizing
such as information retrieval applications. For ex-         strategies for use with authoring applications, strate-
ample, many Web browsers offer users ways to or-            gies for use with information retrieval systems ap-
ganize the addresses of different retrieved webpages.       pear to be driven by attributes of how information
(The organizing features provided by the favorites          sources are structured. For example, a large portion
command in Internet Explorer is one example.)               of the Web comprises densely connected webpages
However, while powers provided by authoring ap-             referred to as the core of the Web. The densely
plications can be provided in other classes of com-         connected structure of information sources in the
puter applications, the strategies that they exploit        core suggests the importance of using a variety of
will tend to be the same.                                   browsing strategies (that rely on using hyperlinks to
                                                            move from one page to another) to locate relevant
Small-Scope Strategies                                      sources. There is also a large portion of the Web that
Small-scope strategies exploit powers provided by           consists of new pages that are not linked to many
particular subclasses of applications. For example,         other pages. Strategies to find these pages therefore
the power of graphic precision is offered mainly by         require the use of different query-based search en-
drawing systems, such as CAD systems. Strategies            gines, given that no single search engine indexes all
that exploit graphic precision enable users to cre-         webpages.
36 ❘❙❚ BERKSHIRE ENCYCLOPEDIA OF HUMAN-COMPUTER INTERACTION



     While there has been much research on                 edge across applications than did students who were
strategies for finding relevant sources of informa-         taught only commands.
tion, one set of strategies works by selecting and or-
dering relevant sources of information based on            New Search Systems
the way information is distributed across sources.         The identification of search strategies to deal with
For example, health care information is typically          the scatter of information across the Web has led
scattered across different health care portals. In this    to the design of a new kind of domain portal
situation a useful strategy is to visit specific kinds     called a Strategy Hub. This type of domain portal
of portals in a particular order to enable compre-         implements the heuristic search strategy of visiting
hensive accumulation of the relevant information.          sources of information in a particular order. Recent
Such strategies become critical when incomplete            studies show that such a system enables users to find
information can have dangerous consequences (as            more comprehensive information on specific topics
is the case with incomplete information on health          when compared to the information retrieved by users
issues).                                                   of other search systems.
     An important difference between strategies for
using authoring applications and strategies for us-        An Analysis Method To Ensure Consistency in
ing information retrieval systems is that search strate-   Capabilities across Applications
gies are fundamentally heuristic—that is, they are         To enable the widest use of strategies across com-
rules of thumb that do not guarantee successful task       puter applications, designers must provide a con-
completion. This is in part because users’ evaluation      sistent set of commands. Therefore, a method called
of what is relevant changes based on what is being         “designs conducive to the use of efficient strategies”
learned during the search process.                         (Design-CUES) has been developed that enables de-
                                                           signers to systematically check if their designs pro-
                                                           vide the commands necessary for users to implement
                                                           efficient and effective strategies.
How the Identification of
Strategies Can Improve
                                                           Looking Forward
Human-Computer Interaction                                 Many years of research has shown that merely learn-
The identification and analysis of application use         ing commands does not make for the best use of
strategies suggests three practical developments: strat-   complex computer applications. The effective and
egy-based instruction, new search systems, and an          efficient use of computer applications often requires
analysis method to ensure consistency in capabili-         the use of strategies in addition to commands. An
ties across applications.                                  important research goal has therefore been to
                                                           identify strategies for using a wide range of computer
Strategy-Based Instruction                                 applications. The strategies that have been identified
Strategies for using authoring applications have           to date have benefited users through strategy-
led to the design of strategy-based instruction.           based instruction, new forms of search systems, and
Strategy-based instruction teaches commands in             new design methods. As research on strategy iden-
combination with the authoring strategies that make        tification continues, we can expect more develop-
use of authoring applications’ powers of iteration,        ments along those lines, all with the ultimate goal of
propagation, and organization. Research has shown          making users more effective and efficient in the
that students who took the strategy-based training         use of complex computer applications.
acquired more efficient and effective strategies and
demonstrated a greater ability to transfer that knowl-                                      Suresh K. Bhavnani
                                                                                                                    ARPANET ❚❙❘ 37



                                                                               Advanced Research Projects Agency (DARPA),
FURTHER READING                                                                funded some of the most important research of
                                                                               the twentieth century.
Bates, M. (1979). Information search tactics. Journal of the American
   Society for Information Science 30(4), 205–214.
Bates, M. J. (1998). Indexing and access for digital libraries and the
   Internet: Human, database, and domain factors. Journal of the               The Arpanet Concept
   American Society for Information Science, 49(13), 1185–1205.
Belkin, N., Cool, C., Stein, A., & Thiel, U. (1995). Cases, scripts, and in-
                                                                               The Arpanet long-distance computer network was
   formation-seeking strategies: On the design of interactive informa-         a collection of ideas, breakthroughs, and people.
   tion retrieval systems. Expert Systems with Applications, 9(3), 379–395.    The roots of the Arpanet can be traced to one of
Bhavnani, S. K. (2002). Domain-specific search strategies for the ef-           ARPA’s most famous managers, J. C. R. Licklider.
   fective retrieval of healthcare and shopping information. In
   Proceedings of CHI’02 (pp. 610–611). New York: ACM Press.
                                                                               In 1962 Licklider was recruited to work at ARPA,
Bhavnani, S. K. (in press). The distribution of online healthcare infor-       then housed in the Pentagon, to start a behavioral
   mation: A case study on melanoma. Proceedings of AMIA ’03.                  sciences program. Although a psychologist by train-
Bhavnani, S. K., Bichakjian, C. K., Johnson, T. M., Little, R. J., Peck,       ing, Licklider had a passion for the emergent field of
   F. A., Schwartz, J. L., et al. (2003). Strategy hubs: Next-generation
   domain portals with search procedures. In Proceedings of CHI                computers and was adamant that the future of com-
   ’03, (pp. 393–400). New York: ACM Press.                                    puting resided in the interactions between humans
Bhavnani, S. K., & John, B. E. (2000). The strategic use of complex            and computers. In his seminal work, a paper enti-
   computer systems. Human-Computer Interaction, 15(2–3), 107–137.
Bhavnani, S. K., Reif, F., & John, B. E. (2001). Beyond command knowl-
                                                                               tled “Man-Computer Symbiosis” written in 1960,
   edge: Identifying and teaching strategic knowledge for using com-           Licklider predicted that computers would not be
   plex computer applications. In Proceedings of CHI ’01 (pp. 229–236).        merely tools for people to use but also extensions of
   New York: ACM Press.                                                        people, forming a symbiotic relationship that would
Drabenstott, K. (2000). Web search strategies. In W. J. Wheeler
   (Ed.), Saving the user’s time through subject access innovation: Papers
                                                                               revolutionize the way people interact with the world.
   in honor of Pauline Atherton Cochrane (pp. 114–161). Champaign:                  Through ARPA Licklider began to interact with
   University of Illinois Press.                                               the brightest minds in computing—scientists at
Mayer, R. E. (1988). From novice to expert. In M. Helander (Ed.),              Stanford, Berkeley, UCLA, MIT, and a handful of com-
   Handbook of human-computer interaction (pp. 781–796).
   Amsterdam: Elsevier Science.                                                panies that made up what Licklider considered to be
O’Day, V., & Jeffries, R. (1993). Orienteering in an information               his “intergalactic computer network.” Of course, this
   landscape: How information seekers get from here to there. In               network existed only in theory because people had
   Proceedings of CHI 93 (pp. 438–445). New York: ACM Press.
Shute, S., & Smith, P. (1993). Knowledge-based search tactics. Infor-
                                                                               no way to bring these resources together other than
   mation Processing & Management, 29(1), 29–45.                               telephone or face-to-face meetings. However, Licklider
Siegler, R. S., & Jenkins, E. (1989). How children discover new strate-        had the vision of gathering these people and resources,
   gies. Hillsdale, NJ: Lawrence Erlbaum Associates.                           making the intergalactic network a physical network
Singley, M., & Anderson, J. (1989). The transfer of cognitive skill.
   Cambridge, MA: Harvard University Press.
                                                                               through an integrated network of computers.
                                                                               Although originally brought on board to work on be-
                                                                               havioral science issues in command-and-control
                                                                               systems, Licklider was directly responsible for trans-
                       ARPANET                                                 forming his command-and-control research office
                                                                               into the Information Processing Techniques Office
The Arpanet, the forerunner of the Internet, was devel-                        (IPTO), which would be responsible for critical ad-
oped by the U.S. Department of Defense’s Advanced                              vanced computing achievements for decades to come.
Research Projects Agency (ARPA) in the early 1960s.                            Although Licklider left ARPA in 1964, he had a last-
ARPA was created in 1958 by President Dwight D.                                ing effect on the field of computing and the devel-
Eisenhower to serve as a quick-response research and                           opment of the Arpanet.
development agency for the Department of Defense,                                   In 1966 another computer visionary, Bob Taylor,
specifically in response to the launch of the Soviet                            became director of IPTO and immediately began
satellite Sputnik. The agency, now the Defense                                 to address the computer networking problem. The
38 ❘❙❚ BERKSHIRE ENCYCLOPEDIA OF HUMAN-COMPUTER INTERACTION



                                                            ple locations), packet switching, dynamic routing al-
  We were just rank amateurs, and we were expecting that    gorithms (computational means of directing data
  some authority would finally come along and say,“Here’s    flows), and network survivability/redundancy. Packet
  how we are going to do it.” And nobody ever came along.   switching would be a critical element of network de-
                     —Vint Cerf on the design of Arpanet    sign because it would allow information to be bro-
                                                            ken down into pieces or “packets” that would be sent
                                                            over the network and reassembled at their final des-
                                                            tination. This design was a much more efficient mes-
computing field at that time suffered from duplica-          saging design, particularly when contrasted to
tion of research efforts, no electronic links between       analogue phone lines. Additionally, the distributed
computers, little opportunity for advanced graph-           network design would be more efficient and robust.
ics development, and a lack of sharing of valuable          Without central nodes (locations that contain all the
computing resources. Taylor asked the director of           resources and then distribute them to the rest of the
ARPA, Charles Herzfeld, to fund a program to create         system), the system could survive a loss of one or
a test network of computers to solve these problems.        more nodes and still route data traffic. This design
Herzfeld granted Taylor’s request, and Taylor’s office       also would allow more efficient data trafficking when
received more than one million dollars to address the       coupled with an adaptive networking algorithm ca-
problems. Thus, the Arpanet project was born.               pable of determining the most efficient path for any
     Taylor needed a program manager for the                packet to travel.
Arpanet project. He recruited Larry Roberts from                 Researchers addressed these issues prior to the
MIT’s Lincoln Labs. Roberts, twenty-nine years old,         Arpanet project. RAND’s Paul Baran recommended a
arrived at the Pentagon in 1966 and was ready to ad-        distributed switching network to the U.S. Air Force in
dress head on the problem of communications be-             1965 for the communications network of the Strategic
tween computers.                                            Air Command, but the network was not developed. In
                                                            the United Kingdom Don Davies was working on
                                                            packet switching and adaptive networking for the
Fundamental Issues in Networking                            Ministry of Defense. The two men independently came
Several fundamental issues existed in the network-          up with many of the same answers that would even-
ing of computers. Networking had been conceived             tually be incorporated into the Arpanet.
of to solve the problem of resource sharing between
computers. During the 1960s computers were ex-
tremely large, expensive, and time consuming to op-         The Arpanet Experiment
erate. ARPA had already invested in computing               Larry Roberts arrived at ARPA in 1966 with the
resources at several computing centers across the           charge to solve the computer networking problem.
country, but these centers had no way to communi-           At an ARPA investigators meeting in Ann Arbor,
cate among one another or to share resources. At the        Michigan, Roberts proposed a networking experi-
same time, Cold War concerns were causing U.S. sci-         ment that would become the Arpanet. He proposed
entists to take a hard look at military communica-          that all of the ARPA time-sharing computers at var-
tions networks across the country and to evaluate           ious sites across the country be connected over dial-
the networks’ survivability in case of a nuclear strike.    up telephone lines. The time-sharing (or host)
In the United Kingdom scientists were looking at            computers would serve double duty—both as re-
networks for purely communications use and were             sources and routers. Meeting participants met
evaluating digital communication methods to work            Roberts’s proposal with a great deal of skepticism.
around the inefficiency of the analogue telephone           Why would people want to spend valuable comput-
system. Both U.S. and United Kingdom scientists             ing resources to communicate between computers
were researching distributed networks (digital data         when people already had all the computing they
communication networks that extend across multi-            needed at their site? At the time, sharing between
                                                                                               ARPANET ❚❙❘ 39



computing centers was a goal of ARPA and not nec-       Lyon, assistant to the president of the University of
essarily of the scientific community itself. In addi-    Texas, unveil the Sputnik-era beginnings of the
tion, researchers would be reluctant to give up         Internet, the groundbreaking scientific work that cre-
valuable computing power just so they could “share”     ated it, and the often eccentric, brilliant scientists and
with other researchers. However, a researcher at        engineers responsible. The team, led by Frank
the meeting, Wes Clark, struck upon a solution          Heart, was dedicated to building the Arpanet on time
that would allow the experiment to be carried out.      and to specifications and had only nine months to de-
Clark recommended keeping the host computers out        liver the first IMP. Despite hardware setbacks, the team
of the networking duties. Instead, he suggested us-     delivered the first IMP to UCLA early. UCLA was also
ing a subnetwork of intermediary computers to han-      the site of the network management center, the “test
dle packet switching and data trafficking. This         track” for the Arpanet. The team was charged with
subnetwork would reduce the computing demand            testing the network’s limits and exposing bugs, flaws,
on the host computers, and the use of a subnetwork      and oddities. The initial Arpanet experiment consisted
of specialized computers would provide uniformity       of four nodes, with an IMP at UCLA, Stanford
and control. This suggestion solved many problems,      Research Institute (SRI), University of Utah, and
both technical and administrative, and would allow      University of California at Santa Barbara. BBN also
ARPA to control the subnetwork. The computers           was responsible for two critical elements: the IMPs
used at the subnetwork level were called “interface     themselves (including IMP-to-IMP communications)
message processors” (IMPs). In addition to design-      and the specifications for the IMP-to-host commu-
ing IMPs, researchers would have to develop pro-        nications. The specifications for the IMP-to-host com-
tocols for how the IMPs would communicate with          munications were drafted by Bob Kahn, who became
host computers and create the network.                  the intermediary between the Arpanet research com-
    ARPA issued a request for proposals (RFP) in        munity and BBN. Graduate students of the host in-
1968, because the specifications for the network had     stitutions digested those specifications and developed
become so detailed. These specifications included:       the code that would serve as the interface between host
                                                        and IMP. They formed the Network Working
 ■ Transfer of digital bits from source to specified
                                                        Group to hammer out the details of protocols, shared
   location should be reliable.
                                                        resources, and data transfer. They created file trans-
 ■ Transit time through the subnetwork should
                                                        fer protocols (which layout the rules for how all com-
   be one-half second or less.
                                                        puters handle the transfer of files) that became the
 ■ The subnetwork had to operate autonomously.
                                                        backbone of the Arpanet and made it functional. This
 ■ The subnetwork had to function even when IMP
                                                        experiment was so successful that the Arpanet was ex-
   nodes went down.
                                                        panded to include other research sites across the coun-
    The ARPA RFP was issued to determine which          try until it grew to twenty-nine nodes. In 1972 the
company could build the Arpanet to these specifi-        Arpanet made its public debut at the International
cations. After much debate, the contract was awarded    Conference on Computer Communication. It was an
in 1969 to the Bolt, Baranek, and Newman company        unequivocal hit, and the computer networking con-
(BBN), which had assembled an amazing team of           cept was validated in the public arena.
scientists to transform this vision into reality. The
choice of BBN was a surprise to many people be-
cause BBN was considered to be a consulting firm,        The Arpanet Evolves
not a computing heavy hitter. However, its proposal     As members of a user community, the researchers
was so detailed and exacting that it could begin work   involved in the Arpanet were always adding, creat-
immediately upon awarding of the contract. BBN          ing, experimenting. The Arpanet became a bargain-
had only twelve months to do the work.                  ing tool in the recruiting of computer science faculty
    In their 1996 book, Where Wizards Stay Up Late,     and an impromptu communication tool for “net-
Katie Hafner, co-author of Cyberpunk, and Matthew       work mail” or electronic mail (e-mail). In 1973 an
40 ❘❙❚ BERKSHIRE ENCYCLOPEDIA OF HUMAN-COMPUTER INTERACTION



ARPA study showed that 75 percent of all traffic
on the Arpanet was e-mail. Researchers eventually                FURTHER READING
wrote dedicated software to handle this “side use” of
the Arpanet.                                                     Adam, J. (1996, November). Geek gods: How cybergeniuses Bob Kahn
                                                                    and Vint Cerf turned a Pentagon project into the Internet and con-
     In 1972 Bob Kahn left BBN and went to work at                  nected the world. Washingtonian Magazine, 66.
ARPA with Larry Roberts. Kahn was now in charge                  Baranek, B., & Newman. (1981, April). A history of the ARPANET: The
of the network that he had helped create. He formed                 first decade. NTIS No. AD A 115440). Retrieved March 23, 2004,
a fruitful collaboration with Vint Cerf of Stanford                 from http://www.ntis.gov
                                                                 Evenson, L. (1997, March 16). Present at the creation of the
( w h o w a s a g r a d u a te s t u d e n t o n t h e U C L A      Internet: Now that we’re all linked up and sitting quietly, Vint Cerf,
Arpanet project) that led to the next evolution of                  one of its architects, describes how the Internet came into being.
networking. Together they tackled the problem of                    San Francisco Chronicle (p. 3ff).
                                                                 Hafner, K., & Lyon, M. (1996). Where wizards stay up late: The origins
packet switching in internetworking, which would                    of the Internet. New York: Simon & Schuster.
eventually become the Internet. In 1975 Vint Cerf                Hughes, T. J. (1998). Rescuing Prometheus. New York: Pantheon Books.
went to DARPA to take charge of all of the ARPA                  Norberg, A., & O’Neill, J. (1997). Transforming computer technology.
Internet programs, and the Arpanet itself was trans-                Ann Arbor: Scholarly Publishing Office, University of Michigan
                                                                    Library.
ferred to the Defense Communication Agency, a                    Salus, P. (1995). Casting the Net. Reading, MA: Addison-Wesley.
transfer that upset some people in the non-bu-
reaucratic computing research community. The
Internet was created by the merging of the Arpanet,
SATNET (Atlantic Packet Satellite Network), and a
packet radio network—all based on the transmis-                                   ARTIFICIAL
sion-control protocol/Internet protocol (TCP/IP)
standard that Cerf and Kahn created—and then more                                INTELLIGENCE
and more networks were created and connected un-
til the Internet was born. The Arpanet eventually                Most research in mainstream artificial intelligence
burgeoned to 113 nodes before it adopted the new                 (AI) is directed toward understanding how people
TCP/IP standard and was split into MILNET and                    (or even animals or societies) can solve problems ef-
Arpanet in 1983. In 1989 the Arpanet was officially               fectively. These problems are much more general
“powered down,” and all of the original nodes                    than mathematical or logical puzzles; AI researchers
were transferred to the Internet.                                are interested in how artificial systems can perceive
                                                                 and reason about the world, plan and act to meet
                                                                 goals, communicate, learn, and apply knowledge such
The Internet and Beyond                                          that they can behave intelligently.
The creation of the Arpanet—and then the Internet—                   In the context of human-computer interaction
was the work of many researchers. Only with diffi-                (HCI), research in AI has focused on three general
culty can we imagine our modern society without the              questions:
interconnectedness that we now share. The
                                                                   ■ How can the process of designing and imple-
Arpanet was a testament to the ingenuity of the hu-
                                                                     menting interactive systems be improved?
man mind and people’s perhaps evolutionary de-
                                                                   ■ How can an interactive system decide which
sire to be connected to one another. The Arpanet not
                                                                     problems need to be solved and how they should
only brought us closer together but also brought us
                                                                     be solved?
one step closer to J. C. R. Licklider’s vision of human-
                                                                   ■ How can an interactive system communicate
computer interaction more than four decades ago.
                                                                     most effectively with the user about the prob-
                                                                     lems that need to be solved?
    Amy Kruse, Dylan Schmorrow, and J. Allen Sears
                                                                    The first question deals with the development
See also Internet—Worldwide Diffusion                            process in HCI, the others with user interaction,
                                                                               ARTIFICIAL INTELLIGENCE ❚❙❘ 41



specifically the issues of control and communica-         acting as automated tax advisors, automobile repair
tion. These questions have been a central concern in      advisors, or medical consultants, search oppor-
HCI for the past thirty years and remain critical         tunistically for combinations of if-then rules that de-
today. AI has been able to provide useful insights into   rive plausible conclusions from input data and
how these questions can be answered. In sum,              existing knowledge. Machine learning systems, in-
what AI brings to HCI development is the possi-           cluding neutral networks, incrementally refine an
bility of a more systematic exploration and evalua-       internal representation of their environment, in a
tion of interface designs, based on automated             search for improved performance on given tasks.
reasoning about a given application domain, the           Natural language understanding systems search
characteristics of human problem solving, and gen-        for correct interpretations through a space of am-
eral interaction principles. The AI approach can ben-     biguous word meanings, grammatical constructs,
efit end users because it encourages tailoring the        and pragmatic goals. These brief descriptions are
behavior of an interactive system more closely to         only approximate, but they help us understand how
users’ needs.                                             a system can represent and deal with some of the
                                                          problems that arise in interacting with users or in-
                                                          terface developers in an intelligent way.
The Concept of Search
Almost all techniques for problem solving in AI
are based on the fundamental concept of search. One         ARTIFICIAL INTELLIGENCE (AI) The subfield of computer
way to understand search is by analogy to naviga-           science that is concerned with symbolic reasoning and
tion on the World Wide Web. Imagine that my goal            problem solving.
is to reach a specific webpage starting from my home-
page, and that I have no access to automated facili-
ties such as search engines. I proceed by clicking
on the navigation links on my current page. For each      AI and the Development
new page that comes up, I decide whether I have
reached my goal. If not, then I evaluate the new page,
                                                          of User Interfaces
comparing it with the other pages that I have en-         Considerable attention in AI has focused on the
countered, to see whether I am moving closer to my        process of developing user interfaces. Experienced
goal or farther away. Based on my evaluation, I           developers generally have a working knowledge of
may continue forward or go back to an earlier, more       software engineering practice, interface architectures,
promising point to take a different path.                 graphic design, and related areas, plus information
     An automated search process works in the same        about the purpose for which the interface is to be
way. Pages correspond to states in a search space,        used. If this knowledge can be captured in compu-
or relevant information about the environment; nav-       tational form, an intelligent development environ-
igation actions are operators, which transform one        ment can aid developers by testing and validating
state into another; an evaluation function assesses       design specifications, by producing alternative de-
information about the state to guide the selection of     signs for a given specification, by generating po-
operators for further transformations.                    tential improvements to a design, and by automating
     A large number of AI techniques have been de-        some of the more common implementation tasks.
veloped to address specific classes of search prob-           The motivation for a search-based approach can
lems, representing the problems in different ways.        be seen most clearly in the problem of layout design.
For example, planning systems search for sequences        If an experienced designer were asked to organize
of interdependent operators to reach a set of goals;      ten loosely related items of information (represented
these systems can deal with complex tasks ranging         in text, pictures, and buttons) on a company’s top-
from planning space missions to helping robots nav-       level webpage, the final product might be the re-
igate over unfamiliar terrain. Expert systems, whether    sult of comparing several alternatives, perhaps a few
42 ❘❙❚ BERKSHIRE ENCYCLOPEDIA OF HUMAN-COMPUTER INTERACTION



dozen at most. The number of all possible layouts of          the product in real-world situations). Thanks to AI
ten items, however, runs into the millions and higher;        research, however, it is becoming possible to build ar-
this is much more than a designer can humanly con-            tificial software agents that can stand in for real users.
sider. Most of these layouts will be unacceptable (for             It is common to think about user interaction with
example, all possible orderings of items diagonally           the software in problem-solving terms, as goal-oriented
across the page), but there may be many effective de-         behavior. For example, if my goal is to send an e-mail
signs that are missed simply because the number               message, I divide this into subgoals: entering the re-
of possibilities is so enormous. A system that can            cipient information and subject line information, writ-
search through different spatial relationships and            ing a short paragraph of text, and attaching a picture.
evaluate the results, even without perfect accuracy,          My paragraph subgoal breaks down further into writ-
can give designers a more comprehensive view of the           ing individual sentences, with the decomposition
problem and its solutions.                                    continuing to the point of mouse movements and key
     Automated layout design is just one aspect of in-        presses. In AI terms, these decompositions can be rep-
terface design. Research in the general area of model-        resented by plans to be constructed and executed au-
based interface design aims to support developers in          tomatically. The PATHS system, a system designed to
all stages of the design process. In MOBI-D and               help automate the testing of graphical user interfaces,
Mastermind, which are user interface generation tools,        lets developers specify a beginning state, an end state,
developers build and evaluate abstract models of com-         and a set of goals to be accomplished using the inter-
puter applications (such as word processing appli-            face. PATHS then creates a comprehensive set of plans
cations, spreadsheet applications, or photographic            to achieve the goals. For example, given the goal of mod-
design applications), interaction tasks and actions,          ifying a document, the planner will generate sequences
presentations, even users and workplaces. The goal            of actions for opening the document, adding and delet-
is to give developers decision-making tools that al-          ing text, and saving the results, accounting for all the dif-
low them to apply their design skills but do not overly       ferent ways that each action can be carried out. If a given
restrict their choices. These tools test constraints, eval-   sequence is found not to be supported when it should
uate design implications, present suggestions, track          be, PATHS will record this as an error in the application.
changes, and so forth, facilitating the eventual con-              Similar work is carried out in the related field of
struction of the actual interface. For example, if a de-      cognitive modeling, which shares many concepts
veloper specifies that the user must enter a number            with AI. Cognitive modelers build computational
at some point, MOBI-D can present different inter-            models of human cognitive processing—perception,
face alternatives, such as a slider (the software equiv-      attention, memory, motor action, and so forth—in
alent of a linear volume control) or a text box that          order to gain insight into human behavior. To make
the user can type into directly, for the developer to         valid comparisons between a model’s performance
choose from. In Mastermind, the developer can switch          and human performance, a common experimental
between a number of visual formats, avoiding ones             ground is needed. User interfaces provide that com-
that are cumbersome. Current research in this area            mon ground. Cognitive models comparable to plan-
is helping to improve webpage design and build in-            ning systems have been developed for evaluating user
terfaces that meet the constraints of the next gener-         interfaces, and they have the added benefit of giving
ation of interactive devices, including cell phones and       developers information about the human side of in-
handheld computers.                                           teraction as well as the application side.
     AI research is also helping software companies
with product evaluation. Partially automated testing
of noninteractive software is now commonplace, but            Interaction
conventional techniques are not well suited to test-          The metaphor of tool use has come to dominate the
ing user interfaces. Software companies usually rely          way we understand human interaction with com-
on limited user studies in the laboratory, plus a large       puters, especially with regard to graphical user inter-
population of alpha and beta testers (people who test         faces. Just as a carpenter keeps specialized sets of tools
                                                                                           ARTIFICIAL INTELLEGENCE ❚❙❘ 43




                  A Personal Story—Putting Humans First in Systems Design

   The field of augmented cognition is pushing the integration of human systems and information technology to the fore-
   front, while also attempting to maximize human potential. My current (and anticipated future) experience with using an
   ever-increasing number of technologies during my everyday life compels me (propels me!) to help design a new class of
   systems for the user to interact with. Practitioners of traditional human-systems integration research and design have stead-
   fastly urged that the human must be considered when designing systems for human use.
       An emerging concept is that not only are human beings the weak link in current human-systems relationships, but also
   that the number of systems that a single human interacts with is growing so rapidly that the human is no longer capable
   of using these technologies in truly meaningful ways. This specifically motivates me to develop augmented cognition tech-
   nologies at the Defense Advanced Research Projects Agency (where I am a program manager). I want to decrease the num-
   ber of system interfaces that we need to interact with, and increase the number of advanced systems that individuals are
   capable of using simultaneously. On any given day, I typically wear (carry) five computers: my wristwatch, cell phone, two-
   way pager with e-mailing capability, a personal digital assistant, and a laptop. I find these systems intrusive and the asso-
   ciated demands on my time to be unacceptable. My home is inundated with appliances that are evolving into computer
   devices—these systems have advanced “features” that require significant attention in order to use them optimally. Even
   with the world’s greatest human factors interface, I would never have time to interact with all of these systems that I use
   on a daily basis.
       Having said all of this, I need the systems that support me to exhibit some intelligence; I need them to be able to per-
   ceive and understand what is going on around and inside of me. I do not have time to overtly direct them. Ideally they
   will support me by “sensing” my limitations (and my capabilities) and determining how best to communicate with me if
   absolutely necessary. Augmented cognition technology will imbue into these systems the ability to interact with me. Indeed,
   augmented cognition is about maximizing human potential. If we humans are the “weak link,” it is because our current
   advanced computer systems are actually limiting our performance. In the future, we must have transparent technologies
   addressing our needs, or we will be overwhelmed by meaningless interactions.
                                                                                                              Dylan Schmorrow



for framing a house or building fine furniture, an expe-                 The principles of direct manipulation provide a
rienced computer user has a variety of software tools              foundation for tool-based environments. Direct-ma-
for word processing, analyzing data with spread-                   nipulation interfaces, as defined by Ben Shneiderman,
sheets, or creating graphics and illustrations. User inter-        the founding director of the Human-Computer
faces are often thought of as tool-using environments,             Interaction Laboratory at the University of Maryland,
which has important implications for the involve-                  provide a visual representation of objects, allow rapid
ment of AI in user interaction. Let us extend the car-             operations with visible feedback, and rely mainly on
penter analogy. If I am intent on hammering a                      physical actions (such as selecting and dragging or
nail, I am not constantly reconsidering and recali-                pressing buttons) to initiate actions. Modern graph-
brating the relationship between the hammer and                    ical user interfaces can trace much of their power to
my hand, or the head of the hammer and the nail.                   direct-manipulation principles. Nevertheless, as pow-
Instead, after an initial adjustment, the hammer                   erful as direct-manipulation interfaces can be, they
effectively becomes an extension of my arm, so                     are not appropriate in all situations. For example,
that I can use it without thinking about it. Similarly,            sometimes in using a piece of software I know what
for a tool-based software environment, selecting in-               needs to be done—I can even describe in words what
dividual tools should be intuitive, and applying a tool            I would like to do—but I do not know exactly how to
should quickly become second nature.                               accomplish my task given the tools at hand.
44 ❘❙❚ BERKSHIRE ENCYCLOPEDIA OF HUMAN-COMPUTER INTERACTION



     These potential limitations, among others, have           Other approaches to building intelligent envi-
led AI researchers to consider alternatives to a strict   ronments, such as programming by example (PBE),
tool-based approach. First, it is possible to build in-   involve more significant changes to user interaction.
telligent environments that take a more active role in    PBE systems watch the user perform a procedure a
assisting the user—for example, by automatically          number of times and then automatically general-
adapting their behavior to the user’s goals. Second,      ize from these examples to create a fully functional
intelligent behavior can be encapsulated within a soft-   program that can execute the repetitive actions so
ware agent that can take responsibility for different     the user does not have to.
tasks in the environment, reducing the burden on the           The SMARTedit system is an example of a ma-
user. Third, these agents and environments can com-       chine-learning approach to PBE, in the context of
municate with the user, rather than passively being       a text-editing application. Suppose that the user
acted upon by the user, as tools are.                     moves the cursor to the beginning of the word ap-
                                                          ple, erases the lowercase a, and types an uppercase
Intelligent Environments                                  A. There are several ways that those actions could be
Some intelligent environments work by integrating         interpreted. Perhaps, for example, the user wanted
AI search into an otherwise conventional interface.       to move the cursor forward n characters and replace
One recently developed technique, human-guided            the arbitrary character at that location with A, or
simple search, is intended to solve computationally       perhaps the user wanted to move to the next oc-
intensive problems such as the traveling salesman         currence of the letter a and capitalize it, or to correct
problem. This problem involves a salesman who must        the capitalization of the first word in a sentence, or
visit a number of cities while keeping the distance       some other possibility. Each of these interpretations
traveled as small as possible. Finding the optimal        is a different hypothesis maintained by SMARTedit
route for even a small number of locations is beyond      about the user’s intentions. As the user takes further
what can be done with pencil and paper; for ten           actions, repeating similar sequences on different text,
locations there are over three million possible routes.   ambiguity is reduced. Some hypotheses become more
Large problems are challenging even for the most          plausible while others are pruned away because they
sophisticated computer programs.                          predict actions inconsistent with the user’s behav-
     The user works with the human-guided search          ior. At any point, the user can direct SMARTedit to
(HUGSS) tool kit through a graphical display of           take over the editing process and watch the system
routes that the system has found. By pressing a           apply its most hig hly ranked hy pothesis. If
button, the user activates a search process that com-     SMARTedit carries out a sequence incorrectly, the
putes the best route it can find within a fixed period      user can interrupt and correct the mistake, with
of time. The user examines the solution and modi-         the system learning from the feedback.
fies it by selecting parts of the route that need fur-          Adaptive user interfaces are another type of in-
ther refinement or identifying those parts that already    telligent environment. Their development is moti-
have a reasonable solution. The user brings human         vated by the observation that while the ideal software
perception and reasoning to bear on the problem by        system is tailored to an individual user, for economic
constraining the space that the search process con-       reasons a single system must be designed and released
siders (for example, by temporarily focusing the          to thousands or even millions of users, who differ
search on routes between five specific locations, rather    widely from one another in expertise, interests, needs,
than the entire set). Problem-solving responsibility      and so forth. The solution is a system that can
is explicitly shared between the user and the system,     adapt to its users when in use. A simple example is
with the amount and timing of the system’s effort         adaptive menus. A system can record how often the
always under the user’s control. HUGSS works faster       user selects different menu options, and modify the
than the best fully automated systems currently in        menu structure so that more frequently chosen op-
use, and it produces results of equal quality.            tions can be reached more efficiently. This basic idea
                                                                                ARTIFICIAL INTELLEGENCE ❚❙❘ 45



also works in more sophisticated adaptive systems,              Information retrieval is just one area in which
many of which compile detailed models of users and         agents have become popular. Agents have also ap-
their particular tasks and adapt accordingly. Adaptive     peared in help systems, planning and scheduling aids,
systems have become especially relevant in efforts to      scripting systems, intelligent tutoring systems, col-
personalize the World Wide Web as well as in research      laborative filtering applications, matchmaking ap-
on intelligent tutoring systems and other applications     plications, and electronic auctions. Work on agents
of AI to education.                                        is one of the fastest-growing areas of AI.
                                                                An important topic within research on agents is
                                                           how to make agents interact most effectively with users.
Intelligent Agents                                         Who should take the initiative—the user or the agent?
The engineer Michael Huhns and the computer sci-           And when? Should one ever interrupt the other? These
entist Munindar Singh define intelligent agents as         are questions of mixed-initiative interaction. Some
“active, persistent (software) components that per-        work on these questions is carried out in the area of
ceive, reason, act, and communicate” (Huhns and            rational decision making, wherein rationality is in-
Singh 1997, 1). For our purposes, the most impor-          terpreted in an economic sense. If an agent has knowl-
tant characteristic of an agent is its autonomy—its        edge of the user’s preferences and can reason about
ability to carry out activities without the constant,      the user’s goals, then it can, for example, determine
direct supervision of a human being. Agents in use         that the value of the information it can contribute
at present include animated characters or “believ-         at some point will offset the cost of the user having to
able” agents, autonomous agents such as softbots           deal with an interruption. A different direction is taken
(software agents that perform tasks on the Internet)       by projects that are influenced by the ways that peo-
and physical robots, and mobile agents whose pro-          ple interact with one another, especially in dialogue.
cessing is not limited to a single computer platform.      TRIPS (The Rochester Interactive Planning System)
Agents are also used in multi-agent systems, which         is a mixed-initiative planning and scheduling assis-
may involve mixed teams of humans and agents.              tant that collaborates with a human user to solve prob-
Most relevant to HCI are interface agents, which act       lems in crisis situations, such as planning and
as intelligent assistants within a user interface, some-   managing an evacuation. The COLLAGEN (from
times carrying out tasks on their own but also able        COLLaborative AGENt) system is a collaboration sys-
to take instructions and guidance from the user.           tem that can be incorporated into agents to give them
     Letizia is an interface agent that assists users in   sophisticated collaboration capabilities across a range
browsing the World Wide Web. Letizia operates in           of application domains. TRIPS and COLLAGEN
conjunction with a standard Web browser, main-             agents can interact with users via everyday natural
taining two open windows for its own use. As the           language as well as through multimedia presentations,
user navigates through the Web, Letizia records the        which leads to the topic of communication.
information on each page that the user visits and
performs an independent search of nearby pages that        Communication
the user may not have seen. Letizia’s evaluation func-     Some agents communicate by conventional means
tion compares the information on the pages that it         in a graphical user interface, for example by raising
visits with the information that the user has seen up      dialog windows and accepting typed input and but-
to the current point. In this way Letizia can make         ton presses for responses. A common and reason-
suggestions about what the user might be interested        able expectation, however, is that if a system is
in seeing next. As Letizia visits pages, it displays the   intelligent, we should be able to talk with it as we
most promising ones for a short time in one win-           would with other people, using natural language.
dow and the overall winner it has encountered in the       (Natural language refers to the languages that peo-
other window. The user can watch what Letizia is           ple commonly use, such as English or French, in con-
doing and take control at will.                            trast to programming languages.) Unfortunately,
46 ❘❙❚ BERKSHIRE ENCYCLOPEDIA OF HUMAN-COMPUTER INTERACTION



even a brief treatment of natural-language under-           hance its spoken conversation. Users can communi-
standing and generation, not to mention voice recog-        cate via speech or gesture, even by simply looking at
nition and speech output, is beyond the scope of this       particular objects, nonverbal behavior that is sensed
article. An example, however, may give some idea of         by cameras. Systems like REA aim to make the com-
the issues involved. Consider three remarks from the        puter side of face-to-face human-computer commu-
user’s side of a dialogue with a natural-language sys-      nication as rich and nuanced as the human side.
tem (the bracketed text is not spoken by the user):
    User (1): Show me document.txt.
    User (2): What’s the last modification date [on          Future Directions
the file document.txt]?                                      This article has introduced the reader to AI approaches
    User (3): Okay, print it [i.e., document.txt].          to HCI rather than give a taxonomy of AI systems;
    To respond correctly, the system must be able           many of the systems touched upon are much broader
to reason that modification dates are associated with        in scope than can be conveyed through a category
files and that files rather than dates are usually printed    assignment and a few sentences. Developments that
(“it” could grammatically refer to either.) Reading         do not fit neatly within the categories discussed are
this dialogue, English-speaking humans make these           listed below.
inferences automatically, without effort or even aware-
ness. It is only recently that computer systems have        Smart Rooms and Intelligent Classrooms
been able to match even a fraction of our abilities.        Much of what makes a software environment intelli-
    The QuickSet communication system combines              gent can be generalized to the physical domain. Smart
natural language and other methods of interaction           rooms and intelligent classrooms rely on the same
for use in military scenarios. Shown a map on a tablet      kind of technology as an embodied conversational
PC, the user can say, “Jeep 23, follow this evacuation      agent; they register users’ gestures and spoken com-
route,” while drawing a path on the display. The sys-       mands and adjust thermostats, change lighting, run
tem responds with the requested action. This in-            presentations and the like, accordingly.
teraction is striking for its efficiency: the user has two
simultaneous modes of input, voice and pen-aided            Games and Virtual Environments
gesture, and the ambiguities in one channel (in             Intelligent agents have begun to enrich games and
this example, the interpretation of the phrase “this        virtual environments, acting as teammates or oppo-
route”) are compensated for by information in the           nents. Extending this line of research, the Mimesis
other channel (the drawn path). In general, voice and       system imposes a nonscripted, dynamic narrative
natural language can support a more engaging, nat-          structure onto a virtual gaming environment, so that
ural style of interaction with the interface than ap-       external goals (for example, education on a histor-
proaches that use a single vector of communication.         ical period) can be met without compromising the
    Embodied conversational agents take work in nat-        user’s direct control over the environment.
ural language a step further. When people speak with
one another, communication is not limited to the            Human-Robot Interaction
words that are spoken. Gestures, expressions, and other     Robots are appearing outside the laboratory, in our
factors can modify or even contradict the literal mean-     workplaces and homes. Human-robot interaction
ing of spoken words. Embodied conversational agents         examines issues of interaction with physical agents
attempt to recognize and produce these broader cues         in real-world environments, even in social situations.
in communication. REA, a simulated real estate agent        Robots can be used to explore otherwise inaccessi-
research prototype developed at the Massachusetts           ble environments and in search-and-rescue missions.
Institute of Technology, is represented by a full body          It should be clear from this discussion that the
figure on a large-scale display. REA shows users around      most interesting problems in HCI are no longer
a house, making appropriate use of eye gaze, body           found in software technology, at the level of the
posture, hand gestures, and facial expressions to en-       visible components of the interface. Effective AI
                                                                                                                ASIAN SCRIPT INPUT ❚❙❘ 47



approaches to HCI focus on issues at deeper levels,                          Shneiderman, B. (1998). Designing the user interface: Strategies for
probing the structure of problems that need to be                                effective human-computer interaction. Boston: Addison-Wesley.
                                                                             Shneiderman, B., & Maes, P. (1997). Debate: Direct manipulation
solved, the capabilities and requirements of users,                              vs. interface agents. Interactions, 4(6), 42–61.
and new ways of integrating human reasoning with                             Sullivan, J. W., & Tyler, S. W. (Eds.). (1991). Intelligent user inter-
automated processing.                                                            faces. New York: ACM Press.
                                                                             St. Amant, R., & Healey, C. G. (2001). Usability guidelines for inter-
                                                 Robert St. Amant                active search in direct manipulation systems. In Proceedings of the
                                                                                 International Joint Conference on Artificial Intelligence (IJCAI) (pp.
                                                                                 1179–1184). San Francisco: Morgan Kaufman.
                                                                             Szekely, P., Sukaviriya, P., Castells, P., Muthukumarasamy, J., & Salcher,
FURTHER READING                                                                  E. (1996). Declarative interface models for user interface construc-
                                                                                 tion tools: The Mastermind approach. In L. Bass & C. Unger (Eds.),
Anderson, D., Anderson, E., Lesh, N., Marks, J., Mirtich, B., Ratajczak,
                                                                                 Engineering for human-computer interaction (pp. 120–150). London
   D., et al. (2000). Human-guided simple search. In Proceedings of
                                                                                 and New York: Chapman & Hall.
   the National Conference on Artificial Intelligence (AAAI) (pp.
                                                                             Wolfman, S. A., Lau, T. A., Domingos, P., & Weld, D. S. (2001).
   209–216). Cambridge, MA: MIT Press.
                                                                                 Mixed initiative interfaces for learning tasks: SMARTedit talks back.
Cassell, J. (Ed.). (2000). Embodied conversational agents. Cambridge,
                                                                                 In Proceedings of the International Conference on Intelligent User
   MA: MIT Press.
                                                                                 Interfaces (pp. 67–174). New York: ACM Press.
Cassell, J., Bickmore, T., Billinghurst, M., Campbell, L., Chang, K.,
   Vilhjálmsson, H., et al. (1999). Embodiment in conversational in-
   terfaces: REA. In Proceedings of the ACM Conference on Human
   Factors in Computing Systems (CHI) (pp. 520–527). New York:
   ACM Press.
Cypher, A., (Ed.). (1993). Watch what I do: Programming by demon-
   stration. Cambridge, MA: MIT Press.                                             ASIAN SCRIPT INPUT
Huhns, M. N., & Singh, M. P. (Eds.). (1997). Readings in agents. San
   Francisco: Morgan Kaufmann.                                               The Asian languages that employ the Chinese alphabet
Kobsa, A. (Ed.). (2001). Ten year anniversary issue. User Modeling and
   User-Adapted Interaction, 11(1–2).
                                                                             in their writing systems present difficult challenges
Lester, J. (Ed.). (1999). Special issue on intelligent user interfaces. AI   for entering text into computers and word proces-
   Magazine, 22(4).                                                          sors. Many Asian languages, such as Korean and Thai,
Lieberman, H. (1995). Letizia: An agent that assists Web browsing. In        have their own alphabets, and the Devanagari al-
   Proceedings of the International Joint Conference on Artificial
   Intelligence (IJCAI) (pp. 924–929). San Francisco: Morgan Kaufmann.
                                                                             phabet is used to write Sanskrit, Hindi, and some
Lieberman, H. (Ed.). (2001). Your wish is my command. San Francisco:         other languages of India. Designing keyboards and
   Morgan Kaufmann.                                                          fonts for alphabets of languages—such as Hebrew,
Lok, S., & Feiner, S. (2001). A survey of automated layout tech-             Greek, Russian, and Arabic—that do not employ the
   niques for information presentations. In Proceedings of the First
   International Symposium on Smart Graphics (pp. 61–68). New York:          Roman alphabet used by English and other west-
   ACM Press.                                                                ern European languages is relatively simple. The chal-
Maybury, M. T., & Wahlster, W. (Eds.). (1998). Readings in intelli-          lenge with Chinese, simply put, is that a standard
   gent user interfaces. San Francisco: Morgan Kaufmann.
Memon, A. M., Pollack, M. E., Soffa, M. L. (2001). Hierarchical GUI
                                                                             national database contains 6,763 symbols (called
   test case generation using automated planning. IEEE Transactions          “characters” rather than “letters”), and a keyboard
   on Software Engineering, 27(2), 144–155.                                  with so many keys would be completely unwieldy.
Newell, A., & Simon, H. (1972). Human problem solving. Englewood                 As was the case with ancient Egyptian hiero-
   Cliffs, NJ: Prentice-Hall.
Oviatt, S. L., Cohen, P. R., Wu, L., Vergo, J., Duncan, L., Suhm, B., et
                                                                             glyphics and Mesopotamian cuneiform, Chinese writ-
   al. (2002). Designing the user interface for multimodal speech and        ing began as pictographs that represented particular
   gesture applications: State-of-the-art systems and research di-           things. Evolving through time and modified for grace-
   rections. In J. Carroll (Ed.), Human-computer interaction in the          ful drawing with an ink brush, these pictographs be-
   new millennium (pp. 419–456). Reading, MA: Addison-Wesley.
Puerta, A.R. (1997). A model-based interface development environ-            came the current system of characters representing
   ment. IEEE Software, 14(4), 41–47.                                        concepts and sounds in a complex interplay of func-
Ritter, F. E., & Young, R. M. (Eds.). (2001). Special issue on cognitive     tions. A person fully literate in Chinese today uses
   modeling for human-computer interaction. International Journal
   of Human-Computer Studies, 55(1).
                                                                             3,000 to 4,000 characters; newspapers have 6,000 to
Russell, S., & Norvig, P. (1995). Artificial intelligence: A modern ap-       7,000 available, but some dictionaries list as many as
   proach. Englewood Cliffs, NJ: Prentice-Hall.                              50,000. In 1958 a standardized phonetic system based
48 ❘❙❚ BERKSHIRE ENCYCLOPEDIA OF HUMAN-COMPUTER INTERACTION



on the Roman alphabet and called “pinyin” was in-            play the equivalent hiragana characters instead of
troduced, but it has not replaced the traditional            Western letters on the screen.
system of writing. Japanese employs two phonetic al-
phabets called kana, as well as Chinese characters
called kanji. In 1983 the Japan Industrial Standard          Many Meanings
listed 2,963 commonly used characters plus an-               The writer probably does not want the hiragana but
other 3,384 that appear only rarely. Korean also makes       rather the kanji, but many Japanese words can be ro-
some use of Chinese characters, but the chief form           manized kannsou. Asian languages have many
of writing is with an alphabet historically based on         homonyms (words that sound similar but have dif-
Chinese but phonetically representing the sounds of          ferent meanings), and Chinese characters must rep-
spoken Korean.                                               resent the one intended meaning. The standard way
    Because Japan has been a leader in developing com-       in which word processors handle this awkward fact, in
puter technology for decades, its language is the best       Chinese as well as Japanese, is to open a selection win-
example. Around 1915 Japan began experimenting               dow containing the alternatives. For example, let’s say
with typewriters, but they were cumbersome and rare.         the user typed “kannsou,” then hit the spacebar (which
Typewriters could be made simply for the kana, a cen-        is not otherwise used in ordinary Japanese) to open
turies-old phonetic system for writing Japanese syl-         the selection window with the first choice highlighted.
lables, either in the traditional hiragana form or in the    The user can select the second choice, which is the cor-
equivalent katakana form used for writing foreign            rect Chinese characters for the Japanese word mean-
words or telegrams. Occasionally reformers have sug-         ing “a comment” (one’s thoughts and impressions
gested that Chinese characters should be abandoned           about something). If the user wanted kannsou to mean
in favor of the kana or the Roman alphabet, but this         not “comment,” but rather “dry,” he or she would se-
reform has not happened. Thus, newspapers employed           lect the third choice. The fourth through ninth choices
vast collections of Chinese type, and careful hand-          mean “welcome” and “farewell,”“a musical interlude,”
writing was used in business, schools, and forms of          “completion, as of a race,”“meditate,”“hay”(dry grass),
printing such as photocopying that could duplicate           and “telling people’s fortunes by examining their faces.”
handwriting. During the 1980s word processors                     Good Asian-language word processing software
were introduced that were capable of producing the           presents the choices in descending order of likelihood,
traditional mixture of kanji, hiragana, and katakana,        and if a person selects a particular choice repeatedly it
along with occasional words in Roman script and other        will appear on the top of the list. The word processor
Western symbols. The Macintosh, which was the                can be set so that the first kanji choice, instead of the
first commercially successful computer with bitmapped         hiragana, appears in the text being written. Pressing
(relating to a digital image for which an array of bi-       the spacebar once would transform it to the second
nary data specifies the value of each pixel) screen          choice, and pressing again could select the next choice
and printing, became popular in Japan because it could       and open the selection window. The choices may in-
handle the language, but all Windows-based comput-           clude a katakana choice as well. Many choices exist,
ers can now as well, as, of course, can indigenous           and some Chinese word processors often fill the se-
Japanese word processors.                                    lection window four times over. Thus, research on the
    Kana computer keyboards exist in Japan, but the          frequency of usage of various Chinese words is im-
most common input method for Chinese characters              portant in establishing their most efficient ordering in
in both China and Japan requires the user to enter text      the selection window. Human-computer interaction
into a Western keyboard, romanizing the words.               (HCI) research has explored other ways of making the
Suppose that someone is using Microsoft Word in              word selection, including eye tracking to select the al-
Japanese and wants to type the word meaning “com-            ternative that the user’s eyes focus upon.
ment.” The writer would press the Western keys that               The chief substitutes for keyboard text input are
phonetically spell the Japanese word kannsou. If the         speech recognition and handwriting recognition.
word processor is set to do so, it will automatically dis-   Speech recognition systems developed for English are
                                                                                                ASIAN SCRIPT INPUT ❚❙❘ 49



unsuitable for Asian languages. Notably, spoken                  Modern word processors may change the balance
Chinese is a tonal language in which each syllable           of forces working for or against change in the tradi-
has a characteristic pitch pattern, an important feature     tional Asian scripts. They may degrade people’s Chinese
absent from English. Experts have done a good deal of        character handwriting skills, but they may simultane-
research on computer recognition of Japanese and             ously help people employ more obscure characters. In
Chinese, but speech input introduces errors while            the psychology of memory people have the ability to
requiring the same selection among choices, as does          recognize things they would not have spontaneously
keyboard input. Handwriting recognition avoids the           produced. Chinese-language and Japanese-language
problem of alternative ways of writing homonyms, but         word processors often include character palettes (com-
despite much research it remains excessively error           parable ranges, qualities, or uses of available elements),
prone. Three approaches are being tried with Chinese:        allowing users to select even obscure characters with
recognizing (1) the whole word, (2) the individual           a single click of the mouse, thereby perhaps encour-
characters, or (3) parts of characters, called “radicals,”   aging them to do so. Computer and information sci-
that may appear in many characters. All three ap-            entists and engineers are rapidly producing search
proaches have high error rates because many charac-          engines and a whole host of other tools that are giving
ters are graphically complex, and people vary                the ancient Asian scripts a new life on the Internet and
considerably in how they draw them. Thus, key-               the World Wide Web.
board input remains by far the most popular method.
                                                                      William Sims Bainbridge and Erika Bainbridge

East and West                                                See also Handwriting Recognition and Retrieval;
China, Japan, and Korea have from time to time               Keyboard
considered abandoning the traditional Chinese
characters, with Korea coming the closest to actually
doing so. A phonetic writing system is easier to             FURTHER READING
learn, thus giving students more time to study other
things. The traditional Chinese system supported an          Apple Computer Company. (1993). Macintosh Japanese input method
                                                                 guide. Cupertino, CA: Apple.
entrenched intellectual elite, who feared that a sim-        Asher, R. E., & Simpson, J. M. Y. (Eds.). (1994). The encyclopedia of lan-
ple alphabet might democratize writing. On the other             guage and linguistics. Oxford, UK: Pergamon.
hand, one advantage of the traditional system is that        Fujii, H., & Croft, W. B. (1993). A comparison of indexing techniques
a vast region of the world speaking many dialects                for Japanese text retrieval. In Proceedings of the 16th annual ACM
                                                                 SIGIR Conference on Research and Development in Information
and languages could be united by a single writing sys-           Retrieval (pp. 237–246). New York: ACM Press.
tem, and even today a Chinese person can commu-              Ho, F.-C. (2002). An analysis of reading errors in Chinese language. In
nicate to some extent with a Japanese person—even                L. Jeffrey (Comp.), AARE 2002 conference papers (n.p.).
                                                                 Melbourne, Australia: Australian Association for Research in
though neither knows the other’s spoken lan-                     Education.
guage—by drawing the characters. Fluent bilingual            Li,Y., Ding, X., & Tan, C. L. (2002). Combining character-based bigrams
readers of an Asian language and a Western language              with word-based bigrams in contextual postprocessing for Chinese
sometimes say they can read Chinese characters more              script recognition, ACM Transactions on Asian Language Information
                                                                 Processing, 1(4), 297–309.
quickly because the characters directly represent con-       Shi, D., Damper, R. I., & Gunn, S. R. (2003). Offline handwritten Chinese
cepts, whereas Western letters represent sounds and              character recognition by radical decomposition. ACM Transactions
thus only indirectly relate to concepts. Some writers            on Asian Language Information Processing, 2(1), 27–48.
have conjectured that dyslexia should be rare in             Wang, J. (2003). Human-computer interaction research and practice
                                                                 in China. ACM Interactions, 10(2), 88–96.
Chinese, if difficulties in learning to read are an in-       Wang, J., Zhai, S., & Su, H. (2001). Chinese input with keyboard and eye-
ability to connect letters with sounds. However, dyslexia        tracking. In Proceedings of the SIGCHI Conference on Human Factors
seems to exist in every language, although its causes            in Computing Systems (pp. 349–356). New York: ACM Press.
and characteristics might be somewhat different in
Asian languages than in English.
50 ❘❙❚ BERKSHIRE ENCYCLOPEDIA OF HUMAN-COMPUTER INTERACTION



                                                         like that of current scientific computers: curve-fit-
THE ATANASOFF-BERRY                                      ting, circuit analysis, structural analysis, quantum
                                                         physics, and problems in mechanics and astronomy.
      COMPUTER                                           The desktop calculators of the era were not up to the
                                                         equation-solving task, and Atanasoff identified their
The Atanasoff-Berry Computer (ABC) was the               limits as a common bottleneck in scientific research.
first electronic digital computer and the inspiration     His conception of a high-speed solution made sev-
for the better-publicized 1946 ENIAC. It was con-        eral unprecedented leaps: binary internal arithmetic
ceived in late 1938, prototyped in 1939 at Iowa State    (with automatic binary-decimal conversion), all-
College (now Iowa State University) in Ames,             electronic operation using logic gates, dynamically-
Iowa, and made usable for production computing           refreshed memory separated from the arithmetic
by 1941. John Atanasoff, a professor of mathemat-        units, parallel operation of up to thirty simultane-
ics and physics, collaborated with Clifford Berry, a     ous arithmetic units, and a synchronous system clock.
graduate student, to develop the system.                      The ABC achieved practical success at the curve-
                                                         fitting application. Atanasoff collaborated with a stat-
                                                         istician colleague at Iowa State, George Snedecor,
Physical Description                                     who supplied a steady stream of small linear-system
In contrast to the computers that followed in the        problems to the ABC. Snedecor’s secretary was given
1940s, the ABC was compact, movable, and easily          the task of checking the results by desk calculation,
operated by a single user. The original system no        which was simpler than solving the equations and
longer exists except for a logic module and a mem-       could be performed manually.
ory drum, but a functioning replica was constructed
in the late 1990s.
                                                         Human Interface
The Atanasoff-Berry Computer                             Compared to modern interfaces, the ABC interface
The ABC weighed about 750 pounds. It had the             resembled that of an industrial manufacturing
weight and maneuverability of an upright piano and       machine. The user controlled the system with throw
could roll on four heavy-duty casters. The total power   switches and card readers (decimal for input and bi-
it drew was less than a kilowatt, and the heat gen-      nary for intermediate results). The user was also
erated by its vacuum tubes was low enough to dis-        responsible for moving a jumper from one pair of
sipate without requiring fan-forced air. The ABC         contacts to another to indicate a particular vari-
used ordinary 117-volt line power. An electric mo-       able in the system of equations. The ABC commu-
tor synchronized to standard 60-hertz line voltage       nicated to the user through a few incandescent lamp
served as the system clock. The electromechanical        indicators, an ohmmeter to indicate correct work-
parts of the ABC, like those of a modern com-            ing voltages, a binary punch card output, and a cylin-
puter, were for purposes other than calculation;         drical readout for decimal numbers that resembled
the computing itself was completely electronic. The      a car odometer.
arithmetic modules were identical and could easily           The inventors clearly designed the machine for
be interchanged, removed, and repaired.                  operation by themselves, not general users. None of
                                                         the switches or lamps was labeled; it was up to the
                                                         user to remember what each switch did and what
Intended Applications                                    each lamp meant. One switch instructed the ABC to
                                                         read a base-10 punch card, convert it to binary,
and Production Use                                       and store it in the dynamic memory, for example.
The ABC was intended to solve dense systems of           Furthermore, the open design of the ABC pro-
up to thirty simultaneous linear equations with          vided far less protection from electric shock than a
15-decimal precision. Atanasoff targeted a workload      modern appliance does. Exposed surfaces only a few
                                                                                  ATTENTIVE USER INTERFACE ❚❙❘ 51



centimeters apart could deliver a 120-volt shock to
the unwary.                                                FURTHER READING
     A user entered the coefficients of the equations on
standard punch cards, using an IBM card punch. Each        Atanasoff, J. V. (1984). Advent of electronic digital computing. Annals
coefficient required up to fifteen decimals and a              of the History of Computing, 6(3), 229–282.
                                                           Burks, A. R. (2003). Who invented the computer? The legal battle that
sign, so five numbers fit onto a single eighty-column            changed computing history. Amherst, NY: Prometheus Books.
card. It was in the user’s best interest to scale up the   Burks, A. R., & Burks, A. W. (1989). The first electronic computer: The
values to use all fifteen decimals, since the arithmetic        Atanasoff story. Ann Arbor: University of Michigan Press.
                                                           Gustafson, J. (2000). Reconstruction of the Atanasoff-Berry com-
was fixed-point and accumulated rounding error.                 puter. In R. Rojas & U. Hashagen (Eds.), The first computers:
     Because the ABC could hold only two rows of               History and architectures (91–106). Cambridge, MA: MIT
coefficients in its memory at once, it relied on a mass         Press.
storage medium to record scratch results for later use.    Mackintosh, A. R. (1988, August). Dr. Atanasoff ’s computer. Scientific
                                                               American (pp. 72–78).
(The solution of two equations in two unknowns             Mollenhoff, C. R. (1988). Atanasoff: Forgotten father of the computer.
did not require scratch memory.) Since magnetic stor-          Ames: Iowa State University Press.
age was still in its infancy, Atanasoff and Berry devel-   Randell, R. (Ed.). (1982). The or ig ins of dig ital computers
oped a method of writing binary numbers using                  (pp. 305–325). New York: Springer-Verlag.
                                                           Reconstruction of the Atanasoff-Berr y Computer. (n.d.).
high-voltage arcs through a paper card. The pres-              Retrieved on January 27, 2004, from http://www.scl.ameslab
ence of a hole, representing a 1, was then readable with       .gov/ABC
lower voltage electrodes. Both reading and writing took    Sendov, B. (2003). John Atanasoff: The electronic Prometheus. Sofia,
                                                               Bulgaria: St. Kliment Ohridski University Press.
place at 1,500 bits per second, which was a remarkable     Silag, W. (1984). The invention of the electronic digital computer at
speed for input/output in 1940. However, the reliabil-         Iowa State College, 1930–1942. The Palimpsest, 65(5), 150–177.
ity of this system was such that a 1-bit error would oc-
cur every 10,000 to 100,000 bits, and this hindered the
ability to use the ABC for production computing be-
yond five equations in five unknowns.
     To obtain human-readable results, the ABC con-
verted the 50-bit binary values stored in the memory
to decimals on the odometer readout. The total process                ATTENTIVE USER
of converting a single 15-decimal number and mov-
ing the output dials could take anywhere from 1 sec-                    INTERFACE
ond to 150 seconds depending on the value of the
number. Atanasoff envisioned automating the man-           An attentive user interface is a context-aware hu-
ual steps needed for operation, but enhancement of         man-computer interface that uses a person’s at-
the ABC was interrupted by World War II and never          tention as its primary input to determine and act
resumed.                                                   upon a person’s intent. Although we can read a
     The ABC was a landmark in human-computer in-          person’s attention in her every word and action
teraction by virtue of being the first electronic com-      (even the way a person moves a cursor on a com-
puter. Its use of punch cards for the input of             puter interface shows what she is attending to), we
high-accuracy decimal data, binary internal represen-      usually read attention in what and how people look
tation, operator console, and the management of mass       at things.
storage and volatile storage were major advancements           Visual attentive user interfaces concentrate on
for the late 1930s when Atanasoff and Berry conceived      the autonomic (involuntary) and social responses
and developed it.                                          that eyes communicate and read such eye move-
                                                           ments as a lingering stare, a roving gaze, and a nerv-
                                        John Gustafson     ous blink in a language of ocular attention. Such
                                                           interfaces also monitor the order in which people
See also ENIAC                                             visually scan objects.
52 ❘❙❚ BERKSHIRE ENCYCLOPEDIA OF HUMAN-COMPUTER INTERACTION



Eye Tracking
Eye tracking is a technique that monitors a person’s
eye movements to determine where she is looking.
Eye tracking has long held promise as the ultimate hu-
man-computer interface, although eye tracking prod-
ucts have not been a commercial success. Original eye
tracking approaches used mechanical/optical instru-
ments that tracked mirrored contact lens reflections
or even instruments that measured eye muscle ten-
sion. Newer approaches illuminate the eye with in-
frared light and watch reflections with a camera.
Researchers can indirectly determine where a person’s
eye is focusing by noting that an electroencephalo-
gram (EEG) signal is dominated by an ocular stim-          Researcher Mike Li demonstrates the technology used
ulus. Four or five video strobe rates on different parts    in the Invision eye-tracking experiment. The balls on
of a display can be distinguished in an EEG. When a        the screen have names of companies that move around
person attends to one of them, his EEG pulses at the       as he looks at them. The object under the screen is
video strobe rate. Codings of attention on a screen        the eye tracker.               Photo courtesy of Ted Selker.
can be identified with an EEG frequency counter.

                                                           eye tracker enables the user’s gaze to roughly posi-
Attention Can Be Detected                                  tion the cursor, which the mouse can then manip-
Whereas the advertising and psychology fields              ulate. If the user wants to change the application
have long used eye movement to understand what a           window he is working with, he stares at the applica-
person is looking at, the human-computer interface         tion window that he wants to work in; this stare
field has struggled to use the eye as a controller. How-    “warps” the cursor to that application window.
ever, the breakthrough in visual attentive user inter-     MAGIC pointing speeds up context changes on
faces is in observing what the eye does, not in giving     the screen.
it a tracking task. Interest Tracker is a system that
monitors the time that a person spends gazing
over a title area instead of the time that the person      The Path of Attention Can
spends gazing at a specific character to determine se-
lection. For example, the title of an article is pre-      Demonstrate Intention
sented at the bottom of a computer screen. A user          During the late 1960s it was shown that the way that
might glance down to read the title; if his glance plays   a person’s eyes move while scanning a picture de-
over the title for more than .3 seconds a window           scribes aspects of what she is thinking. When re-
opens on the computer screen with the full article.        searchers asked viewers seven questions about a
That .3 seconds of dwell time is less than the typi-       painting entitled The Unexpected Visitor, seven iden-
cal 1 second that is required for a computer user to       tifiable eye-scan patterns were recognizable. The or-
select something on a screen by using a pointing de-       der in which a person looks at things also is a key to
vice. Interest Tracker registers whether a person is       what that person is thinking. Research on an ex-
paying attention to, for example, news feeds, stock        periment called “Invision” uses this fact in a user in-
prices, or help information and learns what titles         terface to prioritize activities. Invision’s grouping of
to audition at the bottom of the screen.                   things by the way a person looks improves eye track-
     MAGIC (Manual and Gaze Input Cascaded)                ing and uses gaze to group things of interest.
pointing is a technique that lets a computer mouse         Knowing that an eye moves between staring fixa-
manipulate what a user’s eyes look at on a screen. An      tions can help find those fixations. By analyzing the
                                                                            ATTENTIVE USER INTERFACE ❚❙❘ 53



eye-travel vectors between fixation vertices, Invision    a sensor can recognize many aspects of attention.
gains a much more accurate idea of what a person         EyeaRe consists of a Microchip PIC microprocessor
is trying to look at than by analyzing that person’s     that records and runs the system, an LED and a
dwell time on a particular item.                         photo diode looking at the eye, and another
     Attending to the order in which people look at      LED/photo diode pair that measures whether it is
things provides a powerful interface tool. Invision      in front of other EyeaRe devices and communicates
demonstrates that an attentive user interface can be     information. An IR channel communicates to a
driven from insights about where people look.            video base station or a pair of glasses.
Scenarios are created in which the attentive pattern          If an EyeaRe user is staring, the IR reflection off
of the eye gaze can be “understood” by a com-            his eye does not change. Staring at a video base sta-
puter. By watching the vertices of a person’s eye mov-   tion starts a video; glancing away stops it. The video
ing through a visual field of company names, the         image can detect whether a user is paying attention
system notices which ones interest the person.           to it; if the user doesn’t like it and blinks her eyes
The company names aggregate themselves into clus-        in frustration, the system puts up a more pleasing
ters on the screen based on the person’s scanning        image.
patterns.                                                     When two people stare at each other, EyeaRe
     A similar approach uses an ecological interface     uses the IR communication channel to exchange in-
that is an image of a kitchen with several problems.     formation. When one person stares at another per-
On the counter is a dish with some food on it; the       son, the person being stared at receives the contact
oven door is slightly ajar, as are the dishwasher        information of the person who is staring.
and refrigerator doors. The manner in which a per-            People tend to move their eyes until they have
son’s eyes move around the kitchen image allows the      to look 15 degrees to the side; EyeaRe has an 18-
interface to understand whether the person is hun-       degree horizontal field of view. Thus, gaze and blink
gry, thinking of taking care of problems, or think-      detection occurs when a person looks at the EyeaRe
ing about something else in the kitchen. The             base station or glasses. EyeaRe demonstrates that a
interface uses the order in which the person views       system that doesn’t even track the eye can under-
things in the image to bring up a menu and so forth.     stand the intentions of attention.
     This approach aggregates eye motions into a
story of what the person wants to do. The attention
model drives the interface. The vertices of change       A Simple Attentive
in direction of eye movements easily give focus lo-
cations that have eluded most eye tracking research.     Eye-Gesture Language
                                                         To take eye communication one step further, the Eye
                                                         Bed interface uses an eye-gesture language to perform
Ocular Attention                                         tasks that are helpful to a person lying in bed. The Eye
                                                         Bed demonstrates that computers can be attentive to
without Eye Tracking                                     people’s need to be horizontal eight hours a day. The
EyeaRe is an ocular attention system that is based       Eye Bed interface uses eye tracking housed in a con-
on the fact that many of the social cues that are made   verted lamp hanging over the head of the person in
by an eye do not depend on where the eye is look-        bed. This interface easily distinguishes between star-
ing. In fact, EyeaRe has no eye tracking system. It      ing at an object on the ceiling and glancing around in-
simply measures reflected infrared (IR) from the         differently. A language of attentional eye gestures drives
sclera (the opaque white outer coat enclosing the        the scenario. Glancing around shows lack of attention,
eyeball except the part covered by the cornea) and       whereas staring demonstrates attention. Blinking a
pupil to a photo diode. The system uses this reflected    long wink-like blink means selection. Blinking rapidly
infrared to determine whether the eye is open, closed,   means dislike. Closing the eyes could mean that the
blinking, winking, or staring. Without a camera such     user is going to sleep; thus, a sunset and a nighttime
54 ❘❙❚ BERKSHIRE ENCYCLOPEDIA OF HUMAN-COMPUTER INTERACTION



scenario begin. Opening the eyes makes a morning           their attention—even encouraging their ocular focus
and wakeup scenario begin. Intelligent systems ana-        to be on what they want to do. Attentive user inter-
lyze a person’s reactions to media on music and video      faces allow people’s attention to make things happen.
jukeboxes. The media offerings are auditioned to de-
tect the attention shown them. Blinking when one                                                                        Ted Selker
doesn’t like the media makes the system know that
it should choose other music or video to show the per-     See also Eye Tracking
son. Winking or closing the eyes turns off the system.
The reading of eye gestures becomes an attentive user
interface.                                                 FURTHER READING
     Understanding attention requires a model of what
eye movement means. Researchers can make a com-            Bolt, R. A. (1985). Conversing with computers. Technology Review, 88(2),
plexity of interfaces from some simple observations            34–43.
of eye behavior. As an output device the eye is a          Gregory, R. L. (1997). Eye and brain: The psychology of seeing. Oxford,
                                                               UK: Oxford University Press.
simpler user interface tool than is normally described.    Guo, X. (1999). Eye contact—Talking about non-verbal communica-
The eye can easily be used with a language of closing,         tion: A corpus study. Retrieved April 29, 2004, from http://www.lan-
opening, blinking, winking, making nervous move-               guagemagazine.com/internetedition/ma99/sprpt35.html
ments, glancing around, and staring. This language         Maglio, P. P., Barrett, R., Campbell, C. S., & Selker, T. (2000). SUITOR:
                                                               An attentive information system. New York: ACM Press.
can be sensed with eye-tracking cameras or with a          Morimoto, D., & Flickner, M. (2000). Pupil detection using multiple light
simple reflected LED, as the EyeaRe system demon-               sources. Image and Vision Computing, 18, 331–335.
strates.                                                   Nervous TV newscasters blink more. (1999). Retrieved April 29, 2004,
                                                               from http://www.doctorbob.com/news/7_24nervous.html
                                                           Rice, R., & Love, G. (1987). Electronic emotion: Socioemotional content
                                                               in a computer-mediated communication. Communication Research,
Promises of the Future                                         14(1), 85–108.
Attentive user interfaces hold great promise. People       Russell, S., & Norvig, P. (1995). Artificial intelligence: A modern approach.
                                                               Upper Saddle River, NJ: Prentice Hall.
are now in a position to implement and extend such         Selker, T., & Burleson, W. (2000). Context-aware design and interaction
interfaces. The hardware to create and test them is eas-       in computer systems. IBM Systems Journal, 39(3–4), 880–891.
ily accessible. With the use of the eye as a secondary     Shepard, R. N. (1967). Recognition memory for words, sentences and
indicator of intention, researchers can make robust            pictures. Journal of Verbal Learning and Verbal Behavior, 6, 156–163.

and computationally simple visual interfaces.
     Models of human intention and attention are be-
coming part of all human-computer interfaces. The
context of where we are and what we are doing can                             AUGMENTED
accomplish more than automatically opening the gro-
cery store door. Many interfaces can be driven com-                            COGNITION
pletely by noticing a person’s attention.
     Sensors in a given context can detect many things     Augmented cognition is a field of research that seeks
about human attention. For example, a sensor pad in        to extend a computer user’s abilities via technologies
front of an office door can detect if a person has ar-      that address information-processing bottlenecks in-
rived to visit. Many biometrics (relating to the sta-      herent in human-computer interaction (HCI). These
tistical analysis of biological observations and           bottlenecks include limitations in attention, mem-
phenomena) such as EEG changes, sweat responses,           ory, learning, comprehension, visualization abilities,
and heart rate variability are candidates for attentive    and decision-making. Limitations in human cogni-
user interfaces.                                           tion (the act or process of knowing) are due to in-
     People want to focus on what they are doing           trinsic restrictions in the number of mental tasks
and on the people they are with. Attentive user in-        that a person can execute at one time, and these re-
terfaces can detect people’s intentions without taking     strictions may fluctuate from moment to moment
                                                                                    AUGMENTED COGNITION ❚❙❘ 55



depending on a host of factors, including mental fa-         Early Investments in Related Work
tigue, novelty, boredom, and stress.                         Augmented cognition does not draw from just one
    As computational interfaces have become more             scientific field—it draws from fields such as neuro-
prevalent in society and increasingly complex with           science, biopsychology, cognitive psychology, human
regard to the volume and type of information pre-            factors, information technology, and computer
sented, researchers have investigated novel ways to          science. Each of these fields has itself undergone a
detect these bottlenecks and have devised strategies         substantial revolution during the past forty years that
to aid users and improve their performance via tech-         has allowed the challenges raised by researchers to
nologies that assess users’ cognitive status in real time.   begin to be investigated. Although many individ-
A computational interaction monitors the state of a          ual research projects contributed to the general de-
user through behavioral, psychophysiological, and/or         velopment and direction of augmented cognition,
neurophysiological data and adapts or augments the           several multimillion-dollar projects helped shape the
computational interface to significantly improve users’       foundation on which the field is built.
performance on the task at hand.                                 Since the invention of the electronic computer,
                                                             scientists and engineers have speculated about the
                                                             unique relationship between humans and comput-
Emergence of                                                 ers. Unlike mechanized tools, which are primarily
                                                             devices for extending human force and action, the
Augmented Cognition                                          computer became an entity with which humans
The cognitive science and HCI communities have               forged an interactive relationship, particularly as
researched augmented cognition for several decades.          computers came to permeate everyday life. In 1960
Scientific papers in this field increased markedly dur-        one of the great visionaries of intelligent comput-
ing the late 1990s and addressed efforts to build and        ing, J. C. R. Licklider, wrote a paper entitled “Man-
use models of attention in information display and           Computer Symbiosis.” Licklider was director of the
notification systems. However, the phrase “aug-              Information Processing Techniques Office (IPTO)
mented cognition” associated with this research did          at the Defense Department’s Advanced Research
not find widespread use until the year 2000, when             Projects Agency (ARPA) during the 1960s. In his pa-
a U.S. Defense Department Defense Advanced                   per he stated, “The hope is that, in not too many
Research Project Agency (DARPA) Information                  years, human brains and computing machines will
Science and Technology (ISAT) group study and a              be coupled together very tightly, and that the resulting
workshop on the field at the National Academy of              partnership will think as no human brain has ever
Sciences were held. During the year 2002 the num-            thought and process data in a way not approached
ber of papers about augmented cognition increased            by the information-handling machines we know to-
again. This increase was due, in part, to the start of       day” (Licklider 1960, 4). Almost prophetic, this de-
a DARPA research program in augmented cognition              scription of the symbiotic relationship between
in 2001 with a focus on challenges and opportuni-            humans and computers is one of the first descrip-
ties with the real-time monitoring of cognitive states       tions of what could be considered an augmented cog-
with physiological sensors. This substantial invest-         nition computational system. Although research on
ment in these developing technologies helped bring           this topic was not conducted during his tenure at
together a research community and stimulated a set           ARPA during the 1960s, Licklider championed the
of thematically related projects on addressing cog-          research that developed into the now-burgeoning
nitive bottlenecks via the monitoring of cognitive           field of computer science, including creation of the
states. By 2003 the augmented cognition field ex-            Arpanet computer network (forerunner of the
tended well beyond the boundaries of those specific           Internet). His research, vision, and direction had a
Defense Department research projects, but that ini-          significant impact on both computer science and in-
tial investment provided impetus for the infant field         formation technology and set the stage for the
to begin to mature.                                          field of augmented cognition.
56 ❘❙❚ BERKSHIRE ENCYCLOPEDIA OF HUMAN-COMPUTER INTERACTION



    During the early 1960s researchers speculated        corporate any physiological monitoring. Cognitive
that electrical signals emanating from a human brain     modeling was the cornerstone of the pilot-vehicle in-
in the form of electroencephalographic (EEG)             terface (PVI), which had the critical task of manag-
recordings could be used as indicators of specific       ing all pilot interactions with the system by inferring
events in human cognitive processing. Several            the pilot’s intentions and communicating these in-
Department of Defense investigations into detect-        tentions to the other components of the PA system.
ing these signals and other measurements oc-             The PVI was also responsible for modeling pilot work-
curred through the biocybernetics and learning           load to adapt and configure the information displays
strategies programs sponsored by ARPA during             in the cockpit, conveying workload information to
the 1970s and 1980s. The earliest program was            the other subsystems, and compensating for pilot be-
biocybernetics, which tested the hypothesis that EEG     havior that might result in an error. An example of
activity might be able to control military devices and   this work was a PA program at NASA-Ames Research
serve as indicators of user performance. In this pro-    Center that explored the use of probabilistic mod-
gram biocybernetics was defined as a real-time con-       els of a pilot’s goals and workload over time, based
nection between the operator and computational           on multiple inputs and the use of models to con-
system via physiological signals recorded during spe-    trol the content and complexity of displays. Such
cific tasks. Both the biocybernetics and learning        models did not employ physiological measures of a
strategies programs centered around the creation of      pilot’s cognitive status.
closed-loop feedback systems (the relationship be-            Other research occurred in the academic and pri-
tween user and computational system, where changes       vate sectors, including the attentional user interface
in the computational interface are driven by detected    (AUI) project at Microsoft Research during the
changes in the user’s physiological status, which in     late 1990s, which provided conceptual support to ef-
turn change as a result of the new format of the in-     forts in augmented cognition. Researchers developed
terface) between operator and computer for the           methods for building statistical models of attention
selection and training of personnel, display/con-        and workload from data. Researchers built archi-
trol design, and online monitoring of operator sta-      tectures to demonstrate how cognitive models could
tus (although with slightly different military           be integrated with real-time information from mul-
application domains between the two programs).           tiple sensors (including acoustical sensing, gaze and
    In both programs researchers saw the real-time       head tracking, and events representing interaction
identification of cognitive events as critical to un-     with computing systems) to control the timing
derstanding the best methods for aiding military         and communication medium of incoming notifica-
users in a rapid and contextually appropriate way.       tions. AUI work that included psychological studies
However, when this research was begun, both              complemented the systems and architectures work.
computational systems and neuroscience were in
their infancy, and the results of this research were
not incorporated into production military sys-           Foundations of Augmented Cognition
tems. Augmented cognition can be viewed as a de-         In light of these earlier research efforts, the logical
scendant of these early programs.                        question arises: What sets augmented cognition apart
    Another investigation in this field was the Pilot’s   from what has already been done? As mentioned,
Associate (PA) program sponsored by DARPA dur-           augmented cognition relies on many fields whose
ing the 1980s and early 1990s. Pilot’s Associate was     maturity is critical for its success. Although programs
an integrated system of five components that in-         such as biocybernetics during the 1970s had similar
corporated AI (artificial intelligence) techniques and    goals, they did not have access to the advanced com-
cognitive modeling to aid pilots in carrying out their   putational power necessary to process brain sig-
missions with increased situational awareness and        nals in real time, nor did researchers know enough
enhanced decision-making. Unlike biocybernetics,         about those signals to use them to control displays
PA utilized cognitive modeling alone and did not in-     or machines. Likewise, the Pilot’s Associate program
                                                                                    AUGMENTED COGNITION ❚❙❘ 57



during the 1980s shared many aspirations of today’s
augmented cognition, namely to develop adaptive               INTELLIGENT AGENT Software program that actively lo-
interfaces to reduce pilot workload. However, PA              cates information for you based on parameters you set.
could assess the status of a pilot from inferences and        Unlike a search engine or information filter, it actively
models based only on the pilot’s overt behavior and           seeks specific information while you are doing
the status of the aircraft. What distinguishes aug-           other things.
mented cognition is its capitalization on advances
in two fields: behavioral/neural science and com-
puter science.                                              ical sensors, but the field of augmented cognition is
     At the start of the twenty-first century researchers    broadened even further by their inclusion.
have an unparalleled understanding of human brain                As a result of the “Decade of the Brain,” research-
functioning. The depth of this understanding is due to      ers have an increased knowledge of the cognitive lim-
the development of neuroscientific techniques funded         itations that humans face. The HCI field focuses on
by the U.S. National Institutes of Health (NIH) and         the design, implementation, and evaluation of in-
other agencies during the 1990s, a period now referred      teractive systems in the context of a user’s work.
to as the “Decade of the Brain.” The billion-dollar fund-   However, researchers in this field can work only
ing of the fields of neuroscience, cognitive science, and    with the data and observations easily accessible to
biopsychology resulted in some of the greatest advances     them, that is, how people overtly behave while using
in our understanding of the human biological system         interfaces. Through efforts in neuroscience, biopsy-
in the twentieth century. For example, using techniques     chology, and cognitive neuroscience we can locate and
such as functional magnetic resonance imaging (fMRI),       measure activity from the brain regions that are ac-
scientists were able to identify discrete three-dimen-      tively involved in day-to-day information-processing
sional regions of the human brain active during spe-        tasks. Researchers will have a greater understanding
cific mental tasks. This identification opened up the         of the cognitive resources that humans possess and
field of cognitive psychology substantially (into the        how many of these resources are available during a
new field of cognitive neuroscience) and enabled re-         computationally based task, whether or not the com-
searchers to test their theories of the human mind and      putational systems will include advanced sensors. After
associate previously observed human thoughts and            these cognitive resources are identified and their ac-
behaviors with neural activity in specific brain regions.    tivity (or load) measured, designers of computational
Additional investment from the Department of Defense        interfaces can begin to account for these limitations
and other agencies during the twenty-first century has       (and perhaps adapt to their status) in the design of
allowed researchers to develop even more advanced           new HCI systems.
sensors that will eventually be used in augmented cog-           Finally, without advances in computer science
nition systems. Novel types of neurophysiological sig-      and engineering, none of the neuroscientific devel-
nals that are measurable noninvasively include electrical   opments listed here would be possible, and the field
signals—using electroencephalography and event-re-          of augmented cognition would certainly not be fea-
lated potentials (identifiable patterns of activity within   sible. During the past forty years society has experi-
the EEG that occur either before specific behaviors are      enced leaps in computational prowess and the
carried out, or after specific stimuli are encountered)—     sophistication of mathematical algorithms. These
and local cortical changes in blood oxygenation             leaps have been due in part to the miniaturization of
(BOLD), blood volume, and changes in the scattering         transistors and other silicon-based components so
of light directly due to neuronal firing (using near         that more computational power is available per square
infrared [NIR] light). Most of these signals, unlike        inch of hardware. This miniaturization has allowed
fMRI, can be collected from portable measurement            computers to shrink in size until they have perme-
systems in real time, making them potentially avail-        ated the very fabrics that people wear and even
able for everyday use. All augmented cognition systems      their environments. Computer code itself has become
do not necessarily contain advanced neurophysiolog-         smaller and more flexible, with the emergence of
58 ❘❙❚ BERKSHIRE ENCYCLOPEDIA OF HUMAN-COMPUTER INTERACTION



agent-based computing (the instantiation of active,       tention, memory, learning, comprehension, sensory
persistent software components that perceive, rea-        processing, visualization abilities, qualitative judg-
son, act, and communicate in software code), JAVA,        ments, serial processing, and decision-making. For an
and Internet services. Thus, augmented cognition has      augmented cognition system to be successful it
benefited from two computing advances—improve-             must identify at least one of these bottlenecks in real
ments in raw computational resources (CPUs, phys-         time and alleviate it through a performance-enhanc-
ical memory) and improvements in the languages            ing mitigation strategy. Such mitigation strategies are
and algorithms that make adaptive interfaces possi-       conveyed to the user through the adaptive interface
ble. Many other fields have benefited from these          and might involve modality switching (between
advances as well and in turn have fed into the aug-       visual, auditory, and haptic [touch]), intelligent in-
mented cognition community. These fields include           terruption, task negotiation and scheduling, and as-
user modeling, speech recognition, computer vision,       sisted context retrieval via book marking. When a
graphical user interfaces, multimodal interfaces, and     user state is correctly sensed, an appropriate strat-
computer learning/artificial intelligence.                 egy is chosen to alleviate the bottleneck, the inter-
                                                          face is adapted to carry out the strategy, and the
Components of an                                          resulting sensor information indicates that the aiding
                                                          has worked—only then has a system “closed the loop”
Augmented Cognition System                                and successfully augmented the user’s cognition.
At the most general level, augmented cognition har-
nesses computation and knowledge about human
limitations to open bottlenecks and address the bi-       Applications of Augmented Cognition
ases and deficits in human cognition. It seeks to         The applications of augmented cognition are nu-
accomplish these goals through continual back-            merous, and although initial investments in systems
ground sensing, learning, and inferences to under-        that monitor cognitive state have been sponsored by
stand trends, patterns, and situations relevant to a      military and defense agencies, the commercial sec-
user’s context and goals. At its most general level, an   tor has shown interest in developing augmented cog-
augmented cognition system should contain at least        nition systems for nonmilitary applications. As
four components—sensors for determining user              mentioned, closely related work on methods and ar-
state, an inference engine or classifier to evaluate in-   chitectures for detecting and reasoning about a user’s
coming sensor information, an adaptive user in-           workload (based on such information as activity with
terface, and an underlying computational architecture     computing systems and gaze) have been studied
to integrate the other three components. In reality a     for nonmilitary applications such as commercial no-
fully functioning system would have many more             tification systems and communication. Agencies such
components, but these are the most critical.              as NASA also have shown interest in the use of meth-
Independently, each of these components is fairly         ods to limit workload and manage information over-
straightforward. Much augmented cognition research        load. Hardware and software manufacturers are
focuses on integrating these components to “close         always eager to include technologies that make their
the loop” and create computational systems that           systems easier to use, and augmented cognition sys-
adapt to their users.                                     tems would likely result in an increase in worker pro-
     Thus, the primary challenge with augmented cog-      ductivity with a savings of both time and money
nition systems is not the sensors component (although     to companies that purchased these systems. In more
researchers are using increasingly complex sensors).      specific cases, stressful jobs that involve constant in-
The primary challenge is accurately predicting/as-        formation overload from computational sources,
sessing, from incoming sensor information, the cor-       such as air traffic control, would also benefit from
rect state of the user and having the computer select     such technology. Finally, the fields of education and
an appropriate strategy to assist the user at that        training are the next likely targets for augmented
time. As discussed, humans have limitations in at-        cognition technology after it reaches commercial vi-
                                                                                                             AUGMENTED REALITY ❚❙❘ 59



ability. Education and training are moving toward                            Wilson, G. F. (2001). Real-time adaptive aiding using psychophysio-
an increasingly computational medium. With dis-                                 logical operator state assessment. In D. Harris (Ed.), Engineering
                                                                                psychology and cognitive ergonomics (pp. 175–182). Aldershot, UK:
tance learning in high demand, educational systems                              Ashgate.
will need to adapt to this new nonhuman teaching                             Wilson, R. A., & Keil, F. C. (Eds.). (2001). The MIT encyclopedia of the
interaction while ensuring quality of education.                                cognitive sciences (MITECS). Cambridge, MA: MIT Press.
Augmented cognition technologies could be applied
to educational settings and guarantee students a
teaching strategy that is adapted to their style of learn-
ing. This application of augmented cognition could                               AUGMENTED REALITY
have the biggest impact on society at large.
                                                                             Augmented reality is a new field of research that con-
                         Dylan Schmorrow and Amy Kruse                       centrates on integrating virtual objects into the
                                                                             real world. These virtual objects are computer graph-
See also Augmented Reality; Brain-Computer                                   ics displayed so that they merge with the real world.
Interfaces; Information Overload                                             Although in its infancy, augmented reality holds out
                                                                             the promise of enhancing people’s ability to perform
                                                                             certain tasks. As sensing and computing technolo-
FURTHER READING                                                              gies advance, augmented reality is likely to come to
                                                                             play a significant role in people’s daily lives.
Cabeza, R., & Nyberg, L. (2000). Imaging cognition II: An empirical
   review of 275 PET and fMRI studies. Journal of Cognitive
   Neuroscience, 12(1), 1–47.
Dix, A., Finlay, J., Abowd, G., & Beale, R. (1998). Human computer in-
   teraction (2nd ed.). London, New York: Prentice Hall.
                                                                             Augmented Reality
Donchin, E. (1989). The learning strategies project. Acta Psychologica,
   71(1–3), 1–15.
                                                                             and Virtual Reality
Freeman, F. G., Mikulka, P. J., Prinzel, L. J., & Scerbo, M. W. (1999).      An augmented-reality system merges the real scene
   Evaluation of an adaptive automation system using three EEG in-           viewed by the user with computer-generated virtual
   dices with a visual tracking task. Biological Psychology, 50(1), 61–76.
Gevins, A., Leong, H., Du, R., Smith, M. E., Le, J., DuRousseau, D.,
                                                                             objects to generate a composite view for the user.
   Zhang, J., & Libove, J. (1995). Towards measurement of brain func-        The virtual objects supplement the real scene with
   tion in operational environments. Biological Psychology, 40, 169–186.     additional and useful information. Sounds may be
Gomer, F. (1980). Biocybernetic applications for military systems.           added through the use of special headphones that
   Chicago: McDonnell Douglas.
Gray, W. D., & Altmann, E. M. (2001). Cognitive modeling and hu-             allow the user to hear both real sounds and syn-
   man-computer interaction. In W. Karwowski (Ed.), International            thesized sounds. There are also special gloves that
   encyclopedia of ergonomics and human factors (pp. 387–391). New           a user can wear that provide tactile sensation such
   York: Taylor & Francis.
Horvitz, E., Pavel, M., & Schmorrow, D. D. (2001). Foundations of aug-
                                                                             as hardness or smoothness. A user wearing such
   mented cognition. Washington, DC: National Academy of Sciences.           gloves could “feel” virtual furniture in a real room.
Humphrey, D. G., & Kramer, A. F. (1994). Toward a psychophysio-              In an augmented-reality system, users can walk
   logical assessment of dynamic changes in mental workload. Human           around a real room, hear the echo of their footsteps,
   Factors, 36(1), 3–26.
Licklider, J. C. R. (1960). Man-computer symbiosis: IRE transac-
                                                                             and feel the breeze from an air conditioning unit,
   tions on human factors in electronics. HFE-1 (pp. 4–11).                  while at the same time they can see computer-gen-
Lizza, C., & Banks, S. (1991). Pilot’s Associate: A cooperative,             erated images of furniture or paintings.
   knowledge-based system application. IEEE Intelligent Systems, 6(3),           One of the requirements of an augmented-real-
   18–29.
Mikulka, P. J., Scerbo, M. W., & Freeman, F. G. (2002). Effects of a bio-    ity system is that it needs to be interactive in real
   cybernetic system on vigilance performance. Human Factors, 44,            time. Animation, sound, and textures are added in
   654–664.                                                                  real time so that what the user sees, hears, and feels
Prinzel, L. J., Freeman, F. G., Scerbo, M. W., Mikulka, P. J., & Pope,
   A. T. (2000). A closed-loop system for examining psychophysio-
                                                                             reflects the true status of the real world. The most
   logical measures for adaptive task allocation. International Journal      important characteristic of augmented reality is the
   of Aviation Psychology, 10, 393–410.                                      ability to render objects in three-dimensional space,
60 ❘❙❚ BERKSHIRE ENCYCLOPEDIA OF HUMAN-COMPUTER INTERACTION



which makes them much more realistic in the eyes             This reduces the safety risk, since the user can see the
of the user. Virtual objects are drawn in relationship       real world in real time. If there is a power failure, the
to the real objects around them, both in terms of po-        user will still be able to see as well as he or she would
sition and size. If a virtual object is situated partially   when wearing dark sunglasses. If there is some kind
behind a real object (or vice versa) then the user           of hazard moving through the area—a forklift, for
should not see part of the obscured object. Occlusion        example—the wearer does not have to wait for the sys-
of objects is the largest contributor to human depth         tem to process the image of the forklift and display
perception.                                                  it; the wearer simply sees the forklift as he or she
     The major difference between augmented reality          would when not wearing the HMD. One disadvan-
and virtual reality is that in virtual reality everything    tage is that the virtual objects may appear to lag
that is sensed by the user is computer generated.            behind the real objects; this happens because the
Therefore the virtual objects must be rendered as            virtual objects must be processed, whereas real objects
photorealistically as possible in order to achieve the       do not need to be. In addition, some users are reluc-
feeling of immersion. Augmented reality uses both            tant to wear the equipment for fear of harming their
real and synthetic sights, sounds, and touches to con-       vision, although there is no actual risk, and other users
vey the desired scene, so virtual objects do not bear        dislike the equipment’s cumbersome nature. A new
the entire burden of persuading the user that the scene      version of the see-through HMD is being developed
is real, and therefore they do not need to be so pho-        to resemble a pair of eyeglasses, which would make
torealistic. Augmented reality lies in the middle of the     it less cumbersome.
continuum between absolute reality (in which every-               Closed-view HMDs cannot be seen through.
thing sensed is real) and virtual reality (in which          They typically comprise an opaque screen in front
everything that is sensed is created).                       of the wearer’s eyes that totally blocks all sight of the
                                                             real world. This mechanism is also used for tradi-
                                                             tional virtual reality. A camera takes an image of the
Different Types of Displays                                  real world, merges it with virtual objects, and
                                                             presents a composite image to the user. The advan-
for Augmented Reality                                        tage the closed-view has over the see-through ver-
Most people depend on vision as their primary sen-           sion of the HMD is that there is no lag time for
sory input, so here we will discuss several types of         the virtual objects; they are merged with the real
visual displays that can be used with augmented              scene before being presented to the user. The dis-
reality, each with its own advantages and disadvan-          advantage is that there is a lag in the view of the real
tages. Visual displays include head-mounted displays         world because the composite image must be
(HMDs), monitor-based displays, projected images,            processed before being displayed. There are two
and heads-up displays (HUDs).                                safety hazards associated with closed-view HMD.
                                                             First, if the power supply is interrupted, the user is
Head-Mounted Displays                                        essentially blind to the world around him. Second,
HMDs are headsets that a user wears. HMDs can                the user does not have a current view of the real
either be see-through or closed view. The see-through        world. Users have the same concerns and inhibitions
HMD works as its name implies: The user looks                regarding closed-view HMD as they do regarding
through lenses to see the real world, but the lenses         see-through HMD.
are actually display screens that can have graphics
projected onto them. The biggest advantage of the            Monitor-Based Displays
see-through HMD mechanism is that it is simple               Monitor-based displays present information to the
to implement because the real world does not have            user for configuring an augmented-reality system
to be processed and manipulated; the mechanism’s             this way. First, because a monitor is a separate dis-
only task is to integrate the visual augmentations.          play device, more information can be presented to
                                                                                   AUGMENTED REALITY ❚❙❘ 61



the user. Second, the user does not have to wear          involves placing the virtual objects in the proper lo-
(or carry around) heavy equipment. Third, graphi-         cations in the real world. This is an important ele-
cal lag time can be eliminated because the real world     ment of augmented reality and includes sensing,
and virtual objects are merged in the same way they       calibration, and tracking. Appearance concerns what
are for closed-view HMDs. The safety risk is avoided      the virtual objects look like. In order to achieve seam-
because the user can see the real world in true real      less merging of real and virtual objects, the virtual
time.                                                     objects must be created with realistic color and
     There are also some drawbacks to using moni-         texture.
tor-based displays instead of HMDs. First, the user           In virtual-reality systems, tracking the relative
must frequently look away from the workspace in           position and motion of the user is an important
order to look at the display. This can cause a slow-      research topic. Active sensors are widely used to track
down in productivity. Another problem is that the         position and orientation of points in space. The
user can see both the real world and—on the               tracking information thus obtained is fed into the
monitor—the lagging images of the real world. In          computer graphics system for appropriate render-
a worse case situation in which things in the scene       ing. In virtual reality, small errors in tracking can be
are moving rapidly, the user could potentially see a      tolerated, as the user can easily overlook those errors
virtual object attached to a real object that is no       in the entirely computer-generated scene. In aug-
longer in the scene.                                      mented-reality systems, by contrast, the registration
                                                          is performed in the visual field of the user. The
Projected-Image Displays                                  type of display used in the system usually determines
Projected-image displays project the graphics and         the accuracy needed for registration.
annotations of the augmented-reality system onto              One popular registration technique is vision-
the workspace. This method eliminates the need for        based tracking. Many times, there are fiducials
extra equipment and also prevents the user from hav-      (reference marks) marked out in the scene in which
ing to look away from the work area to check the          the virtual objects need to be placed. The system rec-
monitor-based display. The biggest disadvantage is        ognizes these fiducials automatically and determines
that the user can easily occlude the graphics and an-     the pose of the virtual object with respect to the scene
notations by moving between the projector and             before it is merged. There are also techniques that
the workspace. Users also can put their hands and         use more sophisticated vision algorithms to deter-
arms through the projected display, reducing their        mine the pose without the use of fiducials. The mo-
sense of the reality of the display.                      tion of the user and the structure of the scene are
                                                          computed using projective-geometry formulation.
Heads-Up Displays                                         (Projective geometry is the branch of geometry that
Heads-up displays are very similar to see-through         deals with projecting a geometric figure from one
HMDs. They do not require the user to wear special        plane onto another plane; the ability to project points
headgear, but instead display the data on a see-through   from one plane to another is essentially what is
screen in front of the user. As in seethrough HMDs,       needed to track motion through space.)
these systems are easy to implement, however, there           For a seamless augmented-reality system, it is
may be a lag time in rendering the virtual object.        important to determine the geometry of the vir-
                                                          tual object with respect to the real scene, so that
                                                          occlusion can be rendered appropriately. Stereo-based
Challenges in Augmented Reality                           depth estimation and the z-buffer algorithm (an al-
A majority of the challenges facing augmented real-       gorithm that makes possible the representation of
ity concern the virtual objects that are added to the     objects that occlude each other) can be used for
real world. These challenges can be divided into two      blending real and virtual objects. Also, using research
areas: registration and appearance. Registration          results in radiosity (a technique for realistically
62 ❘❙❚ BERKSHIRE ENCYCLOPEDIA OF HUMAN-COMPUTER INTERACTION



simulating how light reflects off objects), it is possi-   is placed on top of it on the television video to make
ble to “shade” the virtual object appropriately so that   it easier for those watching the game on television
it blends properly with the background scene.             to follow the rapid motion of the puck. Augmented
                                                          reality could also make possible a type of virtual set,
                                                          very similar to the blue-screen sets that are used
Applications                                              today to film special effects. Augmented-reality
Augmented reality has applications in many fields.         sets would be interactive, would take up less space,
In medicine, augmented reality is being researched        and would potentially be simpler to build than tra-
as a tool that can project the output of magnetic res-    ditional sets. This would decrease the overall cost of
onance imaging (MRI), computed tomography (CT)            production. Another example, already developed is
scans, and ultrasound imaging onto a patient to           the game AR2 Hockey, in which the paddles and field
aid in diagnosis and planning of surgical operations.     (a table, as in air hockey) are real but the puck is vir-
Augmented reality can be used to predict more ac-         tual. The computer provides visual tracking of the
curately where to perform a biopsy for a tiny tumor:      virtual puck and generates appropriate sound effects
All the information gathered from traditional meth-       when the paddles connect with the puck or when the
ods such as MRIs can be projected onto the patient        puck hits the table bumpers.
to reveal the exact location of the tumor. This en-            One military application is to use the technol-
ables a surgeon to make precise incisions, reducing       ogy to aim weapons based on the movement of the
the stress of the surgery and decreasing the trauma       pilot’s head. Graphics of targets can be superimposed
to the patient.                                           on a heads-up display to improve weapons’ accuracy
     In architecture and urban planning, annotation       by rendering a clearer picture of the target, which
and visualization techniques can be used to show          will be hard to miss.
how the addition of a building will affect the sur-            Many examples of assembly augmented-real-
rounding landscape. Actually seeing the future build-     ity systems have been developed since the 1990s.
ing life sized, in the location it will occupy, gives a   One of the best known is the Boeing wire-bundling
more accurate sense of the project than can be            project, which was started in 1990. Although well
conveyed from a model. Augmented-reality simula-          known, this project has not yet been implemented
tions also make it easier to recognize potential prob-    in a factory as part of everyday use. The goal is
lems, such as insufficient natural lighting for a         relatively straightforward: Use augmented reality to
building.                                                 aid in the assembly of wire bundles used in Boeing’s
     Augmented reality also has the potential to let      747 aircraft.
developers, utility companies, and home owners “see”           For this project, the designers decided to use a
where water pipes, gas lines, and electrical wires        see-through HMD with a wearable PC to allow work-
are run through walls, which is an aid when it comes      ers the freedom of movement needed to assemble
to maintenance or construction work. In order for         the bundles, which were up to 19 meters long. The
this technique to be implemented, the data must           subjects in the pilot study were both computer sci-
be stored in a format the augmented-reality system        ence graduate students who volunteered and Boeing
can use. Simply having a system that can project the      employees who were asked to participate.
images of electrical wiring on a wall would not be             The developers ran into both permanent and
sufficient; the system first must know where all the        temporary problems. One temporary problem, for
wires are located.                                        example, was that the workers who participated in
     Augmented reality has the potential to make a        the pilot program were typically tired because
big impact on the entertainment industry. A sim-          the factory was running the pilot study at one of
ple example is the glowing puck that is now used          the busier times in its production cycle. Workers
in many televised hockey games. In this application,      first completed their normal shift before working
the hockey puck is tracked and a brightly colored dot     on the pilot project. Another temporary problem
                                                                                   AUGMENTED REALITY ❚❙❘ 63



was the curiosity factor: Employees who were               tial calibration that must be performed as part of the
not involved with the project often came over to           start-up process. The calibration is then per-
chat and check out what was going on and how the           formed periodically when the system becomes con-
equipment worked. More permanent problems                  fused or the error rate increases past a certain
were the employees’ difficulties in tracing the wires      threshold. Users seemed to have difficulty keeping
across complex subassemblies and their hesi-               their heads still enough for the sensitive calibra-
tance to wear the headsets because of fear of the          tion process, so a headrest had to be built. Another
lasers located close to their eyes and dislike of the      problem was that the magnetic tracking devices did
“helmet head” effect that came from wearing the            not work well because there were so many metal parts
equipment.                                                 in the assembly. In addition, the speech recogni-
     One of the strongest success points for this pilot    tion part of the system turned out to be too sensi-
study was that the bundles created using the aug-          tive to background noise, so it was turned off.
mented-reality system met Boeing’s quality assur-               The pilot study for this project was used as a
ance standards. Another good thing was that the            demonstration at a trade show in Germany in 1998.
general background noise level of the factory did not      The program ran for one week without difficulty.
interfere with the acoustic tracker. In the pilot study,   Due to time considerations, the system was not cal-
augmented reality offered no improvement in pro-           ibrated for each user, so some people were not as
ductivity and the only cost savings came from no           impressed as the developers had hoped. Also,
longer needing to store the various assembly boards.       even with the headrest, some users never stayed still
(This can be, however, a significant savings.) The de-      long enough for a proper calibration to be per-
velopers concluded that the reason there was no sig-       formed. Their reactions showed researchers that av-
nificant improvement in assembly time was because           erage users require some degree of training if
they still had some difficulty using the system’s in-       they are to use this sort of equipment success-
terface to find specific wires. The developers are work-     fully. Despite setbacks, the developers considered
ing on a new interface that should help to solve           the pilot a success because it brought the technol-
this problem.                                              ogy to a new group of potential users and it gener-
     Augmented reality has also been used in BMW           ated several possible follow-up ideas relating to the
automobile manufacture. The application was de-            door lock assembly.
signed to demonstrate the assembly of a door lock
for a car door, and the system was used as a feasi-
bility study. The annotations and graphics were taken      The Future
from a CAD (computer-aided design) system that             Augmented reality promises to help humans in
was used to construct the actual physical parts for        many of their tasks by displaying the right infor-
the lock and the door. In this case, the augmented-        mation at the right time and place. There are many
reality system uses a see-through HMD and a voice-         technical challenges to be overcome before such
activated computer—in part because the assembly            interfaces are widely deployed, but driven by com-
process requires that the user have both hands free        pelling potential applications in surgery, the mili-
for the assembly process. Because this augmented-          tary, manufacturing, and entertainment, progress
reality system mimicked an existing virtual-reality        continues to be made in this promising form of
version of assembly planning for the door lock as-         human-computer interaction.
sembly, much of the required data was already avail-
able in an easily retrievable format, which simplified                     Rajeev Sharma and Kuntal Sengupta
the development of the augmented-reality system.
     The developers had to overcome certain prob-          See also Augmented Cognition; Virtual Reality
lems with the system in order to make the pilot work.
The first was the issue of calibration. There is an ini-
64 ❘❙❚ BERKSHIRE ENCYCLOPEDIA OF HUMAN-COMPUTER INTERACTION




FURTHER READING                                                                              AVATARS
Aliaga, D. G. (1997). Virtual objects in the real world. Communi-         Avatar derives from the Sanskrit word avatarah,
   cations of the ACM, 40(3), 49–54.
Azuma, R. (1997). A survey of augmented reality. Presence:
                                                                          meaning “descent” and refers to the incarnation—
   Teleoperators and Virtual Environments, 6(3), 355–385.                 the descent into this world—of a Hindu god. A Hindu
Bajura, M., Fuchs, H., & Ohbuchi, R. (1992). Merging virtual objects      deity embodied its spiritual being when interacting
   with the real world: Seeing Ultrasound imagery within the              with humans by appearing in either human or ani-
   patient. Computer Graphics (Proceedings of SIGGRAPH’92),
   26(2), 203–210.                                                        mal form. In the late twentieth century, the term
Das, H. (Ed.). (1994). Proceedings of the SPIE—The International          avatar was adopted as a label for digital represen-
   Society for Optical Engineering. Bellingham, WA: International         tations of humans in online or virtual environments.
   Society for Optical Engineering.
Elvins, T. T. (1998, February). Augmented reality: “The future’s so
                                                                          Although many credit Neal Stephenson with being
   bright I gotta wear (see-through) shades.” Computer Graphics,          the first to use avatar in this new sense in his semi-
   32(1), 11–13.                                                          nal science fiction novel Snow Crash (1992), the term
Ikeuchi, K., Sato, T., Nishino, K., & Sato, I. (1999). Appearance mod-    and concept actually appeared as early as 1984 in on-
   eling for mixed reality: photometric aspects. In Proceedings of the
   1999 IEEE International Conference on Systems, Man, and
                                                                          line multiuser dungeons, or MUDs (role-playing en-
   Cybernetics (SMC’99) (pp. 36–41). Piscataway, NJ: IEEE.                vironments), and the concept, though not the term,
Milgram, P., & Kishino, F. (1994, December). A taxonomy of mixed          appeared in works of fiction dating back to the mid-
   reality visual displays. IEICE Transactions on Information             1970s. This entry explores concepts, research, and
   Systems, E77-D(12), 1321–1329.
Neumann, U., & Majoros, A. (1998). Cognitive, performance, and
   systems issues for augmented reality applications in manufacturing
   and maintenance. In Proceedings of the IEEE 1998 Virtual Reality
   Annual International Symposium (pp. 4–11). Los Alamitos, CA:
   IEEE.                                                                               Embodied               Avatar
Ohshima, T., Sato, K., Yamamoto, H., & Tamura, H. (1998). AR2                           Agent
   hockey: A case study of collaborative augmented reality. In
   Proceedings of the IEEE 1998 Virtual Reality Annual International
   Symposium (pp. 268–275) Los Alamitos, CA: IEEE.
Ong, K. C., Teh, H. C., & Tan, T. S. (1998). Resolving occlusion in im-                            Digital
   age sequence made easy. Visual Computer, 14(4), 153–165.                                    Representation
Raghavan, V., Molineros, J., & Sharma, R. (1999). Interactive evalua-
   tion of assembly sequences using augmented reality. IEEE
   Transactions on Robotics and Automation, 15(3), 435–449.
Rosenblum, L. (2000, January–February). Virtual and augmented re-
   ality 2020. IEEE Computer Graphics and Applications, 20(1), 38–39.
State, A., Chen, D. T., Tector, C., Brandt, A., Chen, H., Ohbuchi, R.,
   et al. (1994). Observing a volume rendered fetus within a pregnant
   patient. In Proceedings of IEEE Visualization 94 (pp. 364–368). Los
                                                                                     Agent                      Live Human
   Alamitos, CA: IEEE.                                                                                              Being
Stauder, J. (1999, June). Augmented reality with automatic illumi-
   nation control incorporating ellipsoidal models. IEEE Transactions
   on Multimedia, 1(2), 136–143.
Tatham, E. W. (1999). Getting the best of both real and virtual worlds.
   Communications of the ACM, 42(9), 96–98.
Tatham, E. W., Banissi, E., Khosrowshahi, F., Sarfraz, M., Tatham, E.,    F I G U R E 1 . A representational schematic of avatars and
   & Ursyn, A. (1999). Optical occlusion and shadows in a “see-           embodied agents. When a given digital representation
   through” augmented reality display. In Proceedings of the 1999 IEEE
   International Conference on Information Visualization (pp. 128–131).   is controlled by a human, it is an avatar, and when it
   Los Alamitos, CA: IEEE.                                                is controlled by a computational algorithm it is an em-
Yamamoto, H. (1999). Case studies of producing mixed reality worlds.      bodied agent. Central to the current definition is the
   In Proceedings of the 1999 IEEE International Conference on Systems,
   Man, and Cybernetics (SMC’99) (pp. 42–47). Piscataway, NJ: IEEE.
                                                                          ability for real-time behavior, in that the digital rep-
                                                                          resentation exhibits behaviors by the agent or hu-
                                                                          man as they are performed.
                                                                                               AVATAARS ❚❙❘ 65



ethical issues related to avatars as digital human rep-    gestures and an agent controls more mundane au-
resentations. (We restrict our discussion to digital       tomatic behaviors.
avatars, excluding physical avatars such as puppets             One should also distinguish avatars from online
and robots. Currently, the majority of digital avatars     identities. Online identities are the distributed dig-
are visual or auditory information though there is         ital representations of a person. Humans are known
no reason to restrict the definition as such.)              to each other via e-mail, chat rooms, homepages, and
                                                           other information on the World Wide Web.
                                                           Consequently, many people have an online identity,
Agents and Avatars                                         constituted by the distributed representation of all
Within the context of human-computer interaction,          relevant information, though they may not have
an avatar is a perceptible digital representation whose    an avatar.
behaviors reflect those executed, typically in real time,
by a specific human being. An embodied agent, by
contrast, is a perceptible digital representation whose    Realism
behaviors reflect a computational algorithm designed        Avatars can resemble their human counterparts along
to accomplish a specific goal or set of goals. Hence,       a number of dimensions, but the two that have re-
humans control avatar behavior, while algorithms           ceived the most attention in the literature are be-
control embodied agent behavior. Both agents and           havioral realism (reflected in the number of a given
avatars exhibit behavior in real time in accordance        human’s behaviors the avatar exhibits) and photo-
with the controlling algorithm or human actions.           graphic realism (reflected in how many of a given hu-
     Figure 1 illustrates the fact that the actual digi-   man’s static visual features the avatar possesses).
tal form the digital representation takes has no bear-         Behavioral realism is governed by the capability
ing on whether it is classified as an agent or avatar:      of the implementation system to track and render
An algorithm or person can drive the same repre-           behavior in real time. Currently, real-time behavioral
sentation. Hence, an avatar can look nonhuman de-          tracking technology, while improving steadily, does
spite being controlled by a human, and an agent can        not meet expectations driven by popular culture; for
look human despite being controlled by an algorithm.       example, online representations of the character Neo
     Not surprisingly, the fuzzy distinction between       in The Matrix (1999), Hiro from Snow Crash (1992),
agents and avatars blurs for various reasons.              or Case from Neuromancer (1984). In those fictional
Complete rendering of all aspects of a human’s ac-         accounts, the movements and gestures of avatars and
tions (down to every muscle movement, sound, and           the represented humans are generally perceptually
scent) is currently technologically unrealistic. Only      indistinguishable. However, in actual practice,
actions that can be tracked practically can be ren-        complete real-time behavior tracking is extremely
dered analogously via an avatar; the remainder are         difficult. Although gesture tracking through vari-
rendered algorithmically (for example, bleeding) or        ous mechanical, optical, and other devices has im-
not at all (minute facial expressions, for instance).      proved, the gap between actual movements and avatar
     In some cases avatar behaviors are under non-         movements remains large, reducing behavioral real-
analog human control; for example, pressing a              ism at least in situations requiring real-time tracking
button and not the act of smiling may be the way           and rendering, such as online social interaction
one produces an avatar smile. In such a case, the be-      (for example, collaborative virtual work groups).
haviors are at least slightly nonanalogous; the                Fewer barriers exist for photographic realism.
smile rendered by the button-triggered computer al-        Three-dimensional scanners and photogrammetric
gorithm may be noticeably different from the actual        software allow for the photographically realistic recre-
human’s smile. Technically, then, a human repre-           ation of static, digital human heads and faces that
sentation can be and often is a hybrid of an avatar        cannot be easily distinguished from photographs and
and an embodied agent, wherein the human con-              videos of the underlying faces. Nonetheless, the
trols the consciously generated verbal and nonverbal       key challenge to avatar designers is creating faces and
66 ❘❙❚ BERKSHIRE ENCYCLOPEDIA OF HUMAN-COMPUTER INTERACTION




Three views of a digital avatar modeled after a human head and face. This avatar is built by creating a three-
dimensional mesh and wrapping a photographic texture around it.               Photo courtesy of James J. Blascovich.



bodies in sufficient detail to allow for the realistic      users interact with one another using either a key-
rendering of behavior, which brings us back to be-          board or a joystick, typing messages back and
havioral realism. In summary, static avatars currently      forth and viewing one another’s avatars as they move
can look quite a bit like their human controllers but       around the digital world. Typically, these are avatars
can only perform a small subset of a dynamic hu-            in the minimal sense of the word; behavioral and
man’s actions in real time.                                 photographic realism is usually quite low. In the case
                                                            of online role-playing games, users typically navi-
                                                            gate the online world using “stock” avatars with lim-
Current Use of Avatars                                      ited behavioral capabilities.
Depending on how loosely one defines digital rep-
resentation, the argument can be made that
avatars are quite pervasive in society. For example,        Avatar Research
sound is transformed into digital information as it         Computer scientists and others have directed much
travels over fiber-optic cables and cellular networks.       effort towards developing systems capable of pro-
Consequently, the audio representation we perceive          ducing functional and effective avatars. They have
over phone lines is actually an avatar of the speaker.      striven to develop graphics, logic, and the tracking
This example may seem trivial at first, but be-             capabilities to render actual movements by humans
comes less trivial when preset algorithms are applied       on digital avatars with accuracy, and to augment
to the audio stream to cause subtle changes in the          those movements by employing control algo-
avatar, for example, to clean and amplify the sig-          rithms that supply missing tracking data or infor-
nal. This can only be done effectively because the          mation about static visual features.
voice is translated into digital information.                    Furthermore, behavioral scientists are examin-
    More often, however, when people refer to               ing how humans interact with one another via
avatars, they are referring to visual representations.      avatars. These researchers strive to understand so-
Currently, millions of people employ avatars in             cial presence, or copresence, a term referring to the
online role-playing games as well as in chat rooms          degree to which individuals respond socially towards
used for virtual conferencing. In these environments,       others during interaction among their avatars, com-
                                                                                                   AVATARS ❚❙❘ 67



pared with the degree to which they respond to ac-           to have systems operators do this for them) within
tual humans.                                                 virtual environments by amplifying or suppressing
     The behavioral scientist Jim Blascovich and his         communication signals.
colleagues have created a theoretical model for so-               TSI algorithms can impact interactants’ abilities
cial influence within immersive virtual environments          to influence interaction partners. For example, sys-
that provides specific predictions for how the inter-         tem operators can tailor the nonverbal behaviors
play of avatars’ photographic and behavioral realism         of online teachers lecturing to more than one stu-
will affect people’s sense of the relevance of the avatar-   dent simultaneously within an immersive virtual
mediated encounter. They suggest that the inclusion          classroom in ways specific to each student inde-
of certain visual features is necessary if the avatar        pendently and simultaneously. Student A might
is to perform important, socially relevant behavioral        respond well to a teacher who smiles, and Student B
actions. For example, an avatar needs to have rec-           might respond well to a teacher with a neutral ex-
ognizable eyebrows in order to lower them in a frown.        pression. Via an avatar that is rendered separately for
     Other data emphasize the importance of behav-           each student, the teacher can be represented si-
ioral realism. In 2001 Jeremy Bailenson and his              multaneously by different avatars to different stu-
colleagues demonstrated that making a digital rep-           dents, thereby communicating with each student in
resentation more photographically realistic does not         the way that is optimal for that student. The psy-
increase its social presence in comparison with an           chologist Andrew Beall and his colleagues have used
agent that is more cartoon-like as long as both types        avatars to employ such a strategy using eye con-
of agents demonstrate realistic gaze behaviors. In           tact; they demonstrated that students paid greater
findings presented in 2003, Maia Garau and her col-           attention to the teacher using TSI.
leagues failed to demonstrate an overall advantage                However, there are ethical problems associated
for more photographically realistic avatars; more-           with TSIs. One can imagine a dismal picture of the fu-
over, these researchers demonstrated that increasing         ture of online interaction, one in which nobody is who
the photographic realism of an avatar can actually           they seem to be and avatars are distorted so much from
cause a decrease in social presence if behavioral re-        the humans they represent that the basis for judging
alism is not also increased.                                 the honesty of the communication underlying social
     In sum, though research on avatars currently is         interactions is lost. Early research has demonstrated
largely in its infancy, investigators are furthering our     that TSIs involving avatars are often difficult to detect.
understanding of computer-mediated human in-                 It is the challenge to researchers to determine the
teraction. As avatars become more commonplace,               best way to manage this issue as the use of avatars
research geared towards understanding these appli-           becomes more prevalent.
cations should increase.

                                                             State of the Art
Ethical Issues                                               Currently, there are many examples of humans in-
Interacting via avatars allows for deceptive interac-        teracting with one another via avatars. For the most
tions. In 2003 Bailenson and his colleagues intro-           part, these avatars are simplistic and behaviorally
duced the notion of transformed social interactions          and photographically unrealistic. The exception oc-
(TSIs). Using an avatar to interact with another per-        curs in research laboratories, in which scientists
son is qualitatively different from other forms of           are beginning to develop and test avatars that are
communication, including face-to-face interaction,           similar in appearance and behavior to their human
standard telephone conversations, and videocon-              counterpart. As avatars become more ubiquitous,
ferencing. An avatar that is constantly rerendered in        it is possible that we may see qualitative changes in
real time makes it possible for interactants to sys-         social interaction due to the decoupling and trans-
tematically filter their appearance and behaviors (or         formation of behavior from human to avatar. While
68 ❘❙❚ BERKSHIRE ENCYCLOPEDIA OF HUMAN-COMPUTER INTERACTION



there are ethical dangers in transforming behav-                                Blascovich, J. (2001). Social influences within immersive virtual
iors as they pass from physical actions to digital rep-                             environments. In R. Schroeder (Ed.), The social life of avatars. Berlin,
                                                                                    Germany: Springer-Verlag.
resentations, there are also positive opportunities                             Blascovich, J., Loomis, J., Beall, A. C., Swinth, K. R., Hoyt, C. L., &
both for users of online systems and for research-                                  Bailenson, J. N. (2001). Immersive virtual environment technol-
ers in human-computer interaction.                                                  ogy as a methodological tool for social psychology. Psychological
                                                                                    Inquiry, 13, 146–149.
                                                                                Brunner, J. (1975). Shockwaver rider. New York: Ballantine Books.
            Jeremy N. Bailenson and James J. Blascovich                         Cassell, J., & Vilhjálmsson, H. (1999). Fully embodied conversa-
                                                                                    tional avatars: Making communicative behaviors autonomous.
See also Animation; Telepresence; Virtual Reality                                   Autonomous Agents and Multi-Agent Systems, 2(1), 45–64.
                                                                                Garau, M., Slater, M.,Vinayagamoorhty,V., Brogni, A., Steed, A., & Sasse,
                                                                                    M. A. (2003). The impact of avatar realism and eye gaze control on
                                                                                    perceived quality of communication in a shared immersive virtual
                                                                                    environment. In Proceedings of the SIGCHI Conference on Human
                                                                                    Factors in Computing Systems (pp. 529–536). New York: ACM Press.
FURTHER READING                                                                 Gibson, W. (1984). Neuromancer. New York: Ace Books.
                                                                                Morningstar, C., & Farmer, F.R. (1991). The lessons of Lucasfilm’s
Badler, N., Phillips, C., & Webber, B. (1993). Simulating humans:                   habitat. In M. Benedikt (Ed.), Cyberspace: First steps. Cambridge,
   Computer graphics, animation, and control. Oxford, UK: Oxford                    MA: MIT Press.
   University Press.                                                            Slater, M., Howell, J., Steed, A., Pertaub, D., Garau, M., & Springel,
Bailenson, J. N., Beall, A. C., Blascovich, J., & Rex, C. (in press).               S. (2000). Acting in virtual reality. ACM Collaborative Virtual
   Examining virtual busts: Are photogrammetrically generated head                  Environments, CVE’2000, 103–110.
   models effective for person identification? PRESENCE: Teleoperators           Slater, M., Sadagic, A., Usoh, M., & Schroeder, R. (2000). Small group
   and Virtual Environments.                                                        behaviour in a virtual and real environment: A comparative study.
Bailenson, J. N., Beall, A. C., Loomis, J., Blascovich, J., & Turk, M. (in          PRESENCE: Teleoperators and Virtual Environments, 9, 37–51.
   press). Transformed social interaction: Decoupling representation            Stephenson, N. (1993). Snow crash. New York: Bantam Books.
   from behavior and form in collaborative virtual environments.                Thalmann, M. N, & Thalmann D. (Eds). (1999). Computer Animation
   PRESENCE: Teleoperators and Virtual Environments.                                and Simulation 99. Vienna, Austria: Springer-Verlag.
Bailenson, J. N., Blascovich, J., Beall, A. C., & Loomis, J. M. (2001).         Turk, M., & Kolsch, M. (in press). Perceptual Interfaces. In G. Medioni
   Equilibrium revisited: Mutual gaze and personal space in virtual                 & S. B. Kang (Eds.), Emerging topics in computer vision. Upper
   environments. PRESENCE: Teleoperators and Virtual Environments,                  Saddle River, NJ: Prentice-Hall.
   10, 583–598.                                                                 Turkle, S. (1995). Life on the screen: Identity in the age of the Internet.
Beall, A. C., Bailenson, J. N., Loomis, J., Blascovich, J., & Rex, C. (2003).       New York: Simon & Schuster.
   Non-zero-sum mutual gaze in immersive virtual environments.                  Yee, N. (2002). Befriending ogres and wood elves: Understanding rela-
   In Proceedings of HCI International 2003 (pp. 1108–1112). New                    tionship formation in MMORPGs. Retrieved January 16, 2004, from
   York: ACM Press.                                                                 http://www.nickyee.com/hub/relationships/home.html
                                                                                          BETA TESTING

                                                                                              BRAILLE

                                                                            BRAIN-COMPUTER INTERFACES

                                                                                            BROWSERS




            B

                                                          will be discussed below, beta tests make it possible for
            BETA TESTING                                  endusers to contribute to the design and develop-
                                                          ment of a product and may represent a shift in the
Beta testing, a stage in the design and development       organization of the production process.
process of computer software and hardware, uses
people outside a company, called “beta testers,” to be
sure that products function properly for typical end-     Definitions of Beta Testing
users outside the firm. Does a piece of software work      A beta test is an early (preshipping or prelaunch), un-
under normal operating conditions? Can users nav-         official release of hardware or software that has already
igate important features? Are there any critical          been tested within the company for major flaws. In
programming flaws? These are the questions beta           theory, beta versions are very close to the final prod-
tests answer.                                             uct, but in practice beta testing is often simply one way
    The widespread use of beta tests warrants the ex-     for a firm to get users to try new software under real
amination of the process. Because the trade literature    conditions. Beta tests expose software and hardware
in computer programming focuses on the mechan-            to real-world configurations of computing platforms,
ics of designing, conducting, and interpreting beta       operating systems, hardware, and users. For exam-
tests, less has been written on the social implications   ple, a beta test of a website is “the time period just
of the growing use of beta testing. For example, as       before a site’s official launch when a fully operational

                                                                                                                69
70 ❘❙❚ BERKSHIRE ENCYCLOPEDIA OF HUMAN-COMPUTER INTERACTION



product is used under normal operating conditions to       ogists Gina Neff and David Stark, establishing a
identify any programming bugs or interface issues”         cycle of testing, feedback, and innovation that facil-
(Grossnickle and Raskin 2001, 351). David Hilbert de-      itates negotiations about what is made can make it
scribes beta testing as a popular technique for evalu-     possible to incorporate broader participation into
ating the fit between application design and use.           the design of products and organizations.
     The term beta testing emerged from the practice           However, in practice, beta tests may be poorly
of testing the unit, module, or components of a sys-       designed to incorporate user feedback. Advice in the
tem first. This test was called alpha, whereas beta re-     trade literature suggests that beta tests may not be
ferred to the initial test of the complete system. Alpha   constructed to provide more than “bug squashing
and beta, derived from earlier nomenclature of hard-       and usability testing” (Grossnickle and Raskin
ware testing, were reportedly first used in the            n.d., 1). Beta tests also present firms with a chance
1960s at IBM. Now alpha typically refers to tests con-     to conduct research on their users and on how their
ducted within the firm and beta refers to tests con-        products are used. Ideally, beta testers are statistically
ducted externally.                                         representative of typical product users. However, em-
     There is ample evidence that beta testing has in-     pirical research suggests that beta testers may not ac-
creased in various forms over the last decade. James       curately reflect end-users because testers tend to have
Daly, a technology business reporter and founder of        more technical training and hold more technical jobs
the magazine Business 2.0, reports that by 1994, 50        than typical office workers.
percent of Fortune 1000 companies in the United States
had participated in beta testing and 20 percent of those
companies had used beta testing widely. However, the       Critical Views of Beta Testing
implementation—and the purposes—of beta testing            The shift from total quality management to a test-
vary by company. An online market-research hand-           ing-driven model of development means that “the
book suggests that “for most ventures, standard            generation and detection of error plays a renewed
beta-testing technique involves e-mailing friends, fam-    and desired role” in the production cycle (Cole 2002,
ily, and colleagues with the URL of a new site”            1052). With the rise of the acceptance of beta ver-
(Grossnickle and Raskin 2001, 351), which clearly          sions, companies and users alike may be more will-
would not produce a statistically representative sam-      ing to tolerate flaws in widely circulated products,
ple of end users. A meta study of beta-test evaluations    and end-users (including beta testers) may bear an
done more than a decade ago found that most beta           increased burden for the number of errors that com-
testing was actually “driven by convenience or tradi-      panies allow in these products. Some criticism has
tion rather than recognition of the costs and benefits      emerged that companies are “releasing products for
involved” (Dolan and Matthews 1993, 318).                  beta testing that are clearly not ready for the mar-
     In addition to determining whether or not a           ket” and are exploiting free labor by “using beta testers
product works, a beta test can be used to increase a       as unpaid consultants to find the bugs in their prod-
firm’s knowledge about the user base for its prod-         ucts” (Garman 1996, 6).
ucts, to support its marketing and sales goals, and to          Users may also be frustrated by the continually
improve product support. More importantly, beta            updated products that beta testing can enable. The
testers’ suggestions may be incorporated into the de-      distribution of software in non-shrink-wrapped ver-
sign of the product or used to develop subsequent          sions means that products are not clean end-versions
generations of the product.                                b u t d e s t a b i l i ze d a n d con s t a n t l y ch a n g i n g .
                                                           Technological advances in distribution, such as
User Participation in                                      online distribution of software products, “makes it
                                                           possible to distribute products that are continually
Product Development                                        updateable and almost infinitely customizable—
Beta testing allows users to become involved in the        products that, in effect, never leave a type of beta
product-development process. According to sociol-          phase” (Neff and Stark 2003, 177).
                                                                                                       BETA TESTING ❚❙❘ 71



Benefits to Beta Testers                                    tion of the product” (Dolan and Matthews 1993, 20),
Because they are willing to risk bugs that could po-       beta tests present crucial opportunities to incorpo-
tentially crash their computers, beta testers accrue       rate user suggestions into the design of a product.
benefits such as getting a chance to look at new
features and products before other users and con-                                                                      Gina Neff
tributing to a product by detecting software bugs or
minor flaws in programming. More than 2 million             See also Prototyping; User-Centered Design
people volunteered to be one of the twenty thousand
beta testers for a new version of Napster. There is also
an increase of beta retail products—early and of-          FURTHER READING
ten cheaper versions of software that are more ad-
                                                           Cole, R. E. (2002). From continuous improvement to continuous
vanced than a traditional beta version but not yet a          innovation. Total Quality Management, 13(8), 1051–1056.
fully viable commercial release. Although Apple’s          Daly, J. (1994, December). For beta or worse. Forbes ASAP, 36–40.
public beta release of OS X, its first completely          Dolan, R. J., & Matthews, J. M. (1993). Maximizing the utility of con-
new operating system since 1984, cost $29.95, thou-           sumer product testing: Beta test design and management. Journal
                                                              of Product Innovation Management, 10, 318–330.
sands downloaded it despite reports that it still          Hove, D. (Ed.). The Free online dictionary of computing. Retrieved
had many bugs and little compatible software was              March 10, 2004, from http://www.foldoc.org
available. These beta users saw the long-awaited new       Garman, N. (1996). Caught in the middle: Online professionals and
operating system six months before its first com-             beta testing. Online, 20(1), 6.
                                                           Garud, R., Sanjay, J., & Phelps, C. (n.d.). Unpacking Internet time in-
mercial release, and Apple fans and the press pro-            novation. Unpublished manuscript, New York University, New
vided invaluable buzz about OS X as they tested it.           York.
    Many scholars suggest that the Internet has com-       Grossnickle, J., & Raskin, O. (2001). Handbook of online marketing
                                                              research. New York: McGraw Hill.
pressed the product-development cycles, especially         Grossnickle, J., & Raskin, O. (n.d.). Supercharged beta test. Webmonkey:
in software, often to the extent that one generation          Design. Retrieved January 8, 2004, from http://hotwired.lycos.com/
of product software is hard to distinguish from the           webmonkey
next. Netscape, for example, released thirty-nine dis-     Hilbert, D. M. (1999). Large-scale collection of application usage data
                                                              and user feedback to inform interactive software development.
tinct versions between the beta stage of Navigator            Unpublished doctoral dissertation, University of California, Irvine.
1.0 and the release of Communicator 4.0.                   Kogut, B., & Metiu, A. (2001). Open source software development and
                                                              distributed innovation. Oxford Review of Economic Policy, 17(2),
                                                              248–264.
                                                           Krull, R. (2000). Is more beta better? Proceedings of the IEEE Professional
Future Developments                                           Communication Society, 301–308.
Production is an “increasingly dense and differenti-       Metiu, A., & Kogut, B. (2001). Distributed knowledege and the global
ated layering of people, activities and things, each          organization of software development. Unpublished manu-
                                                              script, Wharton School of Business, University of Pennsylvania,
operating within a limited sphere of knowing and              Philadelphia.
acting that includes variously crude or sophisticated      Neff, G., & Stark, D. (2003). Permanently beta: Responsive organi-
conceptualizations of the other” (Suchman 2003,               zation in the Internet era. In P. Howard and S. Jones (Eds.), Society
62). Given this complexity, beta testing has been wel-        Online. Thousand Oaks, CA: Sage.
                                                           O'Mahony, S. (2002). The Emergence of a new commercial actor: Com-
comed as a way in which people who create prod-               munity managed software projects. Unpublished doctoral disser-
ucts can inter act w ith those who use them.                  tation, Stanford University, Palo Alto, CA. Retrieved on January 8,
Internet communication facilitates this communi-              2004, from http://opensource.mit.edu/
cation, making the distribution of products in ear-        Raymond, E. (1999). The Cathedral and the bazaar: Musings on Linux
                                                              and open source from an accidental revolutionary. Sebastapol, CA:
lier stages of the product cycle both easier and              O'Reilly and Associates.
cheaper; it also facilitates the incorporation of user     Ross, R. (2002). Born-again Napster takes baby steps. Toronto Star,
feedback into the design process.                             E04.
                                                           Suchman, L. (2002). Located accountabilities in technology production.
    While it is true that “most design-change ideas           Retrieved on January 8, 2004, from http://www.comp.lancs.ac.
surfaced by a beta test are passed onto product de-           uk/sociology/soc039ls.html. Centre for Science Studies, Lancaster
velopment for incorporation into the next genera-             University.
72 ❘❙❚ BERKSHIRE ENCYCLOPEDIA OF HUMAN-COMPUTER INTERACTION



Techweb (n.d.). Beta testing. Retrieved on January 8, 2004, from            and too expensive to give its users unlimited and
   http://www.techweb.com/encyclopedia                                      quick access to an increasing amount of printed ma-
Terranova, T. (2000). Free labor: Producing culture for the digital econ-
   omy. Social Text 18(2), 33–58.
                                                                            terial: books, newspapers, leaflets, and so forth.
                                                                            The invention of the transistor in 1947 by three U.S.
                                                                            physicists and of integrated circuits in the late 1960s
                                                                            provided the solution: electromechanical tactile dis-
                        BRAILLE                                             plays. After many attempts, documented by numer-
                                                                            ous patents, electronic Braille was developed
Access to printed information was denied to blind                           simultaneously during the early 1970s by Klaus-Peter
people until the late 1700s, when Valentin Haüy, hav-                       Schönherr in Germany and Oleg Tretiakoff in France.
ing funded an institution for blind children in Paris,
embossed letters in relief on paper so that his pupils
could read them. Thus, two hundred and fifty                                First Electronic Braille Devices
years after the invention of the printing press by the                      In electronic Braille, Braille codes—and therefore
German inventor Johannes Gutenberg, blind people                            Braille books—are stored in numerical binary for-
were able to read but not to write.                                         mat on standard mass storage media: magnetic tapes,
                                                                            magnetic disks, and so forth. In this format the bulk
                                                                            and cost of Braille books are reduced by several orders
Historical Background                                                       of magnitude. To be accessible to blind users, elec-
In 1819 a French army officer, Charles Barbier, in-                          tronically stored Braille codes must be converted into
vented a tactile reading system, using twelve-dot                           raised-dot patterns by a device called an “electro-
codes embossed on paper, intended for nighttime                             mechanical Braille display.” An electromechanical
military communications. Louis Braille, who had                             Braille display is a flat reading surface that has holes
just entered the school for the blind in Paris, learned                     arranged in a Braille cell pattern. The hemispherical
of the invention and five years later, at age fifteen,                      tip of a cylindrical pin can either be raised above the
developed a much easier-to-read six-dot code,                               reading surface to show a Braille dot or lowered under
providing sixty-three dot patterns. Thanks to his in-                       the reading surface to hide the corresponding Braille
vention, blind people could not only read much                              dot. The Braille dot vertical motion must be con-
faster, but also write by using the slate, a simple hand                    trolled by some kind of electromechanical actuator.
tool made of two metal plates hinged together be-                           Two such displays were almost simultaneously put
tween which a sheet of paper could be inserted                              onto the market during the mid-1970s.
and embossed through cell-size windows cut in                                   The Schönherr Braille calculator had eight Braille
the front plate. Six pits were cut in the bottom                            cells of six dots each, driven by electromagnetic ac-
plate to guide a hand-held embossing stylus inside                          tuators and a typical calculator keyboard. The dot
each window.                                                                spacing had to be increased to about 3 millimeters
    In spite of its immediate acceptance by his fel-                        instead of the standard 2.5 millimeters to provide
low students, Braille’s idea was officially accepted                        enough space for the actuators.
only thirty years later, two years after his death in                           The Tretiakoff Braille notebook carried twelve
1852. Eighty more years passed before English-speak-                        Braille standard cells of six dots each, driven by piezo-
ing countries adapted the Braille system in 1932, and                       electric (relating to electricity or electric polarity due
more than thirty years passed before development                            to pressure, especially in a crystalline substance) reeds,
of the Nemeth code, a Braille system of scientific no-                       a keyboard especially designed for blind users, a cas-
tation, in 1965. Braille notation was also adopted by                       sette tape digital recorder for Braille codes storage,
an increasing number of countries.                                          and a communication port to transfer data between
    In spite of its immense benefits for blind people,                       the Braille notebook and other electronic devices. Both
the Braille system embossed on paper was too bulky                          devices were portable and operated on replaceable or
                                                                                                                          BRAILLE ❚❙❘ 73




                  Enhancing Access to Braille Instructional Materials

(ANS)—Most blind and visually impaired children at-               have misunderstood each other's business, he said, which
tend regular school classes these days, but they are often        led to frustration on both sides.
left waiting for Braille and large-print versions of class            Most blind children are mainstreamed into public
texts to arrive while the other children already have the         school classrooms and receive additional help from a cadre
books.                                                            of special teachers of the blind. Technology is also giv-
     There are 93,000 students in kindergarten through            ing blind students more options. Scanning devices now
12th grade who are blind or have limited vision. Because          download texts into Braille and read text aloud. Closed
this group represents a small minority of all school-             circuit television systems can enlarge materials for low-
children, little attention has been paid to updating the          vision students.
cumbersome process of translating books into Braille, ad-             “These kids have very individual problems,” noted
vocates said.                                                     Kris Kiley, the mother of a 15-year-old who has limited
     Traditionally, publishers have given electronic copies       vision.“It's not one size fits all. But if you don't teach them
of their books to transcribers, who often need to com-            to read you've lost part of their potential.”
pletely reformat them for Braille. Lack of a single techno-           New tools also bring with them new problems. For
logical standard and little communication between                 example, the new multimedia texts, which are available
publishing houses and transcribers led to delays in blind         to students on CD-ROM, are completely inaccessible to
students receiving instructional materials, experts said.         blind students. And because graphics now dominate many
     The solution, said Mary Ann Siller, a national pro-          books, lots of information, especially in math, does not
gram associate for the American Foundation for the Blind          reach those with limited vision.
who heads its Textbook and Instructional Materials                    Simply recognizing the challenges faced by the
Solutions Forum, is to create a single electronic file for-        blind would go a long way toward solving the problem,
mat and a national repository for textbooks that would            said Cara Yates. Yates, who recently graduated from law
simplify and shorten the production process. And that's           school, lost her sight at age 5 to eye cancer. She recalls one
exactly what is happening.                                        of her college professors who organized a series of tutors
     In October, the American Printing House for the Blind        to help her “see” star charts when she took astrophysics.
in Louisville, Ky., took the first step in creating a reposi-          “A lot of it isn't that hard,” she said.“It just takes some
tory by listing 140,000 of its own titles on the Internet.        thought and prior planning. The biggest problem for the
The group is now working to get publishers to deposit             blind is they can't get enough information. There's no ex-
their text files, which transcribers could readily access.         cuse for it. It's all available.”
     “Everyone is excited about it,” said Christine Anderson,         Siller said the foundation also hoped to raise aware-
director of resource services for the Kentucky organiza-          ness about educational assessment; the importance of
tion. By having a central database with information about         parental participation; better preparation for teachers; a
the files for all books available in Braille, large print, sound   core curriculum for blind students in addition to the
recording or computer files, costly duplications can be           sighted curriculum; and better Braille skills and a reduced
eliminated, she said.                                             caseload for teachers who often travel long distances to
     Pearce McNulty, director of publishing technology at         assist their students.
Houghton Mifflin Co. in Boston, which is a partner in the                                                      Mieke H. Bomann
campaign, said he is hopeful the repository will help solve          Source: Campaign seeks to end blind students' wait for Braille
                                                                     textbooks. American News Service, December 16, 1999.
the problem. Publishers and Braille producers historically
74 ❘❙❚ BERKSHIRE ENCYCLOPEDIA OF HUMAN-COMPUTER INTERACTION



rechargeable batteries. The Tretiakoff Braille note-         cells used in Tretiakoff 's extremely portable Braille
book, called “Digicassette,” measured about 20 by            notebook, the P-Touch. In these “vertical” cells each
25 by 5 centimeters.                                         piezoelectric actuator was located underneath the
    A read-only version of the Digicassette was man-         corresponding tactile dot, allowing tactile dots to be
ufactured for the U.S. National Library Services for         arranged in arrays of regularly spaced rows and
the Blind of the Library of Congress.                        columns for the electronic display of graphics. These
                                                             vertical cells were about twice as high as conventional
                                                             “horizontal” cells and no less expensive. Multiline or
Personal Braille Printers                                    graphic displays were thus made technically feasible
Braille books consist of strong paper pages embossed         but remained practically unaffordable at about $12
with a Braille dot pattern by high-speed machines            per dot for the end user as early as 1985.
and then bound together much like ordinary books.
A typical Braille page can carry up to twenty-five lines
of forty Braille characters each and can be explored         Active versus Passive Reading
rapidly from left to right and from top to bottom by         Since Louis Braille, blind people have performed tac-
a blind reader. Electronic Braille displays consist gen-     tile reading by moving the tip of one to three fingers
erally of a single line comprising usually from eight-       across a Braille page or along a Braille line while ap-
een to forty Braille characters to keep the displays         plying a small vertical pressure on the dot pattern in
portable and affordable for individual users. The shift      a direction and at a speed fully controlled by the
from a full page to a single line delayed the accept-        reader, hence the name “active reading.”
ance of Braille displays in spite of their ability to pro-        Louis Braille used his judgment to choose tactile
vide easy and high-speed access to electronically            dot height and spacing; research performed during
stored information.                                          the last thirty years has shown that his choices
     Personal Braille printers, also made possible by        were right on the mark. Objective experiments, in
the development of integrated circuits, appeared             which the electrical response of finger mechanore-
soon after the first personal computers to fill the gap        ceptors (neural end organs that respond to a me-
between industrially produced Braille books and sin-         chanical stimulus, such as a change in pressure) is
gle-line Braille displays. Similar in concept to dot-        measured from an afferent (conveying impulses
matrix ink printers, personal Braille printers allowed       toward the central nervous system) nerve fiber, have
a blind user to emboss on a sheet of strong paper a          shown that “stroking”—the horizontal motion of
few lines of Braille characters per minute from Braille      the finger—plays an essential role in touch resolu-
codes received from an external source.                      tion, the ability to recognize closely spaced dots.
                                                                  Conversely, if a blind reader keeps the tip of one
                                                             or more fingers still on an array of tactile dots that
Tactile Graphics                                             is moved in various patterns up or down under the
Although the first personal Braille printers were de-         fingertips, this is called “passive reading.” Passive read-
signed to print only a regularly spaced Braille pat-         ing has been suggested as a way to reduce the num-
tern—at .6 centimeter spacing between characters             ber of dots, and therefore the cost of tactile displays,
—some were outfitted with print heads capable of              by simulating the motion of a finger across a wide ar-
printing regularly spaced dots, in both the hori-            ray of dots by proper control of vertical dot motion
zontal and the vertical directions, allowing the             under a still finger. The best-known example of this
production of embossed tactile graphics.                     approach is the Optacon (Optical to Tactile Converter),
    Although the first electronic Braille displays were       invented during the mid-1970s by John Linvill to give
built with horizontally stacked piezoelectric reeds,         blind people immediate and direct access to printed
whose length—about 5 centimeters—prevented the               material. The Optacon generated a vibrating tactile
juxtaposition of more than two Braille lines, the mid-       image of a small area of an object viewed by its
1980s brought the first “vertical” piezoelectric Braille      camera placed and moved against its surface.
                                                                                    BRAIN-COMPUTER INTERFACES ❚❙❘ 75



    Research has shown that touch resolution and
reading speed are significantly impaired by passive                         BRAIN-COMPUTER
reading, both for raised ordinary character shapes
and for raised-dot patterns.                                                 INTERFACES
                                                                     A brain-computer interface (BCI), also known as a
Current and Future                                                   direct brain interface (DBI) or a brain-machine in-
                                                                     terface (BMI), is a system that provides a means
Electronic Tactile Displays                                          for people to control computers and other devices
At the beginning of the twenty-first century, sev-                   directly with brain signals. BCIs fall into the cate-
eral companies make electromechanical tactile cells,                 gory of biometric devices, which are devices that de-
which convert electrical energy into mechanical en-                  tect and measure biological properties as their basis
ergy and vice versa, but the dominant actuator tech-                 of operation. Research on brain-computer interfaces
nology is still the piezoelectric (relating to electricity           spans many disciplines, including computer science,
or electric polarity due to pressure, especially in a                neuroscience, psychology, and engineering. BCIs
crystalline substance) bimorph reed, which keeps                     were originally conceived in the 1960s, and since the
the price per tactile dot high and the displays bulky                late 1970s have been studied as a means of provid-
and heavy. The majority of electronic tactile dis-                   ing a communication channel for people with very
plays are single-line, stand-alone displays carry-                   severe physical disabilities. While assistive technol-
ing up to eighty characters or Braille computers                     ogy is still the major impetus for BCI research, there
carrying from eighteen to forty characters on a sin-                 is considerable interest in mainstream applications
gle line. Their costs range from $3,000 to more than                 as well, to provide a hands-free control channel that
$10,000. A small number of graphic tactile mod-                      does not rely on muscle movement.
ules carrying up to sixteen by sixteen tactile dots                      Despite characterizations in popular fiction, BCI
are also available from manufacturers such as KGS                    systems are not able to directly interpret thoughts or
in Japan.                                                            perform mind reading. Instead, BCI systems mon-
     Several research-and-development projects, us-                  itor and measure specific aspects of a user’s brain
ing new actuator technologies and designs, are un-                   signals, looking for small but detectable differences
der way to develop low-cost g raphic tactile                         that signal the intent of the user. Most existing BCI
displays that could replace or complement visual dis-                systems depend on a person learning to control an
plays in highly portable electronic communication                    aspect of brain signals that can be detected and meas-
devices and computers.                                               ured. Other BCI systems perform control tasks, such
                                                                     as selecting letters from an alphabet, by detecting
                                               Oleg Tretiakoff       brain-signal reactions to external stimuli.
                                                                         Although BCIs can provide a communications
See also Sonification; Universal Access                               channel, the information transmission rate is low
                                                                     compared with other methods of control, such as
                                                                     keyboard or mouse. The best reported user per-
FURTHER READING                                                      formance with current BCI systems is an informa-
                                                                     tion transfer rate of sixty-eight bits per minute, which
American Council of the Blind. (2001). Braille: History and Use of   roughly translates to selecting eight characters per
   Braille. Retrieved May 10, 2004, from http://www.acb.org/re-
   sources/braille.html
                                                                     minute from an alphabet. BCI studies to date have
Blindness Resource Center. (2002). Braille on the Internet.          been conducted largely in controlled laboratory set-
   Retrieved May 10, 2004, from http://www.nyise.org/braille.html    tings, although the field is beginning to target real-
                                                                     world environments for BCI use.
                                                                         A driving motivation behind BCI research has been
                                                                     the desire to help people with severe physical disabil-
76 ❘❙❚ BERKSHIRE ENCYCLOPEDIA OF HUMAN-COMPUTER INTERACTION



ities such as locked-in syndrome, a condition caused       Sensorimotor Cortex Rhythms
by disease, stroke, or injury in which a person remains    Cortical rhythms represent the synchronized activ-
cognitively intact but is completely paralyzed and un-     ity of large numbers of brain cells in the cortex
able to speak. Traditional assistive technologies for      that create waves of electrical activity over the brain.
computer access depend on small muscle movements,          These rhythms are characterized by their frequency
typically using the limbs, eyes, mouth, or tongue to ac-   of occurrence; for example, a rhythm occurring
tivate switches. People with locked-in syndrome            between eight and twelve times a second is denoted as
have such severely limited mobility that system in-        mu, and one occurring between eighteen and twenty-
put through physical movement is infeasible or un-         six times a second is referred to as beta. When
reliable. A BCI system detects tiny electrophysiological   recorded over the motor cortex, these rhythms are
changes in brain signals to produce control instruc-       affected by movement or intent to move. Studies have
tions for a computer, thereby making it unnecessary        shown that people can learn via operant-condition-
for a user to have reliable muscle movement.               ing methods to increase and decrease the voltage
     Researchers have created applications for non-        of these cortical rhythms (in tens of microvolts) to
disabled users as well, including gaming systems and       control a computer or other device. BCIs based on
systems that allow hands-free, heads-up control of         processing sensorimotor rhythms have been used to
devices, including landing an aircraft. Brain signal       operate a binary spelling program and two-dimen-
interfaces have been used in psychotherapy to mon-         sional cursor movement.
itor relaxation responses and to teach meditation,
although these are biofeedback rather than control
interfaces.                                                Slow Cortical Potentials
                                                           Slow cortical potentials (SCPs) are low-frequency
                                                           shifts of cortical voltage that people can learn to con-
Brain Signal Characteristics                               trol with practice. SCP shifts can occur in durations
Brain signals are recorded using two general ap-           of a hundred milliseconds up to several seconds. SCP
proaches. The most ubiquitous approach is the elec-        signals are based over the frontal and central cor-
troencephalogram (EEG), a recording of signals             tex area, and are typically influenced by emotional
representing activity over the entire surface of the       or mental imagery, as well as imagined movement.
brain or a large region of the brain, often incorpo-       SCPs are recorded both from electrodes on the scalp
rating the activity of millions of neurons. An EEG         using an operant conditioning approach and from
can be recorded noninvasively (without surgery)            positive reinforcement to train users to alter their
from electrodes placed on the scalp, or invasively (re-    SCPs. Both nondisabled and locked-in subjects have
quiring surgery) from electrodes implanted inside          been able to learn to affect their SCP amplitude, shift-
the skull or on the surface of the brain. Brain signals    ing it in either an electrically positive or negative di-
can also be recorded from tiny electrodes placed di-       rection. Locked-in subjects have used SCPs to
rectly inside the brain cortex, allowing researchers       communicate, operating a spelling program to write
to obtain signals from individual neurons or small         letters and even surfing with a simple web browser.
numbers of colocated neurons.
    Several categories of brain signals have been ex-
plored for BCIs, including rhythms from the senso-         Evoked Potentials
rimotor cortex, slow cortical potentials, evoked           The brain's responses to stimuli can also be detected
potentials, and action potentials of single neurons.       and used for BCI control. The P300 response occurs
A BCI system achieves control by detecting changes         when a subject is presented with something famil-
in the voltage of a brain signal, the frequency of a       iar, such as a photo of a loved one, or of interest, such
signal, and responses to stimuli. The type of brain        as a letter selected from an alphabet. The P300 re-
signal processed has implications for the nature of        sponse can be evoked by almost any stimulus, but
the user’s interaction with the system.                    most BCI systems employ either visual or auditory
                                                                          BRAIN-COMPUTER INTERFACES ❚❙❘ 77



stimuli. Screening for the P300 is accomplished            by learning to raise or lower some aspect of his or
through an “oddball paradigm,” where the subject           her brain signals, usually amplitude or frequency.
views a series of images or hears a series of tones, at-   Continuous transducers have enabled users to per-
tending to the one that is different from the rest. If     form selections by raising or lowering a cursor to hit
there is a spike in the signal power over the parietal     a target on a screen. A continuous transducer is anal-
region of the brain approximately 300 millisec-            ogous to a continuous device, such as a mouse or joy-
onds after the “oddball” or different stimulus, then       stick, that always reports its current position.
the subject has a good P300 response. One practical             A discrete transducer is analogous to a switch de-
application that has been demonstrated with P300           vice that sends a signal when activated. Discrete trans-
control is a spelling device. The device works by flash-    ducers produce a single value upon activation. A user
ing rows and columns of an alphabet grid and aver-         typically activates a discrete transducer by learning
aging the P300 responses to determine which letter         to cause an event in the brain that can be detected
the subject is focusing on. P300 responses have            by a BCI system. Discrete transducers have been used
also been used to enable a subject to interact with        to make decisions, such as whether to turn in navi-
a virtual world by concentrating on flashing vir-          gating a maze. Continuous transducers can emulate
tual objects until the desired one is activated.           discrete transducers by introducing a threshold that
                                                           the user must cross to “activate” the switch.
                                                                Direct-spatial-positioning transducers produce a
Action Potentials of Single Neurons                        direct selection out of a range of selection choices.
Another approach to BCI control is to record from          These transducers are typically associated with evoked
individual neural cells via an implanted electrode.        responses, such as P300, that occur naturally and do
In one study, a tiny hollow glass electrode was im-        not have to be learned. Direct transducers have been
planted in the motor cortices of three locked-in sub-      used to implement spelling, by flashing letters arranged
jects, enabling neural firings to be captured and          in a grid repeatedly and averaging the brain signal re-
recorded. Subjects attempted to control this form of       sponse in order to determine which letter the user was
BCI by increasing or decreasing the frequency of neu-      focusing on. A direct spatial positioning transducer is
ral firings, typically by imagining motions of para-        analogous to a touch screen.
lyzed limbs. This BCI was tested for controlling                BCI system architectures have many common
two-dimensional cursor movement in communica-              functional aspects. Figure 1 shows a simplified model
tions programs such as virtual keyboards. Other ap-        of a general BCI system design as described by Mason
proaches utilizing electrode arrays or bundles of          and Birch (2003).
microwires are being researched in animal studies.              Brain signals are captured from the user by an
                                                           acquisition method, such as scalp electrodes or im-
                                                           planted electrodes. The signals are then processed by
Interaction Styles With BCIs                               an acquisition component called a feature extractor
How best to map signals from the brain to the              that identifies signal changes that could signify in-
control systems of devices is a relatively new area        tent. A signal translator then maps the extracted sig-
of study. A BCI transducer is a system component           nals to device controls, which in turn send signals to
that takes a brain signal as input and outputs a con-      a control interface for a device, such as a cursor, a
trol signal. BCI transducers fall into three general       television, or a wheelchair. A display may return feed-
categories: continuous, discrete, and direct spatial       back information to the user.
positioning.                                                    Feedback is traditionally provided to BCI users
    Continuous transducers produce a stream of             through both auditory and visual cues, but some test-
values within a specified range. These values can be        ing methods allow for haptic (touch) feedback and
mapped to cursor position on a screen, or they can di-     electrical stimulation. Which feedback mechanisms
rectly change the size or shape of an object (such as a    are most effective usually depends on the abilities
progress bar). A user activates a continuous transducer    and disabilities of the user; many severely disabled
78 ❘❙❚ BERKSHIRE ENCYCLOPEDIA OF HUMAN-COMPUTER INTERACTION




                                                                            FIGURE 1.   BCI system architecture

users have problems with vision that can be com-         Neural Prosthetics
pensated for by adding auditory cues to BCI tasks.       A BCI application with significant implications is
Some research teams have embraced usability test-        neural prostheses, which are orthoses or muscle stim-
ing to determine what forms of feedback are most         ulators controlled by brain signals. In effect, a neu-
effective; this research is under way.                   ral prosthesis could reconnect the brain to paralyzed
                                                         limbs, essentially creating an artificial nervous sys-
                                                         tem. BCI controls could be used to stimulate mus-
Applications for BCIs                                    cles in paralyzed arms and legs to enable a subject to
As the BCI field matures, considerable interest has       learn to move them again. Preliminary work on a
arisen in applying BCI techniques to real-world prob-    neurally controlled virtual hand was reported in 2000
lems. The principal goal has been to provide a com-      with implanted electrodes: a noninvasive BCU has
munication channel for people with severe motor          been demonstrated to control a hand-grasp ortho-
disabilities, but other applications may also be         sis for a person whose hand was paralyzed. An
possible. Researchers are focusing on applications       SSVEP-based BCI has also been used to control a
for BCI technologies in several critical areas:          functional electrical stimulator to activate paralyzed
                                                         muscles for knee extension.
Communication
Making communication possible for a locked-in per-       Mobility
son is a critical and very difficult task. Much of the    Restoring mobility to people with severe disabilities
work in BCI technology centers around communi-           is another area of research. A neurally controlled
cation, generally in the form of virtual keyboards or    wheelchair could provide a degree of freedom and
iconic selection systems.                                greatly improve the quality of life for locked-in peo-
                                                         ple. Researchers are exploring virtual navigation tasks,
Environmental Control                                    such as virtual driving and a virtual apartment, as
The ability to control the physical environment is       well as maze navigation. A noninvasive BCI was used
also an important quality-of-life issue. Devices         to direct a remote-control vehicle, with the aim of
that permit environmental control make it possi-         eventually transferring driving skills to a power
ble for locked-in people to turn a TV to a desired       wheelchair.
channel and to turn lights on and off, as well as con-
trolling other physical objects in their world.
                                                         Issues and Challenges for BCI
Internet Access                                          There are many obstacles to overcome before BCIs
The Internet has the potential to enhance the lives      can be used in real-world scenarios. The minute elec-
of locked-in people significantly. Access to the         trophysiological changes that characterize BCI con-
Internet can provide shopping, entertainment, edu-       trols are subject to interference from both electrical
cation, and sometimes even employment opportu-           and cognitive sources. Brain-signal complexity and
nities to people with severe disabilities. Efforts are   variability make detecting and interpreting changes
under way to develop paradigms for BCI interaction       very difficult except under controlled circumstances.
with Web browsers.                                       Especially with severely disabled users, the effects of
                                                                                BRAIN-COMPUTER INTERFACES ❚❙❘ 79



medications, blood sugar levels, and stimulants such
as caffeine can all be significant. Cognitive distrac-       FURTHER READING
tions such as ambient environmental noise can af-
fect a person’s ability to control a BCI in addition to     Bayliss, J. D., & Ballard, D. H. (2000). Recognizing evoked potentials
                                                               in a virtual environment. Advances in Neural Information Processing
increasing the cognitive load the person bears.                Systems, 12, 3–9.
Artifacts such as eye blinks or other muscle move-          Birbaumer, N., Kubler, A., Ghanayim, N., Hinterberger, T., Perelmouter,
ments can mask control signals.                                J. Kaiser, J., et al. (2000). The thought translation device (TTD) for
    BCIs and other biometric devices are also plagued          completely paralyzed patients. IEEE Transactions on Rehabilitation
                                                               Engineering, 8(2), 190–193.
by what is termed the Midas touch problem: How              Birch, G. E., & Mason, S. G. (2000). Brain-computer interface research
does the user signal intent to control when the brain          at the Neil Squire Foundation. IEEE Transactions on Rehabilitation
is active constantly? Hybrid discrete/continuous               Engineering, 8(2), 193–195.
                                                            Chapin, J., & Nicolelis, M. (2002). Closed-loop brain-machine in-
transducers may be the answer to this problem,                 terfaces. In J. R. Wolpaw & T. Vaughan (Eds.), Proceedings of Brain-
but it is still a major issue for BCIs in the real world.      Computer Interfaces for Communication and Control: Vol. 2. Moving
    Another important issue currently is that BCI              Beyond Demonstration, Program and Abstracts (p. 38). Rensse-
systems require expert assistance to operate. As BCI           laerville, NY.
                                                            Donchin, E., Spencer, K., & Wijesinghe, R. (2000). The mental
systems mature, the expectation is that more of the            prosthesis: Assessing the speed of a P300-based brain-computer
precise tuning and calibration of these systems                interface. IEEE Transactions on Rehabilitation Engineering, 8(2),
may be performed automatically.                                174–179.
    Although BCIs have been studied since the mid-          Kandel, E., Schwartz, J., & Jessell, T. (2000). Principles of neural sci-
                                                               ence (4th ed.). New York: McGraw-Hill Health Professions Division.
1980s, researchers are just beginning to explore their      Kennedy, P. R., Bakay, R. A. E., Moore, M. M., Adams, K., & Goldwaithe,
enormous potential. Understanding brain signals and            J. (2000). Direct control of a computer from the human central
patterns is a difficult task, but only through such an          nervous system. IEEE Transactions on Rehabilitation Engineering,
                                                               8(2), 198–202.
understanding will BCIs become feasible. Currently          Lauer, R. T., Peckham, P. H., Kilgore, K. L., & Heetderks, W. J.
there is a lively debate on the best approach to acquir-       (2000). Applications of cortical signals to neuroprosthetic control:
ing brain signals. Invasive techniques, such as im-            A critical review. IEEE Transactions on Rehabilitation Engineering,
planted electrodes, could provide better control               8(2), 205–207.
                                                            Levine, S. P., Huggins, J. E., BeMent, S. L., Kushwaha, R. K., Schuh,
through clearer, more distinct signal acquisition. Non-        L. A., Rohde, M. M., et al. (2000). A direct-brain interface based
invasive techniques, such as scalp electrodes, could           on event-related potentials. IEEE Transactions on Rehabilitation
be improved by reducing noise and incorporating so-            Engineering, 8(2), 180–185.
phisticated filters. Although research to date has fo-       Mankoff, J., Dey, A., Moore, M., & Batra, U. (2002). Web accessibility
                                                               for low bandwidth input. In Proceedings of ASSETS 2002 (pp. 89–96).
cused mainly on controlling output from the brain,             Edinburgh, UK: ACM Press.
recent efforts are also focusing on input channels.         Mason, S. G., & Birch, G. E. (In press). A general framework for brain-
Much work also remains to be done on appropriate               computer interface design. IEEE Transactions on Neural Systems
                                                               and Rehabilitation Technology.
mappings to control signals.                                Moore, M., Mankoff, J., Mynatt, E., & Kennedy, P. (2002). Nudge
    As work in the field continues, mainstream ap-              and shove: Frequency thresholding for navigation in direct brain-
plications for BCIs may emerge, perhaps for peo-               computer interfaces. In Proceedings of SIGCHI 2001Conference on
ple in situations of imposed disability, such as jet           Human Factors in Computing Systems (pp. 361–362). New York:
                                                               ACM Press.
pilots experiencing high G-forces during maneuvers,         Perelmouter, J., & Birbaumer, N. (2000). A binary spelling interface
or for people in situations that require hands-free,           with random errors. IEEE Transactions on Rehabilitation Engineer-
heads-up interfaces. Researchers in the BCI field are           ing, 8(2), 227–232.
just beginning to explore the possibilities of real-        Pfurtscheller, G., Neuper, C., Guger, C., Harkam, W., Ramoser, H.,
                                                               Schlögl, A., et al. (2000). Current trends in Graz brain-computer in-
world applications for brain signal control.                   terface (BCI) research. IEEE Transactions on Rehabilitation Engi-
                                                               neering, 8(2), 216–218.
                Melody M. Moore, Adriane B. Davis,          Tomori, O., & Moore, M. (2003). The neurally controllable Internet
                                                               browser. In Proceedings of SIGCHI 03 (pp. 796–798).
                             and Brendan Allison            Wolpaw, J. R., Birbaumer, N., McFarland, D., Pfurtscheller, G., &
                                                               Vaughan, T. (2002). Brain-computer interfaces for communica-
See also Physiology                                            tion and control. Clinical Neurophysiology, 113, 767–791.
80 ❘❙❚ BERKSHIRE ENCYCLOPEDIA OF HUMAN-COMPUTER INTERACTION



Wolpaw, J. R., McFarland, D. J., & Vaughan, T. M. (2000). Brain-com-    this ethnographic style of contextually immersed in-
  puter interface research at the Wadsworth Center. IEEE Transactions   vestigation is that of Michael Byrne and his colleagues
  on Rehabilitation Engineering, 8(2), 222–226.
                                                                        (1999), who used their observations to create a
                                                                        taxonomy of Web-browsing tasks. Their method in-
                                                                        volved videotaping eight people whenever they used
                                                                        a browser in their work. The participants were en-
                   BROWSERS                                             couraged to continually articulate their objectives
                                                                        and tasks, essentially thinking aloud. A total of five
For millions of computer users worldwide, a browser                     hours of Web use was captured on video and tran-
is the main interface with the World Wide Web, the                      scribed, and a six-part taxonomy of stereotypical
world’s foremost Internet information exchange serv-                    tasks emerged:
ice. Banking, shopping, keeping in contact with
                                                                          1. Use information: activities relating to the use
friends and family through e-mail, accessing news,
                                                                             of information gathered on the Web;
looking words up in the dictionary, finding facts and
                                                                          2. Locate on page: searching for particular in-
solving puzzles—all of these activities and many
                                                                             formation on a page;
more can be carried out on the Web.
                                                                          3. Go to: the act of trying to get the browser to dis-
     After the 1993 release of the first graphical user
                                                                             play a particular URL (Web address);
interface Web browser (NCSA Mosaic), the Web rap-
                                                                          4. Provide information: sending information to a
idly evolved from a small user base of scientists ac-
                                                                             website through the browser (for example, pro-
cessing a small set of interlinked text documents to
                                                                             viding a billing address or supplying search
approximately 600 million users accessing billions
                                                                             terms to a search engine);
of webpages that make use of many different media,
                                                                          5. Configure browser: changing the configuration
including text, graphics, video, audio, and anima-
                                                                             of the browser itself; and
tion. Economies of scale clearly apply to the effec-
                                                                          6. React to environment: supplying information
tiveness of Web browsers.
                                                                             required for the browser to continue its opera-
     Although there has been substantial work on the
                                                                             tion (for example, responding to a dialog box
“webification” of sources of information (for exam-
                                                                             that asks where a downloaded file should be
ple, educational course materials), there has been
                                                                             saved).
surprisingly little research into understanding and
characterizing Web user’s tasks, developing better                          Although these results were derived from only a
browsers to support those tasks, and evaluating the                     few hours of Web use by a few people, they provide
browsers’ success. But ethnographic and field stud-                      initial insights into the tasks and actions accom-
ies can give us a contextual understanding of Web                       plished using a browser.
use, and longitudinal records of users’ actions make                        Another approach to studying how people use
possible long-term quantitative analyses, which in                      the Web is to automatically collect logs of users’
turn are leading to low-level work on evaluating and                    actions. The logs can then be analyzed to provide a
improving browsers.                                                     wide variety of quantitative characterizations of Web
                                                                        use. Although this approach cannot provide insights
                                                                        into the context of the users’ actions, it has the ad-
What Do Web Users Do?                                                   vantage of being implementable on a large scale.
The best way to understand fully what users do with                     Months or years of logged data from dozens of users
their browsers, why they do it, and the problems they                   can be included in an analysis.
encounter is to observe and question users directly                         Two approaches have been used to log Web-use
as they go about their everyday work. Unfortunately                     data. Server-side logs collect data showing which pages
this approach puts inordinate demands on re-                            were served to which IP address, allowing Web de-
searchers’ time, so it is normally used only with small                 signers to see, for instance, which parts of their sites
sets of participants. The study that best demonstrates                  are particularly popular or unpopular. Unfortunately,
                                                                                               BROWSERS ❚❙❘ 81



server-side logs only poorly characterize Web usability     of participants). Cockburn and McKenzie’s log analy-
issues.                                                     sis suggested that bookmark use had evolved, with
     The second approach uses client-side logs, which       users either maintaining large bookmark collections
are established by equipping the Web browser (or a          or almost none: The total number of bookmarks in
client-side browser proxy) so that it records the exact     participants’ collections ranged from 0 to 587, with
history of the user’s actions with the browser. The first    a mean of 184 and a high standard deviation of 166.
two client-side log analyses of Web use were both con-           A final empirical characterization of Web use
ducted in 1995 using the then-popular XMosaic               from Cockburn and McKenzie’s log analysis is that
browser. The participants in both studies were pri-         Web browsing is surprisingly rapid, with many or
marily staff, faculty, and students in university com-      most webpages being visited for only a very brief pe-
puting departments. Lara Catledge and James Pitkow          riod (less than a couple of seconds). There are two
logged 3 weeks of use by 107 users in 1995, while Linda     main types of browsing behavior that can explain
Tauscher and Saul Greenberg analyzed 5 to 6 weeks           the very short page visits. First, many webpages are
of use by 23 users in 1995. The studies made several        simply used as routes to other pages, with users
important contributions to our understanding of what        following known trails through the series of links
users do with the Web. In particular, they revealed that    that are displayed at known locations on the pages.
link selection (clicking on links in the Web browser)       Second, users can almost simultaneously display a
accounts for approximately 52 percent of all webpage        series of candidate “interesting” pages in inde-
displays, that webpage revisitation (returning to           pendent top-level windows by shift-clicking on the
previously visited webpages) is a dominant naviga-          link or by using the link’s context menu. For exam-
tion behavior, that the Back button is very heavily used,   ple, the user may rapidly pop up several new win-
and that other navigation actions, such as typing URLs,     dows for each of the top result links shown as a result
clicking on the Forward button, or selecting book-          of a Google search.
marked pages, were only lightly used. Tauscher and
Greenberg also analyzed the recurrence rate of page
visits—“the probability that any URL visited is a re-       Improving the Web Browser
peat of a previous visit, expressed as a percentage”
(Tauscher and Greenberg 1997, 112). They found a            User Interface
recurrence rate of approximately 60 percent, mean-          The studies reported above inform designers
ing that on average users had previously seen approxi-      about what users do with the current versions of their
mately three out of five pages visited. In a 2001 study,     browsers. Naturally, there is a chicken-and-egg prob-
Andy Cockburn and Bruce McKenzie showed that the            lem in that stereotypical browser use is strongly af-
average recurrence rate had increased to approximately      fected by the support provided by browsers. Browser
80 percent—four out of five pages a user sees are ones       interfaces can be improved both by designing to bet-
he or she has seen previously. Given these high recur-      ter support the stereotypes and by innovative design
rence rates, it is clearly important for browsers to pro-   that enables previously difficult or impossible tasks.
vide effective tools for revisitation.                           The problems of hypertext navigation were
     The 1995 log analyses suggested that people rarely     well known long before the Web. As users navigate
used bookmarks, with less than 2 percent of user            through the richly interconnected information nodes
actions involving bookmark use. However, a survey           of the Web (or any hypertextual information space)
conducted the following year (Pitkow, 1996) indicates       their short-term memory becomes overloaded with
that users at least had the intention of using book-        the branches made, and they become “lost in hy-
marks, with 84 percent of respondents having more           perspace.” In the late 1980s many researchers were
than eleven bookmarks. Pitkow reported from a               experimenting with graphical depictions of hyper-
survey of 6,619 users that organizing retrieved in-         text spaces in order to help users orient themselves:
formation is one of the top three problems people re-       For example, the popular Apple language Hypercard
port relating to using the Web (reported by 34 percent      provided a thumbnail graphical representation of the
82 ❘❙❚ BERKSHIRE ENCYCLOPEDIA OF HUMAN-COMPUTER INTERACTION



recent cards displayed, and gIBIS provided a network               Another problem users have with current Web
diagram of design argumentation. Soon after the Web           browsers is that they misunderstand the behavior
emerged in 1991, similar graphical techniques were            of the Back button. An experiment showed that eight
being constructed to aid Web navigation. Example              of eleven computer scientists incorrectly predicted
systems included MosaicG, which provided                      the behavior of Back in simple Web navigation tasks.
thumbnail images of the visited pages arranged in a           The problem stems from users believing that Back
tree hierarchy, WebNet, which drew a hub-and-spoke            provides access to a complete history of previously
representation of the pages users visited and the links       visited pages, rather than the stack-based subset that
available from them, and the Navigational View                can actually be accessed. Cockburn and his colleagues
Builder, which could generate a wide variety of               describe the behavior and make an evaluation of a
two-dimensional and three-dimensional represen-               true history-based Back system, but results indicate
tations of the Web.                                           that the pros and cons of the technique are closely
    Despite the abundance of tools that provide               balanced, such that the advantages do not outweigh
graphical representations of the user’s history, none         the difficulties inherent in making a switch from cur-
have been widely adopted. Similarly, log analyses of          rent behavior.
Web use show that users seldom use the history tools               The World Wide Web revolution has been a great
provided by all of the main Web browsers. Given that          success in bringing computer technology to the
Web revisitation is such a common activity, why               masses. The widespread adoption and deployment
are these history tools so lightly used? The best ex-         of the Web and the browsers used to access it hap-
planation seems to be that these tools are not needed         pened largely without input from researchers in hu-
most of the time, so they are unlikely to be on per-          man-computer interaction. Those researchers are
manent display, where they would compete with other           now improving their understanding of the usability
applications for screen real estate. Once iconified, the       issues associated with Web browsers and browsing.
tools are not ready to hand, and it is overhead for           As the technology and understanding matures, we
users to think of using them, take the actions to             can expect browser interfaces to improve, enhanc-
display them, orient themselves within the informa-           ing the efficiency of Web navigation and reducing
tion they display, and make appropriate selections.           the sensation of becoming lost in the Web.
    While the projects above focus on extending
browser functionality, several other research projects                                                           Andy Cockburn
have investigated rationalizing and improving browsers’
current capabilities. The interface mechanisms for re-        See also Mosaic; Website Design
turning to previously visited pages have been a par-
ticular focus. Current browsers support a wide range
of disparate facilities for revisitation, including the       FURTHER READING
Back and Forward buttons and menus, menus that al-
low users to type or paste the URLs of websites the user      Abrams, D., Baecker R., & Chignell, M. (1998). Information archiv-
wants to visit, the history list, bookmarks or lists of fa-      ing with bookmarks: Personal Web space construction and or-
                                                                 ganization. In Proceedings of CHI'98 Conference on Human Factors
vorites, and the links toolbar. Of these utilities, log          in Computing Systems (pp. 41–48). New York: ACM Press.
analyses suggest that only the Back button is heavily         Ayers, E., & Stasko, J. (1995). Using graphic history in browsing the
used. The WebView system and Glabster both demon-                World Wide Web. In Proceedings of the Fourth International World
strate how history facilities and bookmarks can be en-           Wide Web Conference (pp. 451–459). Retrieved January 19, 2004,
                                                                 from http://www.w3j.com/1/ayers.270/paper/270.html
hanced and integrated within the Back menu, providing         Bainbridge, L. (1991). Verbal protocol analysis. In J. Wilson & E. Corlett
a powerful and unified interface for all revisitation             (Eds.), Evaluation of human work: A practical ergonomics method-
tasks. Both WebView and Glabster automatically cap-              ology (pp. 161–179). London: Taylor and Francis.
                                                              Byrne, M., John, B., Wehrle, N., & Crow, D. (1999). The tangled
ture thumbnail images of webpages, making it easier              Web we wove: A taskonomy of WWW Use. In Proceedings of CHI'99
for the user to identify previously visited pages from           Conference on Human Factors in Computing Systems (pp. 544–551).
the set displayed within the back menu.                          New York: ACM Press.
                                                                                                                                BROWSERS ❚❙❘ 83



Catledge, L., & Pitkow, J. (1995). Characterizing browsing strategies in the   Conklin, J., & Begeman, M. (1988). gIBIS: A hypertext tool for ex-
   World Wide Web. In Computer systems and ISDN systems: Proceedings              ploratory discussion. ACM Transactions on Office Information
   of the Third International World Wide Web Conference, 27, 1065–1073).          Systems, 6(4), 303–313.
Chi, E., Pirolli, P., & Pitkow, J. (2000). The scent of a site: A system for   Coulouris, G., & Thimbleby, H. (1992). HyperProgramming. Woking-
   analyzing and predicting information scent, usage, and usability               ham, UK: Addison-Wesley Longman.
   of a Web site. In Proceedings of CHI'2000 Conference on Human               Fischer, G. (1998). Making learning a part of life: Beyond the 'gift-
   Factors in Computing Systems (pp.161–168). New York: ACM Press.                wrapping' approach of technology. In P. Alheit & E. Kammler (Eds.),
Cockburn, A., Greenberg, S., McKenzie, B., Jason Smith, M., & Kaasten,            Lifelong learning and its impact on social and regional development
   S. (1999). WebView: A graphical aid for revisiting Web pages. In               (pp. 435–462). Bremen, Germany: Donat Verlag.
   Proceedings of the 1999 Computer Human Interaction Specialist               Kaasten, S., & Greenberg, S. (2001). Integrating Back, History and
   Interest Group of the Ergonomics Society of Australia (OzCHI'91)               bookmarks in Web browsers. In Proceedings of CHI'01 (pp. 379–380).
   (pp. 15–22). Retrieved January 19, 2004, from http://www.cpsc.ucal-            New York: ACM Press.
   gary.ca/Research/grouplab/papers/1999/99-WebView.Ozchi/                     Mukherjea, S., & Foley, J. (1995). Visualizing the World Wide Web with
   Html/webview.html                                                              the navigational view builder. Computer Systems and ISDN Systems,
Cockburn, A., & Jones, S. (1996). Which way now? Analysing and eas-               27(6), 1075–1087.
   ing inadequacies in WWW navigation. International Journal of                Nielsen, J. (1990). The art of navigating through HyperText: Lost in
   Human-Computer Studies, 45(1), 105–129.                                        hyperspace. Communications of the ACM, 33(3), 296–310.
Cockburn, A., & McKenzie, B. (2001). What do Web users do? An em-              Pirolli, P., Pitkow, J., & Rao, R. (1996). Silk from a sow's ear: Extracting
   pirical analysis of Web use. International Journal of Human-                   usable structures from the Web. In R. Bilger, S. Guest, & M. J. Tauber
   Computer Studies, 54(6), 903–922.                                              (Eds.), Proceedings of CHI'96 Conference on Human Factors in
Cockburn, A., McKenzie, B., & Jason Smith, M. (2002). Pushing Back:               Computing Systems (pp. 118–125). New York: ACM Press.
   Evaluating a new behaviour for the Back and Forward buttons                 Pitkow, J. (n.d.). GVU's WWW User Surveys. Retrieved January 19,
   in Web browsers. International Journal of Human-Computer Studies,              2004, from http://www.gvu.gatech.edu/user_surveys/
   57(5), 397–414.                                                             Tauscher, L., & Greenberg, S. (1997). How people revisit Web pages:
Conklin, J. (1988). Hypertext: An introduction and survey. In I. Greif            Empirical findings and implications for the design of history
   (Ed.), Computer supported cooperative work: A book of readings                 systems. International Journal of Human Computer Studies, 47(1),
   (pp. 423–475). San Mateo, CA: Morgan-Kauffman.                                 97–138.
                                                                                    CATHODE RAY TUBES

                                                                                                 CAVE

                                                                                           CHATROOMS

                                                                                 CHILDREN AND THE WEB

                                                                                          CLASSROOMS




            C
                                                                           CLIENT-SERVER ARCHITECTURE

                                                                               COGNITIVE WALKTHROUGH

                                                                                     COLLABORATORIES

                                                                                            COMPILERS

                                                                 COMPUTER-SUPPORTED COOPERATIVE WORK

                                                                               CONSTRAINT SATISFACTION

                                                                              CONVERGING TECHNOLOGIES

                                                                                    CYBERCOMMUNITIES

                                                                                             CYBERSEX

                                                                                              CYBORGS




                                                         sion and radar over the next fifty years provided the
    CATHODE RAY TUBES                                    impetus for continual improvements. With the emer-
                                                         gence of desktop computing in the 1980s, the CRT
The cathode ray tube (CRT) has been the dominant         market expanded, and its performance continued to
display technology for decades. Products that utilize    evolve. As portability has come to be more and more
CRTs include television and computer screens in the      important in the consumer electronics industry, the
consumer and entertainment market, and electronic        CRT has been losing ground. The development of
displays for medical and military applications. CRTs     flat panel technologies such as liquid crystal displays
are of considerable antiquity, originating in the        and plasma displays for portable products, computer
late nineteenth century when William Crookes             screens, and television makes the CRT very vulner-
(1832–1919) studied the effects of generating an elec-   able. Because of the CRT’s maturity and compara-
trical discharge in tubes filled with various gases.     tively low cost, however, its application will be assured
(The tubes were known as discharge tubes.) It was        for many years to come.
over thirty years later in 1929 that the CRT was
utilized to construct actual imagery for television
applications by Vladimir Zworykin (1889–1982) of         How Cathode Ray Tubes Work
Westinghouse Electric Corporation. The further de-       A CRT produces images when an electron beam
velopment and optimization of the CRT for televi-        is scanned over a display screen in a pattern that is

                                                                                                               85
86 ❘❙❚ BERKSHIRE ENCYCLOPEDIA OF HUMAN-COMPUTER INTERACTION



determined by a deflection mechanism. The display            must be and on when the beam should be moved
screen is coated with a thin layer of phosphor that         across different portions of the screen.
luminesces under the bombardment of electrons.
By this means the display screen provides a two-
dimensional visual display, corresponding to in-            Displaying Color
formation contained in the electron beam. There are         One of the most important tasks of the modern dis-
four major components of a CRT display: the                 play is rendering full-color images. Shadow-masking
vacuum tube, the electron source (known as the elec-        configurations are by far the most successful way
tron “gun”), the deflection mechanism, and the phos-         to create full color images in CRT displays. The
phor screen.                                                shadow mask CRT typically uses three electron beams
    The tube (sometimes referred to as a bulb) is           deflected by one coil (the simplest configuration).
maintained at a very high vacuum level to facili-           The electron beams traverse a perforated metal mask
tate the flow of electrons in the electron beam. The         (shadow mask) before impinging on selected
front surface of the tube defines the visual area           phosphor materials (there are three sorts of phos-
of the display, and it is this front surface that is cov-   phor that can emit red, green, and blue light). The
ered with phosphor, which is in turn covered by             shadow mask apertures are typically configured as
the anode (the electron-collecting electrode). The          stripes, circles, or slots. The arrangement of the elec-
tube has three main sections: the front surface, the        tron optics and the deflection system is such that
funnel, and the neck. The entire tube is typically          three electron beams converge onto the screen af-
made of glass so that very high vacuums can be              ter passing through the shadow mask, each beam
sustained, but in some cases the funnel and neck            impinging on one phosphor, which, when bom-
can be fabricated from metal or ceramic. For de-            barded with electrons, emits red, green, or blue
manding applications that require additional ro-            visible light. The red, green, and blue phosphors are
bustness, an implosion-proof faceplate may be               spatially arranged on the viewing screen.
secured to the front tube surface for durability. This          The Tr init ron desig n, invented by Sony
typically comes at the expense of optical through-          Corporation, uses vertical stripe arrays rather than
put, but antireflection coatings are often used to          circular or slotted apertures. These arrays alternate
improve contrast and to compensate for the trans-           red, green, and blue when viewed from the faceplate
mission losses.                                             side of the tube. There is a single electron source,
    The electron source, a hot cathode at the far end       rather than three, which eliminates the problem of
from the front surface, generates a high-density            beam convergence. The Trinitron also has superior
electron beam whose current can be modulated.               resolution in the vertical direction since its apertures
The electron beam can be focused or reflected—              are not limited in that direction. The only negative
deflected—by electrostatic or magnetic methods,             attribute of the Trinitron is that the mask is not self-
and this deflection steers the electron beam to des-         supporting, which ultimately limits the size of the
ignated positions of the front surface to create visual     vacuum tube. The advantages of CRT displays in-
imagery.                                                    clude their maturity, their well-understood manu-
    The phosphor screen on the inside front surface         facturing process, their ability to provide full-color
of the tube converts the electron beam into visible         and high-resolution imaging, and the comparatively
light output. On top of the phosphor particles is the       low cost for high information content. CRTs are vul-
thin layer of conducting material (usually aluminum)        nerable to competition from liquid crystal displays
that serves as the anode, drawing the electrons toward      and plasma displays (both of which make possible
the screen. The directions on how to manipulate the         flat-panel displays), however, because CRTs are bulky,
electron stream are contained in an electronic sig-         heavy, and big power consumers. In addition to
nal called a composite video signal. This signal con-       the utility of flat-panel display for portable applica-
tains information on how intense the electron beam          tions for which CRTs could never be considered, flat-
                                                                                                                          CAVE ❚❙❘ 87



panel displays have made significant inroads into                        left eye image should be seen by the left eye and are
desktop monitors and large-area televisions. As the                      opaque otherwise. Similarly, the right eye gets the right
price of flat-panel displays continues to plummet,                        image. Images need to be generated at 100 to 120 hertz
they are certain to capture even more of the CRT                         so each eye can get a flicker-free 50- to 60-hertz dis-
market in the future.                                                    play. All screens need to be synchronized so that each
                                                                         eye sees the same phase stereo image on every screen,
                                     Gregory Philip Crawford             a requirement that until 2003 meant that only the
                                                                         most expensive SGI (Silicon Graphics, Inc.) computer
See also Liquid Crystal Displays                                         graphics systems could be used. Synchronizing PC
                                                                         graphics cards now reduce the cost of CAVE com-
                                                                         puting and image generation by 90 percent.
FURTHER READING                                                               The CAVE’s projection onto the screens does not
                                                                         need to keep up with the viewer’s head motion nearly
Castelliano, J. (1992). Handbook of display technology. San Diego, CA:   as much as is required in a head-mounted VR dis-
   Academic Press.
Keller, P. A. (1997). Electronic display measurement. New York:
                                                                         play (HMD), which needs to have small screens at-
   Wiley SID.                                                            tached in front of the eyes. Of course, any movement
MacDonald, L. W., & Lowe, A. C. (1997). Display systems: Design          of the viewer’s body within the space requires up-
   and applications. New York: Wiley SID.                                dating the scene perspective, but in normal inves-
                                                                         tigative use, the CAVE needs to keep up only with
                                                                         body motion, not head rotation; the important re-
                                                                         sult is that the delay of trackers is dramatically less
                                                                         of a problem with CAVEs than with HMDs. In ad-
                           CAVE                                          dition, although only one viewer is tracked, other
                                                                         people can share the CAVE visuals at the same time;
The CAVE is a virtual reality (VR) room, typically 3                     their view is also in stereo and does not swing with
by 3 by 3 meters in size, whose walls, floor, and some-                   the tracked user’s head rotation, although their per-
times ceiling are made entirely of computer-projected                    spective is still somewhat skewed. Often the person
screens. Viewers wear a six-degree-of-freedom loca-
tion sensor called a tracker so that when they move
within the CAVE, correct viewer-centered perspec-
tives and surround-stereo projections are produced
fast enough to give a strong sense of 3D visual im-
mersion. Viewers can examine details of a com-
plex 3D object simply by walking up to and into it.
The CAVE was invented in 1991 for two reasons:
to help scientists and engineers achieve scientific in-
sight without compromising the color and distortion-
free resolution available then on workstations and
to create a medium worthy of use by fine artists.
    CAVE viewers see not only projected computer-
generated stereo scenes but also their own arms and
bodies, and they can interact easily with other people.
The CAVE uses active stereo, which produces differ-
ent perspective views for the left and right eyes of the                 The CAVE is a multi-person, room-sized, high-
viewer in synchrony with special electronic shutter                      resolution, 3D video and audio environment. Photo cour-
glasses that go clear in front of the left eye when the                  tesy of National Center for Supercomputing Applications.
88 ❘❙❚ BERKSHIRE ENCYCLOPEDIA OF HUMAN-COMPUTER INTERACTION



                                                           CAVE Variants
                                                           Variants of the CAVE include the ImmersaDesk,
                                                           a drafting-table-size rear-projected display with a
                                                           screen set at an angle so that the viewer can look
                                                           down as well as forward into the screen; looking
                                                           down gives a strong sense of being in the scene.
                                                           PARIS uses a similarly angled half-silvered screen
                                                           that is projected from the top; the viewer’s hands
                                                           work under the screen and are superimposed on the
                                                           3D graphics (rather than blocking them, as with nor-
                                                           mal projections).
                                                               The CAVE originally used three-tube stereo pro-
The Personal Augmented Reality Immersive System            jectors with special phosphors to allow a 100- to 120-
(PARIS) has a half-silvered mirror at an angle in front    hertz display without ghosting from slow green phos-
of the user. The screen, above the desk facing down,       phor decay. Tube projectors are now rather dim by
superimposes a stereo image on the user’s hands work-      modern standards, so the CAVE was rebuilt to use
ing beyond the mirror. Photo courtesy of the Electronic    bright digital mirror-based projectors, like those used
Visualization Laboratory.                                  in digital cinema theaters. Projectors require sig-
                                                           nificant alignment and maintenance; wall-sized flat-
in the role of guide or instructor handles the con-        panel screens will be welcomed since they need no
trols (a 3D mouse called “Wanda”) and the student          alignment and have low maintenance and no pro-
wears the tracker to get the best view, a mode of          jection distance.
usage that is quite functional for both learning and           The GeoWall, a passive stereo device, works
demonstrations.                                            differently, polarizing the output of two projectors
     The CAVE uses a rear-screen projection for the        onto a single screen. Viewers wear the throw-away
walls so the viewer does not block the light and cast      polarized glasses used in 3D movies to see stereo.
shadows. The floor is typically projected down from             In addition to visual immersion, the CAVE has
the top, which creates a small shadow around the           synchronized synthetic and sampled surround sound.
viewer’s feet. A CAVE with three walls and a floor min-     The PARIS system features a PHANTOM tactile de-
imally requires a 13- by 10-meter space with a ceiling     vice, which is excellent for manipulating objects the
4.5 meters high. Six-sided CAVEs have rear projec-         size of a bread box or smaller.
tions from every direction, which require much higher
ceilings, more elaborate support structures, and floor
screens that can withstand the weight of several people.   CAVEs for Tele-Immersion
Someday, 3-square-meter flat-panel displays sus-           The CAVE was originally envisioned as a tele-
pended as a ceiling, positioned vertically as walls, and   immersive device to enable distance collaboration
tough enough to walk on would allow CAVEs in nor-          between viewers immersed in their computer-
mal rooms. However, current technology panel-              generated scenes, a kind of 3D phone booth. Much
displays refresh too slowly to use shutter glasses, so     work has gone into building and optimizing ultra-
they must be otherwise modified for stereo display.         high-speed computer networks suitable for shar-
The Varrier method involves placing a barrier screen       ing gigabits of information across a city, region,
so that the computed views to each eye are seen            nation, or indeed, the world. In fact, scientists, en-
through perfectly placed thin black bars, that is, the     gineers, and artists in universities, museums, and
correctly segmented image is placed in dynamic             commercial manufacturing routinely use CAVEs and
perspective behind the barrier in real time. Varrier       variants in this manner.
viewers wear no special glasses since the image sepa-
ration is performed spatially by the barrier screen.                             Tom DeFanti and Dan Sandin
                                                                                                                   CHATROOMS ❚❙❘ 89



See also Virtual Reality; Telepresence; Three-
Dimensional Graphics                                                                        CHATROOMS
                                                                               Defined most broadly, chatrooms are virtual spaces
FURTHER READING                                                                where conversations occur between two or more users
                                                                               in a synchronous or nearly synchronous fashion. Many
Cruz-Neira, C., Sandin, D., & DeFanti, T. A. (1993). Virtual reality:          different types of chat spaces exist on the Internet.
   The design and implementation of the CAVE. Proceedings of the
   SIGGRAPH 93 Computer Graphics Conference, USA, 135–142.
                                                                               One type is Internet Relay Chat (IRC), a multiuser
Czernuszenko, M., Pape, D., Sandin, D., DeFanti, T., Dawe, G. L., &            synchronous chat line often described as the citi-
   Brown, M. D. (1997). The ImmersaDesk and Infinity Wall                      zens band radio of the Internet. Another type of vir-
   projection-based virtual reality displays [Electronic version].             tual space where computer-mediated communication
   Computer Graphics, 31(2), 46–49.
DeFanti, T. A., Brown M. D., & Stevens, R. (Eds.). (1996). Virtual re-         (CMC) takes place is Multi-User Domains (MUDs,
   ality over high-speed networks. IEEE Computer Graphics &                    sometimes called “Multi-User Dungeons,” because of
   Applications, 16(4), 14–17, 42–84.                                          their origin as virtual locations for a Dungeons and
DeFanti, T., Sandin, D., Brown, M., Pape, D., Anstey, J., Bogucki, M.,
   et al. (2001). Technologies for virtual reality/tele-immersion ap-
                                                                               Dragons role-playing type of networked gaming).
   plications: Issues of research in image display and global net-             MUDs were initially distinguished from IRC by their
   working. In R. Earnshaw, et al. (Eds.), Frontiers of Human-Centered         persistence, or continued existence over time, and their
   Computing, Online Communities and Virtual Environments                      malleability, where users may take part in the build-
   (pp. 137–159). London: Springer-Verlag.
Johnson, A., Leigh, J., & Costigan, J. (1998). Multiway tele-immersion
                                                                               ing of a community or even a virtual world, de-
   at Supercomputing ’97. IEEE Computer Graphics and Applications,             pending on the tools and constraints built into the
   18(4), 6–9.                                                                 architecture of their particular MUD. Web-based cha-
Johnson, A., Sandin, D., Dawe, G., Qiu, Z., Thongrong, S., & Plepys,           trooms are a third type of chat space where users may
   D. (2000). Developing the PARIS: Using the CAVE to prototype
   a new VR display [Electronic version]. Proceedings of IPT 2000,             converse synchronously in a persistent location hosted
   CD-ROM. Korab H., & Brown, M. D. (Eds.). (1995). Virtual                    by Internet Service Providers (ISPs) or websites, which
   Environments and Distributed Computing at SC’95: GII Testbed and            may be either large Web portals like Yahoo.com or
   HPC Challenge Applications on the I-WAY. Retrieved November 5,
   2003, from http://www.ncsa.uiuc.edu/General/Training/SC95/GII
                                                                               small individual sites.
   .HPCC.html
Lehner, V. D., & DeFanti, T. A. (1997). Distributed virtual reality:
   Supporting remote collaboration in vehicle design. IEEE Computer              UNIX was not designed to stop people from doing stu-
   Graphics & Applications (pp. 13–17).
Leigh, J., DeFanti, T. A., Johnson, A. E., Brown, M. D., & Sandin,
                                                                                 pid things, because that would also stop them from do-
   D. J. (1997). Global tele-immersion: Better than being there. ICAT            ing clever things.
   ’97, 7th Annual International Conference on Artificial Reality and                                                       —Doug Gwyn
   Tele-Existence, pp. 10–17. University of Tokyo, Virtual Reality Society
   of Japan.
Leigh, J., Johnson, A., Brown, M., Sandin, D., & DeFanti, T. (1999).
   Tele-immersion: Collaborative visualization in immersive envi-                  Another type of chat function on the Internet is
   ronments. IEEE Computer Graphics & Applications (pp. 66–73).                instant messaging (IM), which allows users to chat
Sandin, D. J., Margolis, T., Dawe, G., Leigh, J., and DeFanti, T. A. (2001).
   The Varrier™ auto-stereographic display. Proceedings of Photonics
                                                                               with individuals (or invited groups) in “real time,”
   West 2001: Electronics Imaging, SPIE. Retrieved on November 5,              provided that they know a person’s user name. Instant
   2003, from http://spie.org/web/meetings/programs/pw01/                      messaging is distinguished from other chat functions
   home.html                                                                   in that it is often used to hold multiple, simultane-
Stevens, R., & DeFanti, T. A. (1999). Tele-immersion and collabora-
   tive virtual environments. In I. Foster & C. Kesselman (Eds.),
                                                                               ous, private one-on-one chats with others. IM is also
   The grid: Blueprint for a new computing infrastructure (pp. 131–158).       unusual in that the user can also monitor a list of
   San Francisco: Morgan Kaufmann.                                             online friends to see when they are logged in to the
                                                                               instant messaging service. IM chats also differ
                                                                               from other types of chatrooms in that they are not
                                                                               persistent—that is, a user cannot log in to the same
                                                                               chat after the last chatterer has logged off. Instant
90 ❘❙❚ BERKSHIRE ENCYCLOPEDIA OF HUMAN-COMPUTER INTERACTION



message chats are more likely to occur among a group       nated by a hatch mark (#) and a number. Because
of people with some personal or professional con-          that proved difficult to use as IRC expanded, each
nection than among a group of strangers with only          channel was also given a text label, like #hottub
one shared interest who happen to be in the same           or #gayboston. IRC channels were originally not
virtual space at the same time.                            persistent—anyone could create a channel on any
                                                           conceivable topic, and when the last person logged
                                                           out of that channel it ceased to exist. Only with the
History of Internet Chat                                   introduction in 1996 of Undernet and later DalNet
The first function that allowed for synchronous or          did it become possible to create persistent channels.
nearly synchronous communication over a network                IRC runs through client software—the client soft-
was Talk, available on UNIX machines and the               ware is what allows the user to see the text in the chat
networks that connected them. Developed in the             channel that they’re using and to see who else is cur-
early 1980s, Talk allowed for the nearly synchronous       rently in that channel. The most popular client is
exchange of text between two parties; however, un-         mIRC, a windows-compatible client; others include
like its descendents, it displayed text messages as they   Xircon and Pirch.
were written, character by character, rather than as           IRC does not have a central organizing system;
completed messages posted to the discussion all at         organizations like universities and research groups
once. Talk and its sister program Phone fell into dis-     simply run the software on their servers and make it
use after the introduction of the World Wide Web in        available to their users. In the late 1990s, IRC de-
1991 and the introduction of graphical and multi-          centralized architecture contributed to a system
user interfaces.                                           breakdown. In mid-1996, when one IRC server op-
                                                           erator, based in North America, started abusing
                                                           the IRC system, other North American IRC server
  Home computers are being called upon to perform many     operators expelled the abuser; however, when he
  new functions, including the consumption of homework     disconnected his server they discovered that
  formerly eaten by the dog.                               he was also the main link between North American
                                         —Doug Larson      and European IRC networks. After weeks of negoti-
                                                           ations between North American and European
                                                           IRC server operators, who disagreed over the han-
Internet Relay Chat                                        dling of the expulsion, the impasse was not resolved.
Jarkko Oikarinen, a Finnish researcher, developed          While interconnectivity between continents has been
IRC in 1988 based on the older Bulletin Board              restored, the two IRC networks remain separate (IRC
System (BBS). BBSs were central locations where            net and Efnet [Eris Free Net]); they have their own
users could dial in to a central server using a mo-        separate channels and have developed separately.
dem and leave messages and hold discussions on             Other networks, including DALnet and Undernet,
this central server, usually dedicated to a certain        have developed since the separation.
topic or interest group. Oikarinen wrote the IRC
program to allow users to have “real-time” discus-         MUDs
sions not available on the BBS. First implemented          Pavel Curtis, a researcher at Xerox PARC who spe-
on a server at the University of Oulu where Oikarinen      cializes in virtual worlds, gives this definition of a
worked, IRC quickly spread to other Finnish uni-           Multi-User Domain: “A MUD is a software program
versities, and then to universities and ISPs through-      that accepts ‘connections’ from multiple users across
out Scandinavia and then the World.                        some kind of network and provides to each user ac-
    Each “channel” on IRC (the name was taken from         cess to a shared database of ‘rooms’, ‘exits’ and other
the Citizen’s Band radio community) represents a           objects. Each user browses and manipulates this data-
specific topic. Initially each channel was desig-          base from inside one of the rooms. A MUD is a kind
                                                                                                       CHATROOMS ❚❙❘ 91




                                       A Personal Story—Life Online

  In the mid 1990s, I went to visit my first online chatroom as part of a larger project on computer-mediated communi-
  cation. I had no idea what to expect—whether the people would be who they said they were, whether I’d have anything
  in common with other visitors, or what it would be like to interact in a text-based medium. I found myself enjoying the
  experience of talking to people from all over the world and came to spend much time in this virtual community.
       I soon learned that the community was much larger than the chatroom I had visited, connected by telephone, e-mail,
  letters, and occasional face-to-face visits. Over the past five years, I’ve spoken or emailed with many new acquaintances,
  and have had the pleasure of meeting my online friends in person when my travels take me to their part of the country.
       Participation in a virtual community has provided me opportunities to talk in depth with people from around the
  world, including Australia, New Zealand, South America, Mexico, Europe, and even Thailand. The virtual community
  also brings together people from a wide range of socioeconomic backgrounds that might ordinarily never have mixed.
  It’s been fascinating to get to know such a diverse group of individuals.
       My personal experiences in an online community have helped shape my research into the societal dimensions of
  computing and computer-mediated communication. One of my current projects investigates the effects of participation
  in online support communities on people’s mental and physical well-being. In addition, the success with which I’ve been
  able to meet and become acquainted with others using a text-only medium has had a strong impact on my theories about
  how technologies can successfully support remote communication and collaboration.
                                                                                                            Susan R. Fussell




of virtual reality, an electronically represented ‘place’        Web-Based Chat
that users can visit” (Warschauer 1998, 212).                    Web-based chatting became possible after the World
    MUDs provide more virtual places to visit, hang              Wide Web was first released onto the Internet in
out, socialize, play games, teach, and learn than IRC            December 1990, but it didn’t became popular until
or Web-based chatrooms do. Some MUDs have been                   after the release of the Java programming language
used to hold meetings or conferences because they                a few years later. Java allowed developers to create
allow participants to convene without travel hassles—            user-friendly graphical interfaces to chat spaces on
virtual conferences may have different rooms for dif-            websites or ISP portals that could function across
ferent topics and a schedule of events similar to that           different computing and Internet browsing plat-
of a geographically located conference.                          forms. Web-based chatting, like IRC, tends to be
    Two British graduate students, Richard Bartle and            based around common themes, issues, or specific
Roy Trubshaw, developed the first MUD in 1979, as a               discussion topics—it has given rise to rooms like
multiuser text-based networked computer game.                    Love@AOL or sports- or hobby-themed rooms
Other MUDs followed, and new subtypes grew, in-                  like The Runners Room on Yahoo Chats. Other chat
cluding MOOs (Multiuser domains Object Oriented),                spaces may be on an individual website devoted to a
used primarily for educational purposes, and MUSHs               common theme (like the chat on the Atlantic County
(Multi-user Shared Hallucinations). MOOs allow for               Rowing Association site, hosted by ezteams).
greater control because the users of the virtual space
can build objects and spaces as well as contribute text.
Because MUDs are complex virtual environments that               Chatroom Users
users visit to master commands and understand pro-               All the different iterations of chatrooms discussed
tocols, rules, and mores, their use and appeal has been          here have some common elements to help users nav-
limited to a tech-savvy group of users.                          igate and quickly understand how to use the software.
92 ❘❙❚ BERKSHIRE ENCYCLOPEDIA OF HUMAN-COMPUTER INTERACTION



After entering a chatroom, channel, or domain, a           Web-based chat, however, there is the expectation
user is confronted with a screen that is split into two    that users are presenting themselves honestly.
or more parts: One side, usually the left, shows the            Nevertheless, all chat spaces give users the op-
discussion in progress. In another box on the screen       portunity to explore portions of their identity,
is a list of who is logged in to the room. Generally       whether it is by choosing to have the opposite gen-
below these is the box where the user enters text or       der, a different race, or a different set of personal
commands to begin the conversation or move about           experiences, or in the case of some games, by ex-
the space (in the case of MUDs). In some chat spaces,      ploring what it is like to be something other than
users can create their own private chat with a sin-        human. Anonymity or pseudonymity on line gives
gle individual from the chatroom. In some Web-             many users a feeling of freedom and safety that al-
based tools, the chatroom is designed to use an            lows them to explore identities that they dare not
instant messaging program to conduct one-on-               assume in the offline world. Users are separated by
one chats. In others the private chat tool is built-in—    geographic distances so it is unlikely that actions
in MUDs, a user uses the “whisper” command to              taken or phrases uttered will come back to haunt
direct a comment or conversation to a particular in-       them later. And finally, in chat environments with-
dividual, and in some Web-based chats a private chat       out audio or video, communication is mediated by
may be opened in another smaller window in the             the technology so there are none of the cues that can
same chatting interface.                                   make a conversation emotional. All of this leads to
     In a survey in the summer of 2002, the Pew            lower levels of inhibitions, which can either create
Internet & American Life Project found that only one-      greater feelings of friendship and intimacy among
quarter of Internet users had ever visited a chatroom      chat participants or lead to a greater feeling of ten-
or participated in an online discussion, and only 4 per-   sion and lend an argumentative, even combative qual-
cent had visited a chatroom on a typical day. Men          ity to a chat space.
are more likely to use chatrooms than women, as are
those who are less well off; those earning less than
$50,000 a year are much more likely to chat than those     The Future of Chat
earning more. Younger people are also more likely to       In 1991 researchers at Cornell University created
chat, particularly those between eighteen and twenty-      CUSeeMe, the first video chat program to be dis-
nine, although among teens, particularly adolescent        tributed freely online. Video and audio chat did not
girls, chatting is frequently perceived as unsafe.         truly enter mainstream use until the late 1990s, and
Nevertheless, in spite of (or because of) chat’s rep-      with the advent of Apple’s iChat and Microsoft’s im-
utation, 55 percent of young people between twelve         proved chatting programs and web cams, video chat
and seventeen have visited a chatroom.                     utilizing speakers and web cams looks to be the fu-
     Chatrooms have become the favorite play-              ture direction of chatting. Today Yahoo.com and
grounds of many Internet users because they enable         other portal-provided Web-based chatrooms allow
them to assume a character or a role different             audio and video chat in their rooms, though the
from the one they play in their offline life. As social    number of users taking advantage of the technology
psychologist Erving Goffman noted in his 1959 book         is still relatively small. A user’s bandwidth and hard-
Presentation of Self in Everyday Life, we present          ware capabilities are still limiting factors in the use
different images of ourselves to different people, and     of the bandwidth-intensive video chat, but as broad-
some theorists have described chatrooms as                 band Internet connectivity percolates through the
spaces of performance where an identity is “per-           population, the quality of video Web-based chatting
formed” for the audience of other chatters. In cer-        available to most users will improve, and its adop-
tain chatrooms, like MUDs, where gaming or                 tion will undoubtedly become more widespread.
role-playing is often the reason users go there, it is     MUDs and MOOs are also moving into HTML-
expected that visitors do not bear any resem-              based environments, which will make it much eas-
blance to their selves at the keyboard. In IRC and         ier for the average Internet user to adopt them,
                                                                                                           CHILDREN AND THE WEB ❚❙❘ 93



and will perhaps move Multi-User Domains from                                Taylor, T.L. (1999). Life in virtual worlds: Plural existence, multi-
the subculture of academics and devotees into                                   modalities and other online research challenges. American Behavioral
                                                                                Scientist, 4(3).
everyday use.                                                                Turkle, S. (1995). Life on the screen: Identity in the age of the Internet.
                                                                                New York: Simon & Schuster.
                                                 Amanda Lenhart              Warshauer, S. C. (1998). Multi-user environment studies: Defining
                                                                                a field of study and four approaches to the design of multi-user
                                                                                environments. Literary and Linguistic Computing, 13(4).
See also E-mail, MUDs                                                        Young, J. R. (1994). Textuality and cyberspace: MUD’s and written
                                                                                experience. Retrieved July 31, 2003, from http://ftp.game.org/pub/
                                                                                mud/text/research/textuality.txt

FURTHER READING

Bartle, R. (1990). Early MUD History. Retrieved July 31, 2003, from
   http://www.ludd.luth.se/mud/aber/mud-history.html                                           CHILDREN
Bevan, P. (2002). The circadian geography of chat. Paper presented at the
   conference of the Association of Internet Researchers, Maastricht,
   Netherlands.
                                                                                              AND THE WEB
Campbell, J. E. (2004). Getting it on online: Cyberspace, gay male sex-
   uality and embodied identity. Binghamton, NY: The Haworth Press.          Children are among the millions of people who have
Dibbell, J. (1998). A rape in cyberspace. In My tiny life: Crime and pas-    been introduced to new ways of accessing infor-
   sion in a virtual world. Owl Books, chapter 1. Retrieved July 31,         mation on the World Wide Web, which was launched
   2003, from http://www.juliandibbell.com/texts/bungle.html
IRC.Net. IRC net: Our history. Retrieved July 30, 2003, from http://         in 1991 and began to become popular with the adop-
   www.irc.net/                                                              tion of a graphical user interface in 1993. The fact
Hudson, J. M., & Bruckman, A. S. (2002). IRC Francais: The cre-              that the Web utilizes hypertext (content with active
   ation of an Internet-based SLA community. Computer Assisted
   Language Learning, 1(2), 109–134.
                                                                             links to other content) and a graphical user interface
Kendall, L. (2002). Hanging out in the virtual pub: Masculinities and        have made it more congenial and much easier to use
   relationships online. Berkeley: University of California Press.           than earlier menu-driven, text-based interfaces (i.e.,
Lenhart, A., et al. (2001). Teenage life online: The rise of the instant     Gopher, Jughead, Veronica) with the Internet.
   message generation and the Internet’s impact on friendships and
   family relationships. Pew Internet & American Life Project, retrieved
   August 21, 2003, from http://www.pewinternet.org/
Murphy, K. L., & Collins, M. P. (1997). Communication conven-                Children’s Web Use
   tions in instructional electronic chats. First Monday, 11(2).             Children use the Web inside and outside the class-
Pew Internet & American Life Project. (2003). Internet activities (Chart).
   Retrieved July 31, 2003, from http://www.pewinternet.org/reports/         room, and they navigate it to find information for
   index.asp                                                                 both simple and complex projects. They recognize
Pew Internet & American Life Project. (2003). Unpublished data from          the Web as a rich source of up-to-date information,
   June-July 2002 on chatrooms. Author.
Reid, E. M. (1994). Cultural formation in text-based virtual realities.
                                                                             hard-to-find information, and compelling images.
   Unpublished doctoral dissertation, University of Melbourne,               Research by Dania Bilal (2000) and Jinx Watson
   Australia. Retrieved July 31, 2003, from http://www.aluluei.com/          (1998) has revealed that children who use the Web
   cult-form.htm                                                             have a sense of independence, authority, and con-
Rheingold, H. (1993). The virtual community: Homesteading on the
   electronic frontier. Cambridge, MA: MIT Press.
                                                                             trol. They are motivated, challenged, and self-
Rheingold, H. (1998). Building fun online learning communities.              confident. They prefer the Web to print sources due
   Retrieved July 30, 2003, from http://www.rheingold.com/texts/             to the vast amount of information available and their
   education/moose.html                                                      ability to search by keyword and browse subject hi-
Schaap, F. (n.d.). Cyberculture, identity and gender resources (online
   hyperlinked bibliography). Retrieved July 31, 2003, from http://          erarchies quickly. Research conducted for the Pew
   fragment.nl/resources/                                                    Internet & American Life Project revealed that both
Surkan, K. (n.d.). The new technology of electronic text: Hypertext and      parents and children believe that the Internet helps
   CMC in virtual environments. Retrieved July 31, 2003, from http://
   english.cla.umn.edu/GraduateProfiles/Ksurkan/etext/etable.html
                                                                             with learning. While these positive perceptions of
Talk mode (n.d.). The jargon file. Retrieved November 1, 2002, from           the Internet are encouraging, children’s success in
   http://www.tuxedo.org/~esr/jargon/html/entry/talk-mode.html               finding information on the Web is questioned. Given
94 ❘❙❚ BERKSHIRE ENCYCLOPEDIA OF HUMAN-COMPUTER INTERACTION




                        Two of the many books available that
                           educate children on the perils and
                           promise of the Web. Safety on the
                         Internet is geared to ages 6–9, while
                         Cyber Space is meant for ages 9–12.

the Web’s increasing complexity and the abundance          cult vocabulary for elementary schoolchildren to un-
of information available there, it is worth asking how     derstand. Children in that age range preferred sites
well children handle the challenges of using the Web.      with high visual content, animation, and short, sim-
    Researchers from library and information sci-          ple textual content. In 1998 the researchers John
ence, educational psychology, sociology, cognitive         Schacter, Gregory Chung, and Aimee Dorr studied
science, and human-computer interaction have stud-         the effect of types of tasks on the success of fifth and
ied children’s interaction with the Web. In the field       sixth graders in finding information. They found
of information science, researchers have investigated      that children browsed more than they searched by
children’s search strategies, their relative preferences   keyword and performed better on open-ended (com-
for browsing and searching, their successes and fail-      plex) than factual (simple) tasks. By contrast, in 2000
ures, the nature of tasks and success, Web design, and     Terry Sullivan and colleagues found that middle and
children’s navigational skills, relevance judgment,        high school students were more successful on simple
and affective states (feelings, perception, motivation).   tasks than complex ones. Results from Dania Bilal’s
Findings and conclusions from these studies have           research in 2000–2002 echoed Sullivan’s results
begun to provide a rich framework for improving            and revealed that middle school students were most
system design and developing more effective Web            successful on tasks that they chose themselves than
training programs.                                         they were on tasks that were assigned.
    The first study in library and information sci-             In 1999 Andrew Large, Jamshid Beheshti, and
ence appeared in 1997 when Jasmine Kafai and               Haidar Moukdad examined the Web activities of
Marcia Bates examined elementary schoolchildren’s          Canadian sixth graders. These researchers found that
Web literacy skills. They found that children were         children browsed more than they searched by key-
enthusiastic about using the Web and were able to          word, had difficulty finding relevant information,
scroll webpages and use hyperlinks effectively.            and, although they had been given basic Web
However, the researchers perceived that many web-          training, lacked adequate navigational skills. The
sites had too much text to read and too much diffi-         children’s use of the Netscape “Back” command to
                                                                                 CHILDREN AND THE WEB ❚❙❘ 95



return to the previous page, for example, accounted        tion, despite the fact that most of the concepts they
for 90 percent of their total Web moves; they acti-        employed were appropriate. The failure to find re-
vated online search help only once. In fact, frequent      sults can be attributed largely to the poor indexing
use of the “Back” command is common among chil-            of the Yahooligans! database. Overall, the children
dren and young adults. Various studies in the late         took initiative and attempted to counteract their
1990s and early 2000s found similar results. In a          information retrieval problems by browsing subject
follow-up to a 1999 study, Andrew Large and Jamshid        categories. Indeed, they were more successful when
Beheshti (2000) concluded that children valued             they browsed than when they searched by keyword.
the Web for finding information on hard topics, speed            Children’s low success rates on the assigned tasks
of access, and the availability of color images, but       were attributed to their lack of awareness of the
perceived it as more difficult to use than print sources.   difference between simple and complex tasks, es-
Children expressed frustration with information            pecially in regard to the approach to take to fulfill
overload and with judging relevance of the retrieved       the assignment’s requirements. On the complex
results. Information overload and problems deter-          assigned task, for example, children tended to seek
mining relevance seem to be widespread among chil-         specific answers rather than to develop an under-
dren and young adults using the Web; a study of            standing of the information found. On the positive
elementary, middle, and high school students in            side, children were motivated and persistent in us-
England corroborated Large and Beheshti’s finding.          ing the Web. When asked about reasons for their mo-
     Most children assume that the Web is an efficient      tivation and persistence, children cited convenience,
and effective source for all types of information.         challenge, fun, and ease of use. Ease of use was de-
Consequently, they rarely question the accuracy and        scribed as the ability to search by keyword. On the
authority of what they find. If they retrieve results       negative side, children expressed frustration at both
that are close enough to the topic, they may cease to      information overload and the zero retrieval that
pursue their initial inquiry and take what they get at     resulted from keyword searching. Indeed, this fea-
face value.                                                ture was central to most of the search breakdowns
     Most studies focused on using the Web as a whole      children experienced. Although Yahooligans! is de-
and on search engines that are developed for adult         signed for children aged seven through twelve, neither
users rather than children. Bilal has investigated the     its interface design nor its indexing optimized chil-
information-seeking behavior of children who used          dren’s experience. Children’s inadequate knowledge
Yahooligans!, a search engine and directory specif-        of how to use Yahooligans! and their insufficient
ically designed for children aged seven through            knowledge of the research process hindered their
twelve. She found that 50 percent of the middle            success in finding information.
school children were successful on an assigned, fact-
based task, 69 percent were partially successful on
an assigned, research-based task, and 73 percent were      Optimizing the Web for Children
successful on tasks they selected themselves. The flex-     Children’s experiences with the Web can be greatly
ibility children had in topic selection and modifica-       improved by designing Web interfaces that build
tion combined with their satisfaction with the results     on their cognitive developmental level, information-
may have influenced their success rate on the self-         seeking behaviors, and information needs. Since 2002,
selected task. Children were more motivated, stimu-        Bilal (working in the United States) and Large,
lated, and engaged in completing their tasks when          Beheshti, and Tarjin Rahman (working together in
they selected topics of personal interest.                 Canada), have begun projects that involve children
     The children used concrete concepts (selected         in the design of such interfaces. Both groups have
from the search questions) in their searches and, when     concluded that children are articulate about their in-
these concepts failed to generate relevant informa-        formation needs and can be effective design partners.
tion, they utilized abstract ones (synonyms or related     Based on the ten interfaces that children designed for
terms). The children had trouble finding informa-           search engines, Bilal was able to identify the types
96 ❘❙❚ BERKSHIRE ENCYCLOPEDIA OF HUMAN-COMPUTER INTERACTION



of information architecture and visual design chil-         in its use. Children, too, should be taught how to use
dren needed and the information architecture, func-         the Web effectively and efficiently.With critical-thinking
tionality, and visual design they sought.                   skills and an understanding of how to manipulate the
     In sum, both Bilal and the Canadian-based team         Web, children can move from being active explorers
concluded that children are creative searchers who          of the Web to becoming discerning masters of it.
are more successful when they browse than when                   In discussing how usable Web interfaces are for
they search by keyword. Typically, children prefer          children, Jacob Neilsen notes that “existing Web
keyword searching but resort to browsing when they          [interfaces] are based at best by insights gleaned from
experience continued information-retrieval prob-            when designers observe their own children, who are
lems. Children do not take advantage of the search          hardly representative of average kids, typical Internet
features provided in search engines and rarely acti-        skills, or common knowledge about the Web”
vate the help file for guidance.                             (Neilsen 2002, 1). Thus, it is not surprising to find
     The research also revealed that children have both     that children experience difficulty in using the Web.
positive and negative feelings when it comes to the         System developers need to design interfaces that ad-
Web. They associate the Web with motivation, chal-          dress children’s cognitive developmental level,
lenge, convenience, fun, authority, independence,           information needs, and information-seeking be-
and self-control, but also with frustration, dissatis-      haviors. Developing effective Web interfaces for chil-
faction, and disappointment caused by information           dren requires a team effort involving information
overload, lack of success in searches, and inability to     scientists, software engineers, graphic designers, and
make decisions about document relevance.                    educational psychologists, as well as the active par-
     As to information literacy, children possess           ticipation of representative children.
inadequate information-seeking skills, naïve Web                 We have a growing understanding of the
navigational skills, and an insufficient conceptual un-      strengths and weaknesses of the Web as a tool for
derstanding of the research process. These problems         teaching and learning. We also know much about
cry out to teachers and information specialists to pro-     children’s perceptions of and experiences with the
vide more effective Web training and to design in-          Web, as well as their information-seeking behavior
structional strategies that successfully integrate the      on the Web. The rapid progress made in these ar-
Web into effective learning.                                eas of study is commended.
     With regard to system design, it appears that web-          However, research gaps remain to be filled. We do
sites, Web directories, and search engines are not easy     not have sufficient research on working with children
for children to use. Too much text, difficult vocabu-        as partners in designing Web interfaces. We have in-
lary, long screen display, deep subject hierarchies, in-    vestigated children’s information-seeking behavior in
effective help files, poor indexing, and misleading          formal settings, such as schools, to meet instructional
hyperlink titles all hinder children’s successful use.      needs, but with the exception of Debra J. Slone’s 2002
                                                            study, we have little information on children’s Web be-
                                                            havior in informal situations, when they are using it to
Education, Design, and                                      meet social or entertainment needs. We also lack a
                                                            sophisticated model that typifies children’s information-
Future Research                                             seeking behavior.We need to develop a model that more
Use of the Web in school and its increased use at home      fully represents this behavior so that we can predict suc-
does not ensure that children possess effective skills in   cessful and unsuccessful outcomes, diagnose problems,
using it. Information professionals (e.g., school and       and develop more effective solutions.
public librarians) who serve children need to collab-
orate with teachers to identify how the Web can ef-                                                      Dania Bilal
fectively support meaningful learning. Teachers cannot
make the Web an effective learning and research tool        See also Classrooms; Graphical User Interface; Search
unless they first receive effective, structured training     Engines
                                                                                                                            CLASSROOMS ❚❙❘ 97



                                                                                  Washington, DC: Pew Internet and American Life Project. Retrieved
FURTHER READING                                                                   January 4, 2004, from http://www.pewinternet.org/reports/
                                                                                  pdfs/PIP_Schools_Report.pdf
                                                                               Neilsen, J. (2002). Kids’ corner: Website usability for children. Retrieved
Bilal, D. (1998). Children’s search processes in using World Wide Web
                                                                                  Januar y 4, 2004, from http://www.useit.com/aler tbox/
    search engines: An exploratory study. Proceedings of the 61st
                                                                                  20020414.html
    ASIS Annual Meeting, 35, 45–53.
                                                                               Schacter, J., Chung, G. K. W. K., & Dorr, A. (1998). Children’s Internet
Bilal, D. (1999). Web search engines for children: A comparative study
                                                                                  searching on complex problems: Performance and process
    and performance evaluation of Yahooligans!, Ask Jeeves for
                                                                                  analyses. Journal of the American Society for Information Science,
    Kids, and Super Snooper. Proceedings of the 62nd ASIS Annual
                                                                                  49(9), 840–849.
    Meeting, 36, 70–82.
                                                                               Shenton, A. K., & Dixon, P. (2003). A comparison of youngsters’ use
Bilal, D. (2000). Children’s use of the Yahooligans! Web search engine,
                                                                                  of CD-ROM and the Internet as information resources. Journal of
    I: Cognitive, physical, and affective behaviors on fact-based
                                                                                  the American Society for Information Science and Technology, 54(11),
    tasks. Journal of the American Society for Information Science, 51(7),
                                                                                  1049–2003.
    646–665.
                                                                               Slone, D. J. (2002). The influence of mental models and goals on search
Bilal, D. (2001). Children’s use of the Yahooligans! Web search engine,
                                                                                  patterns during Web interaction. Journal of the American Society
    II: Cognitive and physical behaviors on research tasks. Journal of
                                                                                  for Information Science and Technology, 53(13), 1152–1169.
    the American Society for Information Science and Technology, 52(2),
                                                                               Wallace, R. M., Kupperman, J., and Krajcik, J. (2002). Science on the
    118–137.
                                                                                  Web: Students on-line in a sixth-grade classroom. The Journal of
Bilal, D. (2002). Children’s use of the Yahooligans! Web search engine,
                                                                                  the Learning Sciences, 9(1), 75–104.
    III: Cognitive and physical behaviors on fully self-generated tasks.
                                                                               Watson, J. S. (1998). If you don’t have it, you can’t find it: A close look
    Journal of the American Society for Information Science and
                                                                                  at students’ perceptions of using technology. Journal of the American
    Technology, 53(13), 1170–1183.
                                                                                  Society for Information Science, 49(11), 1024–1036.
Bilal, D. (2003). Draw and tell: Children as designers of Web inter-
    faces. Proceedings of the 66th ASIST Annual Meeting, 40,
    135–141.

                                                                                                CLASSROOMS
Bilal, D. (In press). Research on children’s use of the Web. In C. Cole
    & M. Chelton (Eds.), Youth Information Seeking: Theories, Models,
    and Approaches. Lanham, MD: Scarecrow Press.
Druin, A., Bederson, B., Hourcade, J. P., Sherman, L., Revelle, G., Platner,
    M., et al. (2001). Designing a digital library for young children. In      People have regarded electronic technology through-
    Proceedings of the first ACM/IEEE-CS Joint Conference on Digital            out its evolution as an instrument for improving
    Libraries (pp. 398–405). New York: ACM Press.                              learning in classrooms. Television and video were
Fidel, R., Davies, R. K., Douglass, M. H., Holder, J. K., Hopkins, C. J.,
    Kushner, E. J., et al. (1999). A visit to the information mall: Web
                                                                               examples of early electronic technology used in class-
    searching behavior of high school students. Journal of the American        rooms, and now personal computers have shown
    Society for Information Science, 50(1), 24–37.                             how electronic technology can enhance teaching
Hirsh, S. G. (1999). Children’s relevance criteria and information seek-       and learning. Some people have lauded the new
    ing on electronic resources. Journal of the American Society for
    Information Science, 50(14), 1265–1283.                                    kinds of learning activities afforded by electronic
Kafai, Y. B., & Bates, M. J. (1997). Internet Web-searching instruc-           technology; but other people maintain that such
    tion in the elementary classroom: Building a foundation for                technology can be detrimental in classrooms. Despite
    information literacy. School Library Media Quarterly, 25(2),
    103–111.
                                                                               such criticisms, researchers in different fields—
Large, A., & Beheshti, J. (2000). The Web as a classroom resource:             education, computer science, human-computer
    Reactions from the users. Journal of the American Society for              interaction—continue to explore how such tech-
    Information Science And Technology, 51(12), 1069–1080.                     nology, paired with innovative curricula and teacher
Large, A., Beheshti J., & Moukdad, H. (1999). Information seeking on
    the Web: Navigational skills of grade-six primary school students.
                                                                               training, can improve classrooms.
    Proceedings of the 62nd ASIS Annual Meeting, 36, 84–97.
Large, A., Beheshti, J., & Rahman, R. (2002). Design criteria for chil-
    dren’s Web portals: The users speak out. Journal of the American
    Society for Information Science and Technology, 53(2), 79–94.              Early Visions of
Lenhart, A., Rainie, L., Oliver, L. (2003). Teenage life online: The rise of
    the instant-message generation and the Internet’s impact on                Learning Technologies
    friendships and family relationships. Washington, DC: Pew Internet         Early visions of how technology could be applied to
    and American Life Project. Retrieved January 4, 2004, from http://
    www.pewinternet.org/reports/pdfs/PIP_Teens_Report.pdf
                                                                               learning included so-called behaviorist teaching ma-
Lenhart, A., Simon, M., & Graziano, M. (2001). The Internet and ed-            chines inspired by the U.S. psychologist B. F. Skinner
    ucation: Findings of the Pew Internet & American Life Project.             in the 1960s. Skinner believed that classrooms
98 ❘❙❚ BERKSHIRE ENCYCLOPEDIA OF HUMAN-COMPUTER INTERACTION




                                        History Comes Alive in Cyberspace

       OLD DEERFIELD, Mass. (ANS)—On a blustery spring                     computer labs that are being established throughout
       morning, 18 students from Frontier Regional High                    district schools. But as the trip to Old Deerfield demon-
       School made their way down Main Street here in this colo-           strated, students will also add to the pool of knowledge
       nial village, jotting down notes on the Federal and Italianate      and contribute original data to the web site as well.
       architecture and even getting a look at an early 18th-century           “This is not just an electronic test book,” said Tim
       kitchen.                                                            Neumann, executive director of Pocumtuck Valley
            But this was no ordinary field trip. The students              Memorial Association and one of the project’s designers.
       were gathering information for an Internet-based project            “Students are not just surfing the web but actively en-
       that is integrating state-of-the-art computer technology            gaging with the text and images on the screen.” Stu-
       with the social studies curriculum throughout this rural            dents also address questions posed by teachers and then
       western Massachusetts school district.                              conduct research via the Internet, as well as other field
            The project, titled Turns of the Centuries, focuses on         studies, he said.
       life at the turns of the last three centuries, beginning in             Building the web sites, from teachers’ notes and class-
       1700 and continuing through 1900. It’s an unusual                   room lesson plans, are students and technicians at the
       partnership between three distinct entities—a secondary             University of Massachusetts Center for Computer-Based
       school, a university and a museum.                                  Instructional Technology.
            In the project, the primary sources of the Pocumtuck               The students in Friday morning’s expedition were re-
       Valley Memorial Association, a nationally recognized mu-            sponding to an assignment to choose a colonial family and
       seum of frontier life in this region, will be available to stu-     track them over time, using the resources at the museum.
       dents through a web site that teachers, students and                Those results will eventually be incorporated into the Turns
       researchers are putting together.                                   of the Centuries web site, where other students through-
            Central to the project are the over 30,000 museum ar-          out the K-12 district will be able to access them.
       tifacts—diaries, letters and other ‘primary sources’—made               In addition to helping acquaint students with emerg-
       available to students through the developing web site. The          ing technologies, the Turns of the Centuries project in-
       marriage of technology with the museum archives has                 structs teachers how to teach social studies with a web-based
       made possible new opportunities for “inquiry-based” ed-             curriculum, and how to access these resources in their class-
       ucation, which focuses on developing the student as active          rooms, as well as exploring the potential partnerships
       learner.                                                            among school and communities linked by the information
            In essence, the educational project here is a cyberspace       highway.
       version of the museum, enabling students to access archives                                                       Robin Antepara
       through the Internet, either from their homes or through               Source: Students learn about history with classroom computers of to-
                                                                              morrow. American News Service, June 17, 1999.




suffered from a lack of individual attention and that                    Incorrect responses would prevent advancement
individualized instruction would improve learning.                       to the next level of questions, giving students the op-
The idea was that individual students could use a                        portunity to consider how they could correct their
computer that would teach and test them on dif-                          responses. Software adopting this approach is fre-
ferent topics. Students would receive positive rein-                     quently called “drill and practice” software, but few
forcement from the computer through some reward                          examples of such software exist outside of educa-
mechanism (e.g., praise and advancement to the next                      tional games and other kinds of “flash card” pro-
level of instruction) if they gave correct responses.                    grams that teach skills such as spelling and arithmetic.
                                                                                          CLASSROOMS ❚❙❘ 99



     A different vision is found in the work of           concrete design information to guide software de-
Seymour Papert, an MIT professor who has explored         velopers in developing and assessing effective soft-
technology in education since the 1960s, advocating       ware for learning. Many software projects had little
learning theories proposed by the Swiss psycholo-         real grounding in learning theories and the nature
gist Jean Piaget. Papert’s vision uses computers          of children. Thus, for every successful software
as tools that children use for exploratory and con-       project, many others had educational content that
structive activities. Through these activities children   was lacking and classroom use that was less than
create and shape their own understanding of con-          successful. For instance, many software projects in-
cepts. Papert incorporated these ideas in the Logo        volved the development of educational games
programming language, which was intended to               (sometimes called “edutainment” software) whose
let children write programs to create computer graph-     educational content was dubious and whose ini-
ics by exploring deeper concepts, such as the math-       tial appeal to children soon wore off. Other ap-
ematical concepts needed to draw geographical             proaches involved tools such as HyperCard, which
concepts.                                                 allowed laypeople to create software with the hope
     A related vision came from Alan Kay, a renowned      that educators could create software for their
computer science researcher, who proposed the             students. However, although the idea of teacher-
Dynabook concept during the early 1970s. The              created software was well intentioned and although
Dynabook was envisioned as a device similar to to-        teachers have educational knowledge, they lack
day’s laptop computer that children could use in          software design knowledge, again resulting in
information-rich and constructive activities. The         few major successes.
Dynabook would have basic software core func-                 Other issues were contextual. Many early at-
tionality (using the Smalltalk computer language).        tempts at educational computing were techno-
However, children could extend their Dynabook’s           centric and lacked a full understanding of the
functionality through Smalltalk programming. This         support needed in classrooms. Inadequate training
would allow children to create new tools for creative     for busy teachers to use electronic technology can
expression, information gathering, simulation, and        become a large enough burden that teachers simply
so forth by learning not only programming but also        bypass it. Furthermore, technology has been in-
the fundamentals of the underlying content (e.g., to      troduced into the classroom without a good un-
create a music tool, students would need to learn         derstanding by teachers (and sometimes by
musical concepts).                                        researchers developing the technology) of how the
                                                          technology interacts with the classroom curriculum
                                                          and learning goals. Again, technology has little
Initial Attempts at                                       impact if it is not a good fit with the activities that
                                                          teachers desire. Finally, schools have lacked adequate
Technology-Enhanced Classrooms                            technological resources to make full use of tech-
Each technological vision has brought promises of         nology, so disparities in the number of computers
how technology can improve classrooms and learn-          in classrooms and in network connectivity have hin-
ing. With the advent of personal computers, edu-          dered effective use of technology.
cators rushed to place computers in classrooms with
the hope of implementing different visions of learn-
ing technologies. However, many initial attempts          Designing
of technology-enhanced classrooms fell short of their
promise because of technological and contextual           Learner-Centered Technology
issues in classrooms.                                     Simply developing stand-alone technology for class-
     One issue was that although people had some          rooms is not enough. If effective technology-
ideas about what kinds of learning activities and         enhanced classrooms are to become a reality, then
goals computers might support, people had little          designers must design an overall learning system
100 ❘❙❚ BERKSHIRE ENCYCLOPEDIA OF HUMAN-COMPUTER INTERACTION



that integrates three factors: technology, curricu-        enough challenge that learners still work in the mind-
lum, and teacher support and development. During           ful manner needed for real learning to occur.
the last ten to fifteen years designers have developed           Teacher support and development are also key
many learning systems in the classroom by con-             for technology-enhanced classrooms. Teacher sched-
sidering these three factors. Research in many             ules and the classroom environment are busy, and
content areas, such as science education, is shedding      introducing technology into classrooms can make
light on effective technology-enhanced classrooms.         matters more complex for teachers. Teachers need
In such educational approaches technology acts as          support and development to show them how tech-
a cognitive tool to support learners as they engage        nology works, how they can integrate technology
in curricular activities. For example, many educa-         into their classroom activities, and how they can use
tional approaches in science education use an              technology effectively in the classroom.
inquiry-based technique in which students engage
in the same kinds of scientific activity—finding sci-
entific articles, gathering and visualizing scientific       New Visions of
data, building scientific models, and so forth—in
which experts engage. For example, learners can use        Technology-Enhanced Classrooms
software to search digital libraries for informa-          Current classroom technology includes primarily
tion, use handheld computers with probes to gather         desktop-based software. Some software implements
data in different locations, use software to build         “scaffolding” features that support learners by ad-
graphs and scientific models, and so forth. Such tech-      dressing the difficulties they encounter in their learn-
nology should be designed to support learners in           ing activities. For example, one particular software
mindfully engaging in curricular activities so that        feature implementing a scaffolding approach would
learners can meet the learning goals that their teach-     be a visual process map that displays the space of ac-
ers have outlined.                                         tivities that learners should perform (e.g., the activ-
     Given this motivation, the approach for design-       ities in a science investigation) in a way that helps
ing learne-rcentered technologies shifts from sim-         them understand the structure of the activities. Other
ply designing technologies whose hallmark is “ease         classroom software includes intelligent tutoring sys-
of use” to designing technologies that learners can        tems that oversee students as they engage in new ac-
use in new, information-rich activities. Developing        tivity. Intelligent tutoring systems can sense when
learner-centered technologies requires designers to        students have encountered problems or are working
understand the kinds of work that learners should          incorrectly and can provide “just-in-time” advice to
engage in (i.e., curricular activities) and the learn-     help them see their errors and understand their tasks.
ing goals of those learners. Then designers need to             Aside from traditional desktop-based software,
understand the areas where learners may face dif-          researchers are exploring new technology. For ex-
ficulties in performing those kinds of work (e.g.,         ample, handheld computers (e.g., Palm or PocketPC
learners may not know what kinds of activities com-        computers) are becoming more pervasive among
prise a science investigation or how to do those ac-       students. The mobility of handheld computers lets
tivities) so that designers can create support features    students take them to any learning context, not
that address those difficulties. Furthermore, such         just the classroom. Thus, researchers are exploring
support features differ from usability-oriented tra-       how to develop learning tools for handheld com-
ditional software design. Although ease of use is still    puters. An example of such tools is probes that can
important, learner-centered technologies should not        be attached to handhelds for scientific data gather-
necessarily make tasks as easy as possible. Rather, just   ing (e.g., probes to measure oxygen levels in a
as a good teacher guides students toward an an-            stream). Handhelds with wireless networking capa-
swer without giving the answer outright, learner-          bility can be used to gather information (e.g., access
centered technologies must provide enough support          digital libraries) from a range of locations outside of
to make tasks accessible to novice learners but leave      a classroom. Additionally, handhelds can be part of
                                                                                                        CLASSROOMS ❚❙❘ 101




                        A Personal Story—Learning through Multimedia

  When I talk about why I became interested in exploring computers in education, I like to tell a story from my early grad-
  uate school days in the late 1990s. My research group was working in a local Michigan high school using the MediaText
  software they had developed earlier. MediaText was a simple text editor that made it possible to incorporate different
  media objects, such as images, sounds, or video, into the text one was writing.
       In one class, students had been given an assignment to explain a series of physics terms. One particular student some-
  times had difficulty writing, but with MediaText, she could use other media types for her explanations. For example,
  using video clips from the movie Who Framed Roger Rabbit? she explained potential energy with a clip of a cartoon baby
  sitting on top of a stack of cups and saucers, swaying precariously without falling over. Then she explained kinetic ener-
  gy with a clip of the same baby sliding across the floor of a room.
       What struck me was that it was clear from her choice of video clips that she understood those physics concepts. If she
  had been confined to textual explanations, she might not have been able to convey as much understanding. But because
  she had a choice of media types, she was able to successfully convey that she knew those concepts.
       This episode helped me realize how computers could impact learners by giving them a range of different media types
  for self-expression. Now sometimes this story gets me in trouble with people who say that if you give students all these
  alternatives to writing, they’ll never learn to write correctly. I’ll buy that…to a certain extent. But people are diverse—
  they learn differently and they express themselves differently. My response to the naysayers is that effectively incorporat-
  ing different media in software tools isn’t for flash, but to give people different “languages” to learn from and use. By
  offering these alternatives, we open new educational doors, especially for today’s diverse, tech-savvy kids. After all, if one
  student can explain physics terms using a movie about a cartoon rabbit, then multimedia in the classroom is working.
                                                                                                               Chris Quintana




new kinds of learning activities called “participatory             Meeting the Challenge
simulations” in which groups of students can use the               Technology-enhanced classrooms have had fail-
“beaming” feature of wireless handhelds to be part                 ures as researchers have struggled to understand not
of a simulation in which they exchange information.                only the kinds of effective learning technologies, but
For example, students can explore epidemiological                  also the role of technology in classrooms and the
crises in a simulation in which they “meet” other                  support needed for effective technology-enhanced
people by exchanging information with their hand-                  classrooms. Critics of technology in classrooms still
helds. During the simulation a student’s handheld                  exist. Education professor Larry Cuban has written
might announce that it is “sick,” at which point                   extensively on the problems and failures of technol-
students would engage in a discussion to understand                ogy in classrooms. Scientist and author Clifford Stoll
how disease might spread through a community.                      has also written about the possible adverse effects of
    Researchers are also exploring the other end of                technology and the caution that must be taken for
the spectrum, looking at how large displays and vir-               children.
tual reality can be used as learning tools. Such                        However, successes and new visions of how
tools can help students explore virtual worlds and                 technology-enhanced classrooms can support learn-
engage in activities such as “virtual expeditions.”                ers also exist. Designers of learning technologies need
Students can explore environments that might be                    to understand that designing software for ease of use
difficult or impossible to explore in person (e.g., dif-            is not enough. Designers must understand learning
ferent ecosystems), thus allowing them to engage in                theories, the nature of learners, and the classroom
inquiry-based activities throughout a range of loca-               context to design cognitive learning technologies that
tions and gather otherwise inaccessible information.               students use to mindfully engage in substantive
102 ❘❙❚ BERKSHIRE ENCYCLOPEDIA OF HUMAN-COMPUTER INTERACTION



learning activities. People implementing technology-                     puter (PC) explosion, client-server architecture pro-
enhanced classrooms need to consider other issues,                       vides a distributed synthesis of the highly interactive
such as classroom curriculum augmenting technol-                         personal computer (the client) with a remotely
ogy and teachers having the support and develop-                         located computer providing data storage and com-
ment that they need in order to understand and make                      putation (the server). The goal of client-server archi-
full use of technology.                                                  tecture is to create structures and communication
     As new technology arises, people will always                        protocols between the client computer and the server
attempt to see how that technology can be used to                        computer in order to optimize the access to a set of
enhance learning. By understanding the classroom                         computational resources.
context and the local design issues involved in de-
veloping learner-centered technology, the human-
computer interaction community can make                                  Motivating Example
significant contributions to realizing the promise of                     To understand client-server architecture, one can
technology-enhanced classrooms.                                          consider a manufacturing company using computer
                                                                         technology to support day-to-day business opera-
                                                Chris Quintana           tions and long-range strategic planning. Product
                                                                         orders come from the sales department, inventory is
See also Children and the Internet; Psychology                           maintained by the manufacturing department,
and HCI                                                                  and the raw materials orders are generated by the
                                                                         planning department. Furthermore, the accounting
                                                                         department tracks the money, and the chief execu-
FURTHER READING                                                          tive officer (CEO) wants a perspective on all aspects
                                                                         of the company.
Bransford, J. D., Brown, A. L., & Cocking, R. R. (Eds.). (2000). How         To be judged successful, the software solution
   people learn: Brain, mind, experience, and school (Exp. ed.).
   Washington, DC: National Academy Press.
                                                                         implemented should provide data storage and up-
Cuban, L. (1986). Teachers and machines: The classroom use of tech-      date capability for all aspects of the company oper-
   nology since 1920. New York: Teachers College Press.                  ation. Further, the appropriate database segments
Kay, A., & Goldberg, A. (1977). Personal dynamic media. IEEE             should be accessible by all of the employees based
   Computer, 10(3), 31–41.
Papert, S. (1980). Mindstorms. New York: Basic Books.                    on their particular job responsibility, regardless of
Quintana, C., Soloway, E., & Krajcik, J. (2003). Issues and approaches   where they are physically located. Finally, the ap-
   for developing learner-centered technology. In M. Zelkowitz (Ed.),    plication views of the database should be highly
   Advances in computers: Volume 57. Information Repositories
   (pp. 272–323). New York: Academic Press.
                                                                         usable, interactive, and easy to build and update to
Reiser, B. J. (2002). Why scaffolding should sometimes make tasks        reflect ongoing business growth and development.
   more difficult for learners. Proceedings of CSCL 2002, 255–264.
Soloway, E., Guzdial, M., & Hay, K. E. (1994). Learner-centered de-
   sign: The challenge for HCI in the 21st century. Interactions,
   1(2), 36–48.
                                                                         Conflicting Goals
                                                                         One key feature of any software application is the
                                                                         database, the dynamic state of the application. For
                                                                         example, the status of inventory and orders for a fac-
                                                                         tory would be maintained in a database management
            CLIENT-SERVER                                                system (DBMS). Modern database management tech-
                                                                         nology is quite well developed, supporting database
            ARCHITECTURE                                                 lookup and update in a secure, high performance
                                                                         fashion. DBMS computers, therefore, are typically
Client-server architecture is one of the many ways                       high-performance, focused on the task, and have
to structure networked computer software.                                large permanent storage capacity (disk) and large
Developed during the 1980s out of the personal com-                      working memory. The cost of this hardware, the crit-
                                                                          CLIENT-SERVER ARCHITECTURE ❚❙❘ 103



ical need for consistency, and the complexity of sys-        Common Object Request Broker Architecture
tem management dictate that the DBMS be centrally            (CORBA) or the Component Object Model (COM).
located and administered. This goal was realized in               Returning to our motivating example, the soft-
the mainframe architecture of the 1960s and the              ware solution would include a separate interactive
time-sharing architecture of the 1970s.                      PC application designed for each business function:
     On the other hand, personal computer appli-             sales manufacturing, accounting, planning, and
cations such as the spreadsheet program VisiCalc,            the CEO. Each of these individual PC applications
introduced in 1979, demonstrate the power of highly          would use an RPC call for each query or update
interactive human-computer interfaces. Responding            operation to the company database server. This par-
instantly to a user’s every keystroke and displaying         titioning of function is effective both in terms of
results using graphics as well as text, the PC has           hardware cost performance (relatively inexpensive
widened the scope and number of users whose pro-             client computers for each user versus a relatively ex-
ductivity would be enhanced by access to comput-             pensive database server computer shared between
ing. These inexpensive computers bring processing            all users) and end-user application design.
directly to the users but do not provide the same scal-           As the number of simultaneous users grows,
able, high-performance data-storage capability of            the portion of a server’s computation time spent man-
the DBMS. Furthermore, the goal of information               aging client-server sessions grows as well. To mitigate
management and security is counter to the personal           this processing overhead, it is useful to introduce
computer architecture, in which each user operates           an intermediary server to help handle the client-server
on a local copy of the database.                             requests. Called a “message queuing server,” this soft-
     The network—the tie that binds together the             ware system accepts operations to be performed on
DBMS and the human-computer interface—has                    the database and manages the request queues asyn-
evolved from proprietary system networks, such as            chronously. Priority information allows intelligent
IBM System Network Architecture (SNA), intro-                management and scheduling of the operations. Result
duced in 1974, to local area networks, such as               queues, returning answers back to the requesting
Ethernet, developed at Xerox’s Palo Alto Research            client, provide for asynchronous delivery in the other
Center (PARC) and introduced in 1976, to the                 direction as well. Through a message server the queu-
Internet, which began as the U.S. Department of              ing operations are offloaded from the database server,
Defense’s Advanced Research Projects Agency net-             providing enhanced throughput (output). The mes-
work (Arpanet) in 1972 and continues to evolve.              sage server also leads to increased flexibility be-
A networking infrastructure allows client software,          cause the message queuing provides a layer of
operating on a PC, to make requests of the server for        translation and independence between the client soft-
operations on the user’s behalf. In other words, the         ware and the DBMS server.
network provides for the best of both worlds:
high-performance, high-reliability components pro-
viding centralized data computation and user in-             Business Processes
terface components located on the personal computer          Although PC-client access to a server-based DBMS was
providing high interactivity and thereby enhanced            an early client-server scenario and continues to be im-
usability.                                                   portant, other client-server architectures include other
     Furthermore, by defining standardized message-           types of network services. For example, an application
passing protocols for expressing the requests from           server hosts computation rather than data storage, as
client to server, a level of interoperability is achieved.   with a DBMS server. The business processes for an en-
Clients and servers coming from different vendors            terprise may be implemented using an application
or implementing different applications may com-              server. Like the message queuing server, the application
municate effectively using protocols such as Remote          server sits between the client software and the DBMS,
Procedure Call (RPC) or Standard Query Language              encapsulating functions that may be common across
(SQL), together with binding services such as the            many clients, such as policies and procedures.
104 ❘❙❚ BERKSHIRE ENCYCLOPEDIA OF HUMAN-COMPUTER INTERACTION



The Future of                                                                                        based system. It is crucial to design systems for
                                                                                                     ease of learning, because people generally learn to
Client-Server Computing                                                                              use new computer-based systems by exploration.
Client-server computing will continue to be impor-                                                   People resort to reading manuals, using help systems,
tant long into the future. PCs continue to drop in price,                                            or taking formal training only when they have
and new networked devices such as personal data as-                                                  been unsuccessful in learning to do their tasks by ex-
sistants (PDAs) and the World Wide Web are driving                                                   ploration. CW has been applied to a wide variety
network accessibility to a broader audience. The client-                                             of systems, including automatic teller machines
server architecture, which lives on the network through                                              (ATMs), telephone message and call forwarding sys-
standardized messaging protocols, will continue to                                                   tems, websites, computerized patient-record systems
have wide applicability, especially in business.                                                     for physicians, programming languages, multimedia
                                                                                                     authoring tools, and computer-supported coopera-
                                                                         Mark R. Laff                tive work systems. HCI researcher Andrew J. Ko and
                                                                                                     his associates innovatively applied CW (in lieu of pi-
See also Peer-to-Peer Architecture                                                                   lot experiments) to predict problems that experi-
                                                                                                     mental participants might have with the instructions,
                                                                                                     procedures, materials, and interfaces used in ex-
FURTHER READING                                                                                      periments for testing the usability of a system (the
                                                                                                     system was a visual programming language).
Berson, A. (1992). Client/server architecture. New York: McGraw-Hill.
Berson, A. (1995). Sybase and client/server computing. New York:
    McGraw-Hill.
Comer, D. (1994). Internetworking with TCP/IP: Vol. 3. Client-server                                 Cognitive Walkthrough
    programming and applications. Englewood Cliffs, NJ: Prentice Hall.
C o r b i n , J. R . ( 1 9 9 1 ) . T h e a r t o f d i s t r i b u t e d a p p l i c a t i o n s :   Methodology
    Programming techniques for remote procedure calls. New York:                                     The CW approach was invented in 1990 and has
    Springer-Verlag.
Edelstein, H. (1994). Unraveling client/server architecture. Redwood
                                                                                                     evolved into a cluster of similar methods with the
    City, CA: M & T Publishing.                                                                      following four defining features:
Hall, C. (1994). Technical foundations of client/server systems. New
    York: Wiley.                                                                                       1. The evaluation centers on particular users and
IBM Corporation. (2002). Websphere MQ application message inter-                                          their key tasks. Evaluators start a CW by care-
    face. (SC34-6065-00). Armonk, NY: International Business                                              fully analyzing the distinctive characteristics of
    Machines Corporation.
Krantz, S. R. (1995). Real world client server: Learn how to success-
                                                                                                          a particular user group, especially the relevant
    fully migrate to client/server computing from someone who’s actu-                                     kinds of background knowledge these users can
    ally done it. Gulf Breeze, FL: Maximum Press.                                                         call upon when learning to perform tasks on
Metcalfe, R. M., & Boggs, D. R. (1976). Ethernet: Distributed packet                                      the system. Next, CW evaluators select a set of
    switching for local computer networks. Communications of the
    ACM, 19(5), 395–404.
                                                                                                          key tasks that members of the user group will
Sims, O. (1994). Business objects: Delivering cooperative objects for client-                             do on the system. Key tasks are tasks users do
    server. New York: McGraw-Hill.                                                                        frequently, tasks that are critical even if done
                                                                                                          infrequently, and tasks that exhibit the core ca-
                                                                                                          pabilities of the system.
                                                                                                       2. The steps designers prescribe for doing tasks are
                   COGNITIVE                                                                              evaluated. For each key task, CW evaluators
                                                                                                          record the full sequence of actions necessary to
                  WALKTHROUGH                                                                             do the task on the current version of the sys-
                                                                                                          tem. Then CW evaluators walk through the
The cognitive walkthrough (CW) is a usability eval-                                                       steps, simulating users’ action selections and
uation approach that predicts how easy it will be for                                                     mental processes while doing the task. The sim-
people to learn to do particular tasks on a computer-                                                     plest CW version asks two questions at each
                                                                          COGNITIVE WALKTHROUGH ❚❙❘ 105



   step: (1) Is it likely that these particular users        designers can readily learn CW, but they have
   will take the “right action”—meaning the ac-              a shallower grasp of the underlying theory than
   tion designers expect them to take—at this step?          usability experts trained in cognitive psychol-
   and (2) If these particular users do the “right           ogy and consequently find less than half as many
   action” and get the feedback the system pro-              usability problems. A group CW, including at
   vides (if any), will they know they made a good           least one usability expert trained in cognitive
   choice and realize that their action brought              psychology, can find a higher percentage of us-
   them closer to accomplishing their goal? To an-           ability problems than an individual evaluator—
   swer each question evaluators tell a believable           up to 50 percent of the problems that appear in
   success story or failure story. They record fail-         usability tests of the system.
   ure stories and have the option of adding sug-
                                                            CW was one of the several evaluation methods
   gestions for how to repair the problems and
                                                        pioneered in the early 1990s to meet a practical need,
   turn failures into successes. Anchoring the eval-
                                                        the need to identify and repair usability problems
   uation to the steps specified by designers com-
                                                        early and repeatedly during the product develop-
   municates feedback to designers in their own
                                                        ment cycle. The cost of repairing usability problems
   terms, facilitating design modifications that re-
                                                        rises steeply as software engineers invest more time
   pair the usability problems.
                                                        in building the actual system, so it is important to
3. Evaluators use theory-based, empirically verified
                                                        catch and fix problems as early as possible. For a
   predictions. The foundation for CW is a theory
                                                        product nearing completion the best evaluation
   of learning by exploration that is supported by
                                                        method is usability testing with end users (the people
   extensive research done from the 1960s to the
                                                        who will actually use the system), but CW is ap-
   1980s on how people attempt to solve novel
                                                        propriate whenever it is not possible to do usabil-
   problems when they lack expert knowledge
                                                        ity testing. Early versions of CW were tedious to
   or specific training. According to this theory,
                                                        perform, but the 1992 cognitive jogthrough and
   learning to do tasks on a computer-based sys-
                                                        streamlined CW of 2000, which still preserve all
   tem requires people to solve novel problems by
                                                        the essential CW features, are much quicker to
   using general problem-solving methods, gen-
                                                        perform.
   eral reading knowledge, and accumulated ex-
   perience with computers. “The key idea is
   that correct actions are chosen based on their       Transforming CW to Faster and
   perceived similarity to the user’s current goal”
   (Wharton et al. 1994, 126). For software appli-      More Accurately Predict User Actions
   cations, the theory predicts that a user scans       The cognitive walkthrough for the Web (CWW) has
   available menu item labels on the computer           transformed the CW approach by relying on Latent
   screen and picks the menu item label that is         Semantic Analysis (LSA)—instead of on the subjec-
   most similar in meaning to the user’s current        tive judgments of usability experts and software en-
   goal. CW evaluators answer the first question         gineers—to predict whether users are likely to select
   with a success story if the “right action” desig-    the “right action.” LSA is a computer software sys-
   nated by the designer is highly similar in mean-     tem that objectively measures semantic similarity—
   ing to the user’s goal and if the menu item labels   similarity in meaning—between any two passages of
   on the screen use words familiar to the user.        text. LSA also assesses how familiar words and phrases
4. Software engineers can easily learn how to           are for particular user groups.
   make CW evaluations. It is crucial to involve            While analyzing the distinctive characteristics of
   software engineers and designers in CW, be-          the particular user group, CWW evaluators choose
   cause they are the individuals responsible for       the LSA semantic space that best represents the back-
   revising the design to repair the problems. There    ground knowledge of the particular user group—
   is strong evidence that software engineers and       the space built from documents that these users
106 ❘❙❚ BERKSHIRE ENCYCLOPEDIA OF HUMAN-COMPUTER INTERACTION



are likely to have read. For example, CWW currently        about the experimental design and statistics of these
offers a college-level space for French and five spaces     studies.
that accurately represent general reading knowledge            Relying on LSA opens the door to fully auto-
for English at college level and at third-, sixth-,        mating CWW and increasing its cost-effectiveness.
ninth-, and twelfth-grade levels.                          If other CW methods start to rely on LSA they,
    CWW uses LSA to measure the semantic simi-             too, could be automated. The streamlined CW is
larity between a user’s information search goal            more efficient than earlier CW methods, but it still
(described in 100 to 200 words) and the text labels        consumes the time of multiple analysts and relies on
for each and every subregion of the web page and           subjective judgments of uncertain accuracy.
for each and every link appearing on a web page.
CWW then ranks all the subregions and link labels
in order of decreasing similarity to the user’s goal.      Objectively Predicting Actions for
CWW predicts success if the “right action” is the
highest-ranking link, if that link is nested within the    Diverse Users
highest-ranking subregion, and if the “right action”       Relying on LSA makes it possible for CWW to do
link label and subregion avoid using words that are        something that even usability experts trained in
liable to be unfamiliar to members of the user group.      cognitive psychology can almost never do: objec-
    Relying on LSA produces the same objective an-         tively predict action selections for user groups whose
swer every time, and laboratory experiments con-           background knowledge is very different from
firm that actual users almost always encounter serious      the background knowledge of the human evaluators.
problems whenever CWW predicts that users will             For example, selecting the sixth-grade semantic space
have problems doing a particular task. Furthermore,        enables LSA to “think” like a sixth grader, because
using CWW to repair the problems produces two-             the sixth-grade LSA semantic space contains only
to-one gains in user performance. So far, CWW re-          documents likely to have been read by people who
searchers have tested predictions and repairs only         have a sixth-grade education. In contrast, a college-
for users with college-level reading knowledge of          educated analyst cannot forget the words, skills, and
English, but they expect to prove that CWW gives           technical terms learned since sixth grade and can-
comparably accurate predictions for other user             not, therefore, think like a sixth grader.
groups and semantic spaces.                                     Since CW was invented in 1990, the number and
                                                           diversity of people using computers and the Internet
                                                           have multiplied rapidly. Relying on LSA will enable
  APPLICATION A software program that performs a ma-       the CW approach to keep pace with these changes.
  jor computing function (such as word processing or Web   In cases where none of the existing LSA semantic
  browsing).                                               spaces offers a close match with the background
                                                           knowledge of the target user group, new semantic
                                                           spaces can be constructed for CWW (and potentially
    Research by cognitive psychologist Rodolfo Soto        for CW) analyses—in any language at any level of
suggests that CW evaluations of software applica-          ability in that language. Specialized semantic spaces
tions would be improved by relying on LSA, but to          can also be created for bilingual and ethnic minor-
date CW has consistently relied on subjective judg-        ity user groups and user groups with advanced back-
ments of human evaluators. Consequently the agree-         ground knowledge in a specific domain, such as
ment between any two CW evaluators is typically            the domain of medicine for evaluating systems used
low, raising concerns about the accuracy of CW pre-        by health professionals.
dictions. Many studies have tried to assess the accu-
racy and cost-effectiveness of CW compared to                                        Marilyn Hughes Blackmon
usability testing and other evaluation methods. The
results are inconclusive, because there is controversy     See also Errors in Interactive Behavior; User Modeling
                                                                                                             COLLABORATORIES ❚❙❘ 107



                                                                            Spencer, R. (2000). The streamlined cognitive walkthrough method,
FURTHER READING                                                                working around social constraints encountered in a software de-
                                                                               velopment company. In CHI 2000: Proceedings of the Conference
                                                                               on Human Factors in Computing Systems, 353–359.
Blackmon, M. H., Kitajima, M., & Polson, P. G. (2003). Repairing
                                                                            Wharton, C., Rieman, J., Lewis, C., & Polson, P. (1994). The cognitive
   usability problems identified by the cognitive walkthrough for the
                                                                               walkthrough method: A practitioner’s guide. In J. Nielsen &
   web. In CHI 2003: Proceedings of the Conference on Human Factors
                                                                               R. L. Mack (Eds.), Usability inspection methods (pp. 105–140). New
   in Computing Systems, 497–504.
                                                                               York: Wiley.
Blackmon, M. H., Polson, P. G., Kitajima, M., & Lewis, C. (2002).
   Cognitive walkthrough for the Web. In CHI 2002: Proceedings of
   the Conference on Human Factors in Computing Systems, 463–470.
Desurvire, H. W. (1994). Faster, cheaper!! Are usability inspection

                                                                                        COLLABORATIVE
   methods as effective as empirical testing? In J. Nielsen & R. L. Mack
   (Eds.), Usability inspection methods (pp. 173–202). New York: Wiley.
Gray, W. D., & Salzman, M. D. (1998). Damaged merchandise? A re-
   view of experiments that compare usability evaluation methods.
   Human-Computer Interaction, 13(3), 203–261.                                            INTERFACE
Hertzum, M., & Jacobsen, N. E. (2003). The evaluator effect: A chill-
   ing fact about usability evaluation methods. International Journal       See Multiuser Interfaces
   of Human Computer Interaction, 15(1), 183–204.
John B. E., & Marks, S. J. (1997). Tracking the effectiveness of us-
   ability evaluation methods. Behaviour & Information Technology,
   16(4/5), 188–202.
John, B. E., & Mashyna, M. M. (1997). Evaluating a multimedia au-
   thoring tool. Journal of the American Society for Information Science,
                                                                                    COLLABORATORIES
   48(11), 1004–1022.
Ko, A. J., Burnett, M. M., Green, T. R. G., Rothermel, K. J., & Cook,       A collaboratory is a geographically dispersed or-
   C. R. (2002). Improving the design of visual programming lan-            ganization that brings together scientists, instru-
   guage experiments using cognitive walkthroughs. Journal of Visual
   Languages and Computing, 13, 517–544.
                                                                            mentation, and data to facilitate scientific research.
Kushniruk, A. W., Kaufman, D. R., Patel, V. L., Lévesque, Y., &             In particular, it supports rich and recurring hu-
   Lottin, P. (1996). Assessment of a computerized patient record sys-      man interaction oriented to a common research area
   tem: A cognitive approach to evaluating medical technology. M D          and provides access to the data sources, artifacts, and
   Computing, 13(5), 406–415.
Lewis, C., Polson, P., Wharton, C., & Rieman, J. (1990). Testing a walk-
                                                                            tools required to accomplish research tasks. Collab-
   through methodology for theory-based design of walk-up-and-              oratories have been made possible by new com-
   use interfaces. In CHI ‘90: Proceedings of the Conference on Human       munication and computational tools that enable
   Factors in Computing Systems, 235–242.                                   more flexible and ambitious collaborations. Such
Lewis, C., & Wharton, C. (1997). Cognitive walkthroughs. In
   M. Helander, T. K. Landauer, & P. Prabhu (Eds.), Handbook of             collaborations are increasingly necessary. As science
   human-computer interaction (2nd ed., revised, pp. 717–732).              progresses, the unsolved problems become more
   Amsterdam: Elsevier.                                                     complex, the need for expensive instrumentation in-
Pinelle, D., & Gutwin, C. (2002). Groupware walkthrough: Adding
   context to groupware usability evaluation. In CHI 2002: Proceedings
                                                                            creases, larger data sets are required, and a wider
   of the Conference on Human Factors in Computing Systems, 455–462.        range of expertise is needed. For instance, in high-
Polson, P., Lewis, C., Rieman, J., & Wharton, C. (1992). Cognitive walk-    energy physics, the next generation of accelerators
   throughs: A method for theory-based evaluation of user interfaces.       will require vast international collaborations and will
   International Journal of Man-Machine Studies, 36, 741–773.
Rowley, D. E., & Rhoades, D. G. (1992). The cognitive jogthrough: A
                                                                            have a collaboratory model for remote access. At least
   fast-paced user interface evaluation procedure. In CHI ’92:              150 collaboratories representing almost all areas of
   Proceedings of the Conference on Human Factors in Computing              science have appeared since the mid-1980s.
   Systems, 389–395.                                                            Collaboratories offer their participants a num-
Sears, A., & Hess, D. J. (1999). Cognitive walkthroughs: Understanding
   the effect of task description detail on evaluator performance.          ber of different capabilities that fall into five broad
   International Journal of Human-Computer Interaction, 11(3),              categories: communication (including tools such as
   185–200.                                                                 audio or video conferencing, chat, or instant mes-
Soto, R. (1999). Learning and performing by exploration: Label qual-
   ity measured by Latent Semantic Analysis. In CHI ’99: Proceedings
                                                                            saging), coordination (including tools relating to
   of the Conference on Human Factors and Computing Systems,                access rights, group calendaring, and project man-
   418–425.                                                                 agement), information access (including tools for
108 ❘❙❚ BERKSHIRE ENCYCLOPEDIA OF HUMAN-COMPUTER INTERACTION



accessing online databases, digital libraries, and doc-     As the frontiers of science are pushed back, the in-
ument repositories), computational access (includ-          strumentation required for advances becomes more
ing access to supercomputers), and facility access          and more esoteric, and therefore usually more and
(including tools for remotely accessing specialized         more expensive. Alternatively, certain scientific in-
facilities or instruments, such as a particle acceler-      vestigations require instrumentation in specific geo-
ator or a high-powered microscope).                         graphic settings, such as an isolated or inhospitable
     Research on collaboratories has focused mostly         area. A typical example is the Keck Observatory,
on solving technical problems. However, substantial         which provides access to an astronomical observa-
gains in the practice of science are likely to be the       tory on the summit of Mauna Kea in Hawaii to a
combined effect of social and technical transforma-         consortium of California universities.
tions. The gap between the raw performance capa-
bility of collaboratory tools (based on bandwidth,          Community Data System
storage capacity, processor speed, and so forth)            An especially common collaboratory type is one in
and the realized performance (usage for scientific           which a geographically dispersed community agrees
purposes, which is limited by factors such as usabil-       to share their data through a federated or central-
ity and fit to the work and culture) can limit the po-       ized repository. The goal is to create a more power-
tential of collaboratories. This point will be discussed    ful data set on which more sophisticated or powerful
in greater detail later.                                    analyses can be done than would be possible if the
                                                            parts of the data set were kept separately. A typical
                                                            example of a community data system is the Zebrafish
Types of Collaboratories:                                   Information Network (ZFIN), an online aggregation
                                                            of genetic, anatomical, and methodological infor-
Research-Focused Collaboratories                            mation for zebra fish researchers.
There are a number of different kinds of collabora-
tories. A collaboratory that satisfies all elements of the   Open-Community Contribution System
definition given above is a prototypical collaboratory—      Open-community contribution systems are an emerg-
a distributed research center. Other kinds of collabo-      ing organizational type known as a voluntary asso-
ratories are missing one or more of the elements of         ciation. Interested members of a community (usually
that definition. The following four types of collabo-        defined quite broadly) are able to make small con-
ratories focus on enabling geographically distributed       tributions (the business scholar Lee Sproull calls them
research.                                                   microcontributions) to some larger enterprise. These
                                                            contributions are judged by a central approval or-
Distributed Research Center                                 ganization and placed into a growing repository. The
This type of collaboratory functions like a full-fledged     classic example is open-source software development,
research center or laboratory, but its users are geo-       which involves hundreds or even thousands of con-
graphically dispersed—that is, they are not located         tributors offering bug fixes or feature extensions to a
at the research center. It has a specific area of inter-     software system. In science, such schemes are used to
est and a general mission, with a number of specific         gather data from a large number of contributors. Two
projects. A good example of a distributed research          examples will help illustrate this. The NASA Ames
center is the Alliance for Cellular Signaling, a large,     Clickworkers project invited members of the public
complex distributed organization of universities            to help with the identification of craters on images
whose goal is to understand how cells communicate           from a Viking mission to Mars. They received 1.9 mil-
with one another to make an organism work.                  lion crater markings from over 85,000 contribu-
                                                            tors, and the averaged results of these community
Shared Instrument                                           contributions were equivalent in quality to those of
A shared-instrument collaboratory provides access           expert geologists. A second example is MIT’s Open
to specialized or geographically remote facilities.         Mind Common Sense Initiative, which is collecting
                                                                                      COLLABORATORIES ❚❙❘ 109



examples of commonsense knowledge from mem-                collaboratory is successful. What follow are some of
bers of the public “to help make computers smarter”        the most important factors.
(Singh n.d.).
                                                           Readiness for Collaboration
                                                           Participants must be ready and willing to collaborate.
Types of Collaboratories:                                  Science is by its very nature a delicate balance of
                                                           cooperation and competition. Successful collabora-
Practice-Focused Collaboratories                           tions require cooperation, but collaboration is very
The next two collaboratory types support the pro-          difficult and requires extra effort and motivation.
fessional practice of science more broadly, as op-         Technologies that support collaboration will not be
posed to supporting the conduct of research itself.        used if the participants are not ready or willing to col-
                                                           laborate. Various fields or user communities have
Virtual Community of Practice                              quite different traditions of sharing. For instance,
This is a network of individuals who share a research      upper-atmospheric physicists have had a long tra-
area of interest and seek to share news of profes-         dition of collaboration; the Upper Atmospheric
sional interest, advice, job opportunities, practical      Research Collaboratory (UARC) began with a col-
tips on methods, and the like. A good example of           laborative set of users. On the other hand, several ef-
this kind of collaboratory is Ocean US, which              forts to build collaboratories for biomedical research
supports a broad community of researchers inter-           communities (for instance, for researchers studying
ested in ocean observations. A listserv is another         HIV/AIDS or depression) have had difficulty in part
mechanism that is used to support a virtual com-           because of the competitive atmosphere. Readiness
munity of practice, but much more common                   for collaboration can be an especially important fac-
these days are websites and wikis.                         tor when the collaboratory initiative comes from
                                                           an external source, such as a funding agency.
Virtual Learning Community
This type of collaboratory focuses on learning that        Technical Readiness
is relevant to research, but not research itself. A good   The participants, the supporting infrastructure, and
example is the Ecological Circuitry Collaboratory,         the design of the tools must be at a threshold tech-
whose goal is to train doctoral students in ecology        nical level. Some communities are sufficiently col-
in quantitative-modeling methods.                          laborative to be good candidates for a successful
                                                           collaboratory, but their experience with collaborative
                                                           technologies or the supporting infrastructure is not
Evolution and                                              sufficient. Technical readiness can be of three kinds.

Success of Collaboratories                                 INDIVIDUAL TECHNICAL READINESS       People in various
Collaboratories that last more than a year or two tend     organizations or fields have different levels of ex-
to evolve. For example, a collaboratory may start          perience with collaboration tools. A specific new
as a shared-instrument collaboratory. Those who            technology such as application sharing may be a leap
share the instrument may add a shared database com-        for some and an easy step for others. It is impor-
ponent to it, moving the collaboratory toward a            tant to take account of users’ specific experience when
community data system. Then users may add com-             introducing new tools.
munication and collaboration tools so they can plan
experiments or data analyses, making the collabo-          INFRASTRUCTURE READINESS     Collaborative technolo-
ratory more like a distributed research center.            gies require good infrastructure, both technical and
    Some collaboratories are quite successful, while       social. Poor networks, incompatible workstations, or
others do not seem to work very well. There are a          a lack of control over different versions of software
number of factors that influence whether or not a           can cause major problems. It is also very important
110 ❘❙❚ BERKSHIRE ENCYCLOPEDIA OF HUMAN-COMPUTER INTERACTION



to have good technical support personnel, especially        and also created for the community a spirit of gen-
in the early phases of a collaboratory. The Worm            erosity and collaboration. Although goodwill among
Community System (WCS) was a very early collab-             the community of researchers has been a sufficient
oratory project, intended to support a community of         incentive for participation, ZFIN is now expanding
researchers who studied the organism c. elegans (a          its participation beyond its founders, and it will be in-
type of nematode). Sophisticated software was de-           teresting to see how successful the goodwill incentive
veloped for the WCS on a UNIX platform that was             is in the context of the expanded community.
not commonly used in the laboratories of the sci-
entists. Since the tools were thus not integrated           GOODWILL PLUS KARMA POINTS     Slashdot is a very large
with everyday practice, they were seldom used.              and active community of open-source software
Furthermore, the necessary technical support was not        developers who share and discuss news. Slashdot re-
generally present in the lab, so when there were prob-      wards those who make the most informative con-
lems, they were showstoppers.                               tributions by bringing them more into the center of
                                                            attention and allocating them karma points. Karma
SOCIAL ERGONOMICS OF TOOLS         The social interac-      points are allocated in accordance with how highly
tions that take place in teams are affected both by the     a contributor’s postings are rated by others. These
characteristics of team members and by the tools            karma points give contributors some additional priv-
that are used. The study of the impact of technology        ileges on the site, but their main value is as a tangi-
characteristics on this process may be called social        ble measure of community participation and status.
ergonomics (ergonomics is the application of knowl-         Karma points are a formalization of goodwill, valu-
edge about humans to the design of things). For ex-         able primarily because the members of the com-
ample, video conferencing systems often ignore such         munity value them as an indicator of the quality of
details as screen size, display arrangement in relation     the sharing done by specific individuals.
to participants, camera angle, and sound volume.
But it turns out that these details can have social         REQUIRING CONTRIBUTION AS A PREREQUISITE FOR OTHER
effects. For example, a study conducted by the re-          ACTIVITY  In order to get the details of gene sequences
searchers Wei Huang, Judith Olson, and Gary Olson           out of published articles in journals, a consortium
found that the apparent height of videoconference           of high-prestige journals in biology requires that
participants, as conveyed via camera angle, influ-          those who submit articles to the consortium’s
enced a negotiation task. The apparently taller per-        journals have a GenBank accession number indi-
son was more influential in shaping the final outcome         cating that they have stored their gene sequences
than the apparently shorter person.                         in the shared database.

Aligned Incentives                                          NEW FORMS OF PUBLICATION      The Alliance for Cellular
Aligning individual and organizational incentives is        Signaling has taken a novel approach to providing
an important element of successful collaborations.          researchers with an incentive to contribute molecule
Consider the incentives to participate in a commu-          pages to the Alliance’s database. Because the mole-
nity data system: What motivates a researcher to con-       cule pages represent a lot of work, the Alliance has
tribute data to a shared database? By contributing,         worked out an agreement with Nature, one of the
the researcher gives up exclusive access to the data        high-prestige journals in the field, to count a mol-
he or she has collected. There are a variety of incentive   ecule page as a publication in Nature. Nature
schemes for encouraging researchers to collaborate.         coordinates the peer reviews, and although mole-
                                                            cule-page reviews do not appear in print, the mol-
GOODWILL  ZFIN has relied on the goodwill of its mem-       ecule pages are published online and carry the
bers. Most of the members of this community had a           prestige of the Nature Publishing Group. The
connection to one specific senior researcher who both        Alliance’s editorial director has written letters in sup-
pioneered the use of zebra fish as a model organism          port of promotion and tenure cases indicating that
                                                                                    COLLABORATORIES ❚❙❘ 111



molecule page contributions are of journal-               projects there are tensions between users, who want
publication quality. This agreement is a creative at-     reliable tools that do what they need done, and com-
tempt to ensure that quality contributions will be        puter scientists, who are interested in technical inno-
made to the database; it also represents an interest-     vations and creative software ideas. There is little
ing evolution of the scholarly journal to include new     incentive for the computer scientists to go beyond
forms of scholarly publication.                           the initial demonstration versions of tools to the re-
                                                          liable and supported long-term operational infra-
Data Issues                                               structure desired by the users. In some fields, such
Data are a central component of all collaborations.       as high-energy physics, this tension has been at least
There are numerous issues concerning how data are         partially resolved. The field has used advanced soft-
represented and managed; how these issues are re-         ware for so long that it is understood that the extra
solved affects collaboratory success. For example,        costs associated with having production versions
good metadata—data about data—are critical as             of tools must be included in a project. Other fields
databases increase in size and complexity. Library        are only just discovering this. The organization of
catalogs and indexes to file systems are examples of       the George E. Brown, Jr., Network for Earthquake
metadata. Metadata are key to navigation and search       Engineering Simulation (NEES) project represents
through databases.                                        an innovation in this regard. The National Science
     Information about the provenance or origins of       Foundation, which funds the project, established it
the data is also important. Data have often been          in two phases, an initial four-year system-integration
highly processed, and researchers will want to know       phase in which the tools are developed and tested,
what was done to the original raw data to arrive at       and a ten-year operational phase overseen by a NEES
the processed data currently in the database. Two         consortium of user organizations.
related collaboratories in high-energy physics,                Any large organization faces difficult manage-
GriPhyN and iVDGL, are developing schemes for             ment issues, and practicing scientists may not always
showing investigators the paths of the transfor-          have the time or the skills to properly manage a com-
mations that led to the data in the database. This        plex enterprise. Management issues get even more
will help researchers understand the data and will        complicated when the organization is geographically
also help in identifying and correcting any errors in     distributed. Many large collaboratories have faced
the transformations.                                      difficult management issues. For instance, the two
     For some kinds of collaboratories, the complex       physics collaboratories mentioned earlier, GriPhyN
jurisdictional issues that arise when data are com-       and iVDGL, found that it was necessary to hire a full-
bined into a large database pose an interesting new       time project manager for each collaboratory in or-
issue. The BIRN project is facing just such an issue      der to help the science project directors manage
as it works to build up a database of brain images.       the day-by-day activities of the projects. The Alliance
The original brain images were collected at different     for Cellular Signaling has benefited from a charis-
universities or hospitals under different institutional   matic leader with excellent management skills who
review boards, entities that must approve any hu-         has set up a rich management structure to oversee
man data collection and preservation, and so the          the project. The BIRN collaboratory has an explicit
stipulations under which the original images were         governance manual that contains guidelines for a
collected may not be the same in every case.              host of tricky management issues; it also has a steer-
                                                          ing committee that is responsible for implementing
                                                          these management guidelines.
Other Issues
Many collaboratory projects involve cooperation
between domain scientists, who are the users of           Collaboratories in the Future
the collaboratory, and computer scientists, who are       Geographically distributed research projects are be-
responsible for the development of the tools. In many     coming commonplace in all the sciences. This
112 ❘❙❚ BERKSHIRE ENCYCLOPEDIA OF HUMAN-COMPUTER INTERACTION



proliferation is largely driven by what is required to                         Sproull, L., Conley, C., & Moon, J. Y. (in press). Pro-social behavior on
work at the frontiers of science. In the future, widely                           the net. In Y. Amichai-Hamburger (Ed.), The social net: The social
                                                                                  psychology of the Internet. New York: Oxford University Press.
shared knowledge about how to put together suc-                                Sproull, L. & Kiesler, S. (in press). Public volunteer work on the Internet.
cessful collaboratories will be essential. Of course,                             In B. Kahin & W. Dutton (Eds.), Transforming enterprise. Cam-
scientists are not alone in attempting geographically                             bridge, MA: MIT Press.
distributed collaborations. Similar issues are faced                           Star, S. L., & Ruhleder, K. (1994). Steps towards an ecology of infra-
                                                                                  structure: Complex problems in design and access for large-scale
in industry, education, government, and the non-                                  collaborative systems. In Proceedings of CSCW 94 (pp. 253–264).
profit sector. Good tools for collaboration and the                                New York: ACM Press.
social and organizational knowledge to make effec-                             Teasley, S., & Wolinsky, S. (2001). Scientific collaborations at a dis-
                                                                                  tance. Science, 292, 2254–2255.
tive use of them will be critical in all domains.                              Torvalds, L., & Diamond, D. (2001). Just for fun: The story of an acci-
                                                                                  dental revolutionary. New York: Harper Business.
                                                      Gary M. Olson            Wulf, W.A. (1993). The collaboratory opportunity. Science, 261,
                                                                                  854–855.
See also Computer-Supported Cooperative Work;
Groupware

                                                                                                   COMPILERS
FURTHER READING
                                                                               Compilers are computer programs that translate
Aldhous, P. (1993). Managing the genome data deluge. Science, 262,             one programming language into another. The orig-
   502–3.                                                                      inal program is usually written in a high-level lan-
Birnholtz, J., & Bietz, M. (2003). Data at work: Supporting sharing in         guage by a programmer and then translated into a
   science and engineering. In Proceedings of Group 2003. New
   York: ACM Press.
                                                                               machine language by a compiler. Compilers help
Cinkosky, M. J., Fickett, J. W., Gilna, P., & Burks, C. (1991). Electronic     programmers develop user-friendly systems by al-
   data publishing and GenBank. Science, 252, 1273–1277.                       lowing them to program in high-level languages,
Finholt, T. A. (2002). Collaboratories. In B. Cronin (Ed.), Annual             which are more similar to human language than ma-
   Review of Information Science and Technology, 36, 74–107.
   Washington, DC: American Society for Information Science and
                                                                               chine languages are.
   Technology.
Finholt, T. A., & Olson, G. M. (1997). From laboratories to collabo-
   ratories: A new organizational form for scientific collaboration.
   Psychological Science, 8(1), 28–36.
                                                                               Background
Huang, W., Olson, J. S., & Olson, G. M. (2002). Camera angle affects           Of course, the first compilers had to be written in
   dominance in video-mediated communication. In Proceedings of                machine languages because the compilers needed to
   CHI 2002, short papers (pp. 716–717). New York: ACM Press.                  operate the computers to enable the translation
National Science Foundation. (2003) Revolutionizing science and en-
   gineering through cyberinfrastructure: Report of the National Science
                                                                               process. However, most compilers for new computers
   Foundation blue-ribbon panel on cyberinfrastructure. Retrieved              are now developed in high-level languages, which
   December 24, 2003, from http://www.communitytechnology.org/                 are written to conform to highly constrained syntax
   nsf_ci_report/                                                              to ensure that there is no ambiguity.
Olson, G. M., Finholt, T. A., & Teasley, S. D. (2000). Behavioral aspects
   of collaboratories. In S. H. Koslow & M. F. Huerta (Eds.), Electronic
                                                                                    Compilers are responsible for many aspects of
   collaboration in science (pp. 1–14). Mahwah, NJ: Lawrence Erlbaum           information system performance, especially for
   Associates.                                                                 the run-time performance. They are responsible
Olson, G. M., & Olson, J. S. (2000). Distance matters. Human-Computer          for making it possible for programmers to use the
   Interaction, 15(2–3), 139–179.
Raymond, E. S. (1999). The cathedral and the bazaar: Musing on Linux           full power of programming language. Although com-
   and open source by an accidental revolutionary. Sebastopol, CA: O’Reilly.   pilers hide the complexity of the hardware from
Schatz, B. (1991). Building an electronic community system. Journal            ordinary programmers, compiler development re-
   of Management Information Systems, 8(3), 87–107.
Singh, Push (n.d.). Open mind common sense. Retrieved December
                                                                               quires programmers to solve many practical algo-
   22, 2003, from http://commonsense.media.mit.edu/cgi-bin/                    rithmic and engineering problems. Computer
   search.cgi                                                                  hardware architects constantly create new challenges
                                                                                          COMPILERS ❚❙❘ 113



for compiler developers by building more complex        takes the source code as input and then examines the
machines.                                               source program to check its conformity to the syn-
    Compilers translate programming languages and       tactic and semantic constraints of the language in
the following are the tasks performed by each spe-      which the program was written. During the synthetic
cific compiler type:                                     process, the object code in the target language is gen-
 ■
                                                        erated. Each major process is further divided. The
   Assemblers translate low-level language in-
                                                        analytic process, for example, consists of a character
   structions into machine code and map low-level
                                                        handler, a lexical analyzer, a syntax analyzer, and a
   language statements to one or more machine-
                                                        constraint analyzer. The character handler identifies
   level instructions.
 ■ Compilers translate high-level language in-
                                                        characters in the source text, and the lexical analyzer
                                                        groups the recognized characters into tokens such
   structions into machine code. High-level lan-
                                                        as operators, keywords, strings, and numeric con-
   guage statements are translated into more than
                                                        stants. The syntax analyzer combines the tokens into
   one machine-level instruction.
 ■ Preprocessors usually perform text substitutions
                                                        syntactic structures, and the constraint analyzer
                                                        checks to be sure that the identified syntactic
   before the actual translation occurs.
 ■ High-level translators convert programs written
                                                        structures meet scope and type rules.
                                                            The synthetic process consists of an intermedi-
   in one high-level language into another high-
                                                        ate code generator, a code optimizer, and a code gen-
   level language. The purpose of this translation is
                                                        erator. An intermediate code generator produces code
   to avoid having to develop machine-language-
                                                        that is less specific than the machine code, which will
   based compilers for every high-level language.
 ■ Decompilers and disassembers translate the ob-
                                                        be further processed by another language translator.
                                                        A code optimizer improves the intermediate code
   ject code in a low-level language into the source
                                                        with respect to the speed of execution and the com-
   code in a high-level language. The goal of this
                                                        puter memory requirement. A code generator
   translation is to regenerate the source code.
                                                        takes the output from the code optimizer and then
    In the 1950s compilers were often synonymous        generates the machine code that will actually be
with assemblers, which translated low-level language    executed on the target computer hardware.
instructions into directly executable machine code.
The evolution from an assembly language to a high-
level language was a gradual one, and the FORTRAN       Interpreters and
compiler developers who produced the first successful
high-level language did not invent the notion of pro-   Interpretive Compilers
gramming in a high-level language and then com-         In general, compilers produce the executable ob-
piling the source code to the object code. The first     ject code at the full speed, and compilers are usually
FORTRAN compiler was designed and written be-           designed to compile the entire source code before
tween 1954 and 1957 by an IBM team led by John          executing the resulting object code. However, it is
W. Backus, but it had taken about eighteen person-      common for programmers to expect to execute one
years of effort to develop. The main goal of the team   or more parts of a program before completing the
led by Backus was to produce object code that could     program. In addition, many programmers want to
execute as efficiently as human machine coders could.    write programs using a trial-and-error or what-if
                                                        strategy. These cases call for the use of an interpreter
                                                        in lieu of a traditional compiler because an inter-
Translation Steps                                       preter, which executes one instruction at a time, can
Programming language translators, including com-        take the source program as input and then execute
pilers, go through several steps to accomplish their    the instructions without generating any object code.
task, and use two major processes—an analytic pro-           Interpretive compilers generate simple inter-
cess and a synthetic process. The analytic process      mediate code, which satisfies the constraints of the
114 ❘❙❚ BERKSHIRE ENCYCLOPEDIA OF HUMAN-COMPUTER INTERACTION



practical interpreters. The intermediate code is then     control flow is manipulated by instances of regular
sent as input to an interpreter, which executes the al-   expressions in the input stream. Regular expressions
gorithm embedded in the source code by utilizing a        consist of normal characters, which include upper-
virtual machine. Within the virtual machine setting,      and lower-case letters and digits, and metacharac-
the intermediate code plays the role of executable        ters, which have special meanings. For example, a
machine code.                                             dot is a metacharacter, which matches any one char-
                                                          acter other than the new-line character. There is also
                                                          a table of regular expressions and their associated
Famous Compiler:                                          program pieces, called Lex source, and the resulting
                                                          program is a translation of the table. The program
GNU Compiler Collection (GCC)                             reads the input stream and generates the output
Many high-level language compilers have been im-          stream by partitioning the input into strings that
plemented using the C programming language                match the given regular expression.
and generating C code as output. Because almost all           Yacc is a general tool for describing the source
computers come with a C compiler, source code writ-       code to a program. After the Yacc user specifies the
ten in C is very close to being truly hardware-           structures to be recognized and the corresponding
independent and portable. The GNU Compiler                codes to be invoked, Yacc finds the hierarchical struc-
Collection (GCC) provides code generation for many        tures and transforms their specifications into sub-
programming languages such as C, C++, and Java,           routines that process the input.
and supports more than two hundred different soft-
ware and hardware platforms. The source code of
GCC is free and open, based on GNU General Public         The Future of Compilers
License, which allows people to distribute the            Proebstring’s Law states that “compiler advances
compiler’s source code as long as the original            double computing power every 18 years” (Proebsting,
copyright is not violated and the changes are pub-        n.d., 1). This implies that compiler-optimization
lished under the same license. This license enables       work makes a very minor contribution because it
users to port GCC to their platform of choice.            means that while the processing power of computer
     Presently almost all operating systems for per-      hardware increases by about 60 percent per year, the
sonal computers are supported by GCC and ship the         compiler optimization increases by only 4 percent.
compiler as an integrated part of the platform. For            Furthermore, some people claim that compilers
example, Apple’s Mac OS X is compiled using               will become obsolete with the increased use of script-
GCC 3.1. Other companies such as Sun and The              ing languages, which rely on interpreters or inter-
Santa Cruz Operation also offer GCC as their stan-        pretive compilers. Scripting languages, such as
dard system compiler. These examples show the flex-        Python, are popular among new programmers and
ibility and portability of GCC.                           people who do not care about minute efficiency dif-
                                                          ferences. However, there are arguments for the con-
                                                          tinued existence of compilers. One of the arguments
Compiler Constructor: Lex and Yacc                        is that there has to be a machine code on which the
Roughly speaking, compilers work in two stages. The       interpreters rely in order for a programmer’s intended
first stage is reading the source code to discover its     algorithm to be executed. In addition, there will al-
structure. The second stage is generating the exe-        ways be new and better hardware, which will then
cutable object code based on the identified structure.     rely on new compilers. It will also be impossible to
    Lex, a lexical-analyzer generator, and Yacc, a com-   extinguish the continuing desire to achieve even
piler-compiler, are programs used to discover the         minute performance improvements and compile-
structure of the source code. Lex splits the source       time error-detection capability. One of the proposed
code into tokens and then writes a program whose          future directions for compilers is to aid in increas-
                                                                             COMPUTER-SUPPORTED COOPERATIVE WORK ❚❙❘ 115



ing the productivity of programmers by optimizing                            Pizka, M. (1997). Design and implementation of the GNU INSEL
the high-level code. Another possible direction is                              Compiler gic. Technical Report TUM–I 9713. Munich, Germany:
                                                                                Munich University of Technology.
to make compilers smarter by making them self-                               Proebstring, T. (n.d.). Todd Proebsting’s home page. Retrieved January
steering and self-tuning, which would allow them to                             20, 2004, from http://research.microsoft.com/~toddpro/
adapt to input by incorporating artificial-intelligence                       Rice compiler group. (n.d.). Retrieved January 20, 2004, from http://
techniques.                                                                     www.cs.rice.edu/CS/compilers/index.html
                                                                             Terry, P. D. (1997). Compilers and compiler generators— an introduc-
                                                                                tion with C++. London: International Thomson Computer Press.
                                                        Woojin Paik          The comp.compilers newsgroup. (2002). Retrieved January 20, 2004,
                                                                                from http://compilers.iecc.com/index.html
                                                                             The Lex and Yacc page. (n.d.). Retrieved January 20, 2004, from
See also Programming Languages                                                  http://dinosaur.compilertools.net/
                                                                             Why compilers are doomed. (April 14, 2002). Retrieved January 20,
                                                                                2004, from http://www.equi4.com/jcw/wiki.cgi/56.html
FURTHER READING
Aho, A. V., Sethi, R., & Ulman, J. D. (1986). Compilers: principles, tech-
   niques and tools. Reading, MA: Addison-Wesley.

                                                                             COMPUTER-SUPPORTED
Aho, A. V., & Ulman, J. D. (1977). Principles of compiler design. Reading,
   MA: Addison-Wesley.
Bauer, A. (2003). Compilation of functional programming languages
   using GCC—Tail Calls. Retrieved January 20, 2004, from
   http://home.in.tum.de/~baueran/thesis/baueran_thesis.pdf                   COOPERATIVE WORK
A Brief History of FORTRAN /fortran. (1998). Retrieved January 20,
   2004, from http://www.ibiblio.org/pub/languages/FORTRAN/                  Computer-supported cooperative work (CSCW) is
   ch1-1.html
Catalog of free compilers and interpreters. (1998). Retrieved January
                                                                             the subarea of human-computer interaction con-
   20, 2004, from http://www.idiom.com/free-compilers/                       cerned with the communication, collaboration,
Clodius W. (1997). Re: History and evolution of compilers. Retrieved         and work practices of groups, organizations, and com-
   January 20, 2004, from http://compilers.iecc.com/comparch/article/        munities, and with information technology for
   97-10-008
Compiler Connection. (2003). Retrieved January 20, 2004, from
                                                                             groups, organizations, and communities. As the
   http://www.compilerconnection.com/index.html                              Internet and associated networked computing activ-
Compiler Internet Resource List. (n.d.). Retrieved January 20, 2004,         ities have become pervasive, research in CSCW has
   from http://www.eg3.com/softd/compiler.htm                                expanded rapidly, and its central concepts and vo-
Cooper, K., & Torczon, L. (2003). Engineering a Compiler. Burlington,
   MA: Morgan Kaufmann.                                                      cabulary are still evolving. For the purposes of this
Cooper, K., Kennedy, K., and Torczon, L. (2003). COMP 412 Overview           discussion, we understand cooperative work as any
   of the course. Retrieved January 20, 2004, from http://www.owl-           activity that includes or is intended to include the co-
   net.rice.edu/~comp412/Lectures/L01Intro.pdf
Cranshaw, J. (1997). Let’s build a compiler. Retrieved January 20, 2004,
                                                                             ordinated participation of at least two individuals;
   from http://compilers.iecc.com/crenshaw/                                  we take computer support of such work to be any in-
GCC Homepage. (January 26, 2004). Retrieved January 26, 2004, from           formation technology used to coordinate or carry out
   http://gcc.gnu.org/                                                       the shared activity (including archiving of the records
Free Software Foundation. (1991). GNU General Public License.
   Retrieved January 20, 2004, from http://www.fsf.org/licenses/
                                                                             of an activity to allow subsequent reuse by another).
   gpl.html                                                                       Several themes dominate research and practice
Joch, A. (January 22, 2001). Compilers, interpreters and bytecode.           in CSCW: studies of work, in which activities and
   Retrieved January 20, 2004, from http://www.computerworld.com/            especially tool usage patterns are observed, analyzed,
   softwaretopics/software/story/0,10801,56615,00.html
Lamm, E. (December 8, 2001). Lambda the Great. Retrieved January             and interpreted through rich qualitative descriptions;
   20, 2004, from http://lambda.weblogs.com/2001/12/08                       design and use of computer-mediated communica-
Mansour, S. (June 5, 1999). A Tao of Regular Expressions. Retrieved          tion (CMC) systems and of groupware, designed
   January 20, 2004, from http://sitescooper.org/tao_regexps.html
Manzoor, K. (2001). Compilers, interpreters and virtual machines.
                                                                             to aid with collaborative planning, acting, and sense
   Retrieved January 20, 2004, from http://homepages.com.pk/                 making; and analyses of the adoption and adapta-
   kashman/jvm.htm                                                           tion of CSCW systems.
116 ❘❙❚ BERKSHIRE ENCYCLOPEDIA OF HUMAN-COMPUTER INTERACTION




                 A Personal Story—Social Context in Computer-Supported
                                Cooperative Work (CSCW)

  In the early 1980s, our research group at the IBM Watson Research Center focused on the early stages of learning word pro-
  cessing systems, like the IBM Displaywriter. We carried out an extensive set of studies over several years. In these investi-
  gations, we noticed that people tried to minimize the amount of rote learning they engaged in, preferring to adopt
  action-oriented approaches in their own learning.
      Eventually, we developed a description of the early stages of learning to use computer applications that helped to de-
  fine new design approaches and learning support. But this work also made us wonder what more advanced learning might
  be like.
      To investigate this, my colleague John Gould and I visited an IBM customer site, to observe experienced users of
  Displaywriters as they worked in their everyday environments. These individuals were competent and confident in their
  use of the software. However we observed a pattern of distributed expertise: Each member of the staff had mastered one
  advanced function. Whenever someone needed to use an advanced function, she contacted the corresponding expert for
  personal, one-on-one coaching. This was a win-win situation: the requestors received customized help, and the specialized
  experts earned an increase in status. These field observations taught us the importance of people’s social context in the use
  and evaluation of information technology, something we now take for granted in CSCW.
                                                                                                           Mary Beth Rosson




Studies of Work                                                   and learning support and flexibility in the roles and
A fundamental objective of CSCW is to understand                  responsibilities available to human workers.
how computers can be used to support everyday work                    Studies of work often employ ethnographic
practices. Early research in the 1980s focused on                 methods adapted from anthropology. In ethno-
workflow systems. This approach codifies existing                   graphic research, the activities of a group are ob-
business procedures (for example, relating to the hir-            served over an extended period of time. This allows
ing of a new employee) in a computer model and                    collaborative activity to be seen in context. Thus,
embeds the model in a tracking system that moni-                  tasks are not characterized merely in terms of the
tors execution of the procedures, providing reminders,            steps comprising procedures, but also in terms of
coordination across participants, and assurance that              who interacts with whom to carry out and im-
appropriate steps are followed. Computerized work-                provise procedures, what tools and other arti-
flow systems are highly rational technological tools               facts are used, what information is exchanged and
whose goal is to support the effective execution of               created, and the longer-term collateral outcomes
normative procedures. Ironically, a major lesson that             of activity, such as personal and collective learn-
emerged from building and studying the use of these               ing and the development of group norms and mu-
systems is that exceptions to normative business                  tual trust. This work has demonstrated how, for
procedures are pervasive in real activity, and that               example, the minute interdependencies and
handling such exceptions characteristically involves              personal histories of doctors, nurses, patients,
social interactions that need to be fluid and nuanced              administrators, and other caregivers in the func-
in order to succeed. Indeed, the failure of direct and            tioning of a hospital must be analyzed to properly
rational workflow support was to a considerable ex-                understand actions as seemingly simple as a doc-
tent the starting point for modern CSCW, which now                tor conveying a treatment protocol to a nurse on
emphasizes balance between structured performance                 the next shift.
                                                           COMPUTER-SUPPORTED COOPERATIVE WORK ❚❙❘ 117



     Sometimes the observer tries to be invisible in       ambiguous, entailing clarifications and confirma-
ethnographic research, but sometimes the investi-          tions. And of course informal interactions are also
gator joins the group as a participant-observer.           often unproductive. Balancing direct support for
Typically video recordings of work activities are          work activities with broader support for building and
made, and various artifacts produced in the course         maintaining social networks is the current state of
of the work are copied or preserved to enable later        the classic workflow systems challenge.
analysis and interpretation.
     Ethnographic methods produce elaborate and
often voluminous qualitative descriptions of com-          Computer-Mediated Communication
plex work settings. These descriptions have be-            The central role of communication in the behavior
come central to CSCW research and have greatly             of groups has led to intense interest in how tech-
broadened the notion of context with respect to un-        nology can be used to enable or even enhance com-
derstanding human activity. Theoretical frameworks         munication among individuals and groups. Much
such as activity theory, distributed cognition, and        attention has been directed at communication among
situated action, which articulate the context of ac-       group members who are not colocated, but even
tivity, have become the major paradigms for science        for people who share an office, CMC channels such
and theory in CSCW.                                        as e-mail and text chat have become pervasive. Indeed
     Much of what people do in their work is guided        e-mail is often characterized as the single most suc-
by tacit knowledge. A team of engineers may not re-        cessful CSCW application, because it has been inte-
alize how much they know about one another’s               grated so pervasively into everyday work activities.
unique experience, skills, and aptitudes, or how well          The medium used for CMC has significant con-
they recruit this knowledge in deciding who to call        sequences for the communicators. Media richness
when problems arise or how to phrase a question or         theory suggests that media supporting video or voice
comment for best effect. But if an analyst observes        are most appropriate for tasks that have a subjective
them at work, queries them for their rationale dur-        or evaluative component because the nonverbal cues
ing problem-solving efforts, and asks for reflections       provided by a communicator’s visual appearance or
on why things happen, the tacit knowledge that is          voice tone provide information that helps partici-
uncovered may point to important trade-offs in             pants better understand and evaluate the full impact
building computerized support for their work               of one another’s messages. In contrast, text-based
processes. For instance, directing a question to an        media like e-mail or chat are better for gathering and
expert colleague provides access to the right infor-       sharing objective information. Of course, even text-
mation at the right time, but also establishes and re-     based channels can be used to express emotional con-
inforces a social network. Replacing this social           tent or subjective reactions to some extent; a large
behavior with an automated expert database may             and growing vocabulary of character-based icons
answer the query more efficiently, but may cause em-        and acronyms are used to convey sadness, happiness,
ployees to feel more disconnected from their               surprise, and so on.
organization.                                                  Use of CMC has also been analyzed from the per-
     A persistent tension in CSCW studies of work          spective of the psychologist Herbert Clark’s theory
springs from the scoping of activities to be sup-          of common ground in language—the notion that
ported. Many studies have shown how informal               language production, interpretation, and feedback
communication—dropping by a coworker’s office,              relies extensively on communicators’ prior knowl-
encountering someone in the hall, sharing a coffee—        edge about one another, the natural language they
can give rise to new insights and ideas and is essen-      are using, the setting they are in, and their group and
tial in creating group cohesion and collegiality, social   cultural affiliations. In CMC settings some of this
capital to help the organization face future challenges.   information may be missing. Furthermore, many of
But communication is also time consuming and often         the acknowledgement and feedback mechanisms that
118 ❘❙❚ BERKSHIRE ENCYCLOPEDIA OF HUMAN-COMPUTER INTERACTION



humans take for granted in face-to-face conversa-         and to build trusting relationships. Indeed, there have
tion (for example, head nods and interjected uh-          been a number of well-publicized episodes of cruel
huhs and so on) become awkward or impossible to           behavior in CMC environments such as chatrooms
give and receive in CMC. The theory of common             and MUDs (multiuser dungeons or domains).
ground argues that these simple acknowledgement                During the 1990s, cell phones, pagers, personal
mechanisms are crucial for fluid conversation be-          digital assistants, and other mobile devices rendered
cause they allow conversation partners to monitor         people and their work activities more mobile. As a
and track successful communication: A head nod or         consequence, the context of CMC became quite var-
an uh-huh tells the speaker that the listener under-      ied and unpredictable. A research area that has de-
stands what the speaker meant, is acknowledging           veloped in response to users’ changing environments
that understanding, and is encouraging the speaker        is context-aware computing, wherein the technology
to continue.                                              is used not only to support work activities, but also
     Despite the general acknowledgement that text-       to gather information about the users’ situation. For
based CMC media such as e-mail and chat are rel-          example, it is relatively straightforward to set up dis-
atively poor at conveying emotion and subjective          tinct settings for how a cell phone will operate
content, these channels have advantages that make         (e.g., ring tone or volume) at work, home, outdoors,
them excellent choices for some tasks. E-mail, for ex-    and so on, but it takes time and attention to re-
ample, is usually composed and edited in advance of       member to activate and deactivate them as needed.
sending the message; it can be read and reviewed          Thus the goal is to build devices able to detect changes
multiple times; and it is very easily distributed to      in people’s environment and to activate the appro-
large groups. E-mail is also easy to archive, and its     priate communication options or tasks. Whether such
text content can be processed in a variety of ways        mode changes take place automatically or are man-
to create reusable information resources. Because         aged by the individual, the resulting context infor-
e-mail is relatively emotion-free, it may be appro-       mation can be important to collaborators, signaling
priate for delicate or uncomfortable communication        if and when they can initiate or return to a shared
tasks. With so many CMC options, people are now           activity.
able to make deliberate (or tacit) choices among
CMC channels, using a relatively informal and un-
obtrusive medium like text chat for low-cost inter-       Groupware
action, more formally composed e-mail for business        CSCW software is often categorized by the timing
memos, and video or audio conferencing for im-            of the collaboration it supports: Synchronous group-
portant decision-making tasks.                            ware supports interaction at the same point in time,
     The relative anonymity of CMC (particularly          while asynchronous groupware supports collabo-
with text-based channels) has provoked considerable       ration across time. Another distinction is the col-
research into the pros and cons of anonymous com-         laborators’ relative location, with some groupware
munication. Communicators may use their real              designed for colocated interaction and some for dis-
names or screen names that only loosely convey their      tributed activities. For example, group decision sup-
identity; in some situations virtual identities may be    port systems are typically used for synchronous and
adopted explicitly to convey certain aspects of an in-    colocated interaction: As part of a face-to-face meet-
vented personality or online persona. Anonymity           ing, group members might use a shared online en-
makes it easier to express sensitive ideas and so can     vironment to propose, organize, and prioritize ideas.
be very effective when brainstorming or discussion        In contrast, an online forum might be used for asyn-
is called for but social structures would otherwise in-   chronous discussions among distributed group
hibit a high degree of sharing. However the same fac-     members.
tors that make anonymity an aid to brainstorming              A longstanding goal for many groupware devel-
also can lead to rude or inappropriate exchanges and      opers has been building support for virtual meetings—
may make it difficult to establish common ground           synchronous group interactions that take place
                                                                  COMPUTER-SUPPORTED COOPERATIVE WORK ❚❙❘ 119




                            A Personal Story—Internet Singing Lessons

  Having immigrated to the United States from India at an early age, I have always had a problem mastering the fine melodic
  nuances required to sing traditional Hindi songs. This problem has limited my singing repertoire.
       Last winter, during a religious meeting of Indian immigrants living in the Ann Arbor area, I was struck with how
  well a young Indian man sang a haunting Hindu chant. Later that evening I asked him to help me improve how I sang
  Hindu chants, which he did willingly. However, he soon informed me that he was returning to India the following week
  as he was in the U.S. on a temporary work visa.
       Because I was disappointed in losing such a willing teacher, my friend suggested a technological solution. He suggested
  that I set up an account with Yahoo! Messenger, and to buy a stereo headset through which we could continue our music
  interaction. Yahoo! Messenger is an instant messaging system that enables logged-in users to exchange text messages, and
  to talk free of charge on the Internet.
       When my friend returned to India, we had to deal with two problems. First, we had to deal with the time-difference.
  India is 10½ hours ahead of the U. S. Second, we had to deal with the problem that my friend only had access to an Internet
  connection at the office where he worked. This is because computers and Internet connections are still quite expensive
  for the average Indian. We therefore decided that the best time for undisturbed instant voice messaging would be at 7:30 a.m.
  Indian Standard Time when other employees had not yet arrived in my friend’s office. This time also work out well for
  me because it would be 9:00 p.m. (EST), the time when I liked to pluck on my guitar and sing.
       The above plan worked well—on February 8th, 2004, I had my first “transcontinental singing lesson.” Despite a slight
  delay in sound transmission due to the Internet bandwidth problem, my friend was able to correct the fine melodic nu-
  ances that I missed when I sang my favorite Hindu chant. I can now sing a Hindu chant with nuances approved by a singing
  teacher sitting in front of a computer many oceans away.
                                                                                                              Suresh Bhavnani




entirely online as a substitute for traditional face-to-          viewing, and revising of content that is a natural con-
face meetings. As businesses have become increas-                 sequence of working together online.
ingly international and distributed, support for                      When collaborators meet online, participant au-
virtual meetings has become more important. A vir-                thentication is an important issue. Many work sit-
tual meeting may use technology as simple as a tele-              uations have policies and procedures that must be
phone conference call or as complex as a collaborative            respected; for example, meetings may have a spec-
virtual environment (CVE) that embodies attendees                 ified attendee list or restricted documents, or de-
and their work resources as interactive objects in a              cisions may require the approval of a manager.
three-dimensional virtual world. Because virtual                  Enforcing such restrictions creates work for both
meetings must rely on CMC, attendees have fewer                   the organizer of the activity (who must activate the
communication cues and become less effective at                   appropriate controls) and the participants (who
turn taking, negotiation, and other socially rich in-             must identify themselves if and when required).
teraction. It is also often difficult to access and in-            Depending on a group’s culture and setting, the
teract with meeting documents in a CVE, particularly              meeting organizers may choose to make no restric-
when the meeting agenda is open and information                   tions at all (for example, they may meet in an on-
needs to evolve during the meeting. Some researchers              line chatroom and rely on group members to
have argued that online meetings will never equal                 self-enforce relevant policies and group behavior),
face-to-face interaction, and that researchers should             or they may rely on a set of roles (such as leader, at-
focus instead on the special qualities offered by a               tendee, or scribe) built into the groupware system
virtual medium—for example, the archiving, re-                    to manage information access and interaction.
120 ❘❙❚ BERKSHIRE ENCYCLOPEDIA OF HUMAN-COMPUTER INTERACTION



    A significant technical challenge for synchronous      know which group members are around, available
groupware is ensuring data consistency. When col-         for interaction, and so on. Social awareness can be
laborators are able to communicate or edit shared         provided through mechanisms such as buddy lists,
data in parallel, there is the possibility that simul-    avatars (online representations of group members),
taneous requests will conflict: One participant might      or even regularly updated snapshots of a person in
correct the spelling of a word at the same time that      their work setting. For a shared project that takes
another member deletes a phrase containing the            place over weeks or months, collaborators need ac-
word, for example. The simplest technique for avoid-      tivity awareness: They must be aware of what proj-
ing consistency problems is to implement a floor          ect features have changed, who has done what, what
control mechanism that permits only one partici-          goals or plans are currently active, and how to con-
pant at a time to have the virtual pen, with others       tribute. However, promoting activity awareness re-
waiting until it is passed to them. Because such mech-    mains an open research topic; considerable work is
anisms can be awkward and slow, many groupware            needed to determine how best to integrate across
systems have explored alternatives, including implicit    synchronous and asynchronous interactions, what
locking of paragraphs or individual words, and fully      information is useful in conveying status and
optimistic serialization, which processes all input in    progress, and how this information can be gathered
the order in which it is received, with the assump-       and represented in a manner that supports rather
tion that well-learned social protocols of turn tak-      than interrupts collaborative activities.
ing and coordination will reduce conflict and ensure
smooth operation.
    Many other technical challenges plague the            Adoption and Adaptation of
smooth operation of groupware. For instance, it is
quite common for collaborators to be interacting          CSCW Systems
with rather different hardware and software plat-         Even when great care is taken in the design and
forms. Although work groups may settle on a stan-         implementation of a CSCW system, there is no guar-
dard set of software, not all group members may           antee that it will be successfully adopted and inte-
follow all aspects of the standard, and beyond the        grated into work practices—or that when it is
work group, there may be no standards. Thus inter-        adopted it will work as originally intended. Many
operability of data formats, search tools, editing or     case studies point to a sociotechnical evolution cy-
viewing software, and analysis tools is a constant con-   cle: Initially, delivered CSCW systems do not fit onto
cern. As work settings have become more mobile and        existing social and organizational structures and
dynamic, the variety of technical challenges has in-      processes. During a process of assimilation and ac-
creased: Some members at a virtual meeting may            commodation, the organization changes (for exam-
join by cell phone, while others may use a dedicated      ple, a new role may be defined for setting up and
broadband network connection. It is increasingly          facilitating virtual meetings) in concert with the tech-
common for groupware systems to at least provide          nology (for example, a set of organization-specific
an indicator of such variation, so that collabora-        templates may be defined to simplify agenda setting
tors can compensate as necessary (for example, by         and meeting management).
recognizing that a cell phone participant may not be           Several implications follow from this view of
able to see the slides presented at a meeting).           CSCW adoption. One is that participatory design of
    The general goal of promoting awareness dur-          the software is essential—without the knowledge of
ing CSCW interactions has many facets. During syn-        praxis provided by the intended users, the software
chronous work, groupware often provides some form         will not be able to evolve to meet their specific needs;
of workspace awareness, with telepointers or minia-       furthermore if users are included in the design
turized overviews showing what objects are selected       process, introduction of the CSCW system into the
or in view by collaborators. In more extended col-        workplace will already have begun by the time the
laborations, partners depend on social awareness to       system is deployed. Another implication is that
                                                           COMPUTER-SUPPORTED COOPERATIVE WORK ❚❙❘ 121



CSCW software should have as open an architecture          the new technology, much of the sociotechnical evo-
as possible, so that when the inevitable need for          lution has taken place, context-specific procedures
changes is recognized months or years after deploy-        have been developed and refined in situ, and there
ment, it will be possible to add, delete, or other-        are local experts to assist new users.
wise refine existing services. A third implication is             As more and more of an organization’s activities
that organizations seeking CSCW solutions should           take place online—whether through e-mail or video-
be ready to change their business structures and           conferencing or shared file systems—via CSCW tech-
processes—and in fact should undertake business            nology, the amount of online information about the
process reengineering as they contribute to the de-        organization and its goals increases exponentially.
sign of a CSCW system. A frequent contributing fac-        The increased presence of organizational informa-
tor in groupware failure is uneven distribution of         tion online has generated great interest in the
costs and benefits across organizational roles and re-      prospects for organizational memory or knowledge
sponsibilities. There are genuine costs to collabo-        management. The hope is that one side effect of carry-
ration: When an individual carries out a task, its         ing out activities online will be a variety of records
subtasks may be accomplished in an informal and            about how and why tasks are decomposed and ac-
ad hoc fashion, but distributing the same task among       complished, and that these records can provide guid-
individuals in a group is likely to require more ad-       ance to other groups pursuing similar goals. Of
vance planning and negotiation, recordkeeping, and         course once again, there are important cost-benefit
explicit tracking of milestones and partial results.       issues to consider: Recording enough information
Collaboration implies coordination. Of course the          to be helpful to future groups takes time, especially
benefits are genuine as well: One can assign tasks         if it is to be stored in any useful fashion, and the ben-
to the most qualified personnel, one gains multiple         efit in most cases will be enjoyed by other people.
perspectives on difficult problems, and social recog-       One solution is to give computers the job of record-
nition and rewards accrue when individuals com-            ing, organizing, and retrieving. For example, even
bine efforts to reach a common goal. Unfortunately,        a coarse-grained identification of speakers making
the costs of collaboration are often borne by work-        comments in a meeting can simplify subsequent
ers, who have new requirements for online planning         browsing of the meeting audiotape.
and reporting, while its benefits are enjoyed by man-
agers, who are able to deliver on-time results of higher
quality. Therefore, when designing for sociotechni-        Research Directions
cal evolution, it is important to analyze the expected     Much of the active research in CSCW is oriented
costs and benefits and their distribution within the        toward new technologies that will enhance aware-
organization. Equally important are mechanisms for         ness, integrate multiple devices, populations, and
building social capital and trust, such that individ-      activities, and make it possible to visualize and share
uals are willing to contribute to the common good,         rich data sets and multimedia documents. The need
trusting that others in the group will reward or           to interconnect people who are using diverse devices
care for them when the time comes.                         in diverse settings entails many research challenges,
     Critical mass is another determinant of success-      some related to the general issues of multiplatform
ful adoption—the greater the proportion of indi-           computing and others tied to understanding and
viduals within an organization who use a technology,       planning for the social and motivational differences
the more sense it makes to begin using it oneself.         associated with varied work settings. The rapidly
A staged adoption process is often effective, with a       expanding archives in organizations offer many re-
high-profile individual becoming an early user and          search opportunities related to data processing and
advocate who introduces the system to his or her           analysis as well as information visualization and re-
group. This group chronicles its adoption experience       trieval. At the same time, these digital storehouses
and passes the technology on to other groups, and          raise important questions about individual privacy
so on. By the time the late adopters begin to use          and identity—the more information an organization
122 ❘❙❚ BERKSHIRE ENCYCLOPEDIA OF HUMAN-COMPUTER INTERACTION



collects about an individual, the more opportunity                           Baecker, R. M. (1993). Readings in groupware and computer-supported
there is for inappropriate access to and use of this                            cooperative work: Assisting human-human collaboration. San
                                                                                Francisco: Morgan-Kaufmann.
information.                                                                 Beaudouin-Lafon, M. (Ed). (1999). Computer supported co-operative
    A methodological challenge for CSCW is the de-                              work. Chichester, UK: John Wiley & Sons.
velopment of effective evaluation methods. Field                             Bikson, T. K., & Eveland, J. D. (1996). Groupware implementation:
                                                                                Reinvention in the sociotechnical frame. In Proceedings of the
studies and ethnographic analyses yield very rich
                                                                                Conference on Computer Supported Cooperative Work: CSCW ’96
data that can be useful in understanding system                                 (pp. 428–437). New York: ACM Press.
requirements and organizational dynamics. But                                Carroll, J. M., Chin, G., Rosson, M .B., & Neale, D. C. (2000). The
analyzing such detailed records to answer precise                               development of cooperation: Five years of participatory design in
                                                                                the virtual school. In Designing interactive systems: DIS 2000
questions is time consuming and sometimes im-                                   (pp. 239–251). New York: ACM Press.
possible due to the complexity of real-world settings.                       Carroll, J. M., & Rosson, M.B. (2001). Better home shopping or new
Unfortunately, the methods developed for study-                                 democracy? Evaluating community network outcomes. In
ing individual computer use do not scale well to the                            Proceedings of Human Factors in Computing Systems: CHI 2001
                                                                                (pp. 372–379). New York: ACM Press.
evaluation of multiple users in different locations.                         Dourish, P., & Bellotti, V. (1992). Awareness and coordination in shared
Because social and organizational context are a key                             workspaces. In Proceedings of the Conference on Computer Supported
component of CSCW activities, it is difficult to sim-                            Cooperative Work: CSCW ’92 (pp. 107–114). New York: ACM Press.
                                                                             Grudin, J. (1994). Groupware and social dynamics: Eight challenges
ulate shared activities in a controlled lab setting.                            for developers. Communications of the ACM, 37(1), 92–105.
Groupware has been evolving at a rapid rate, so there                        Gutwin, C., & Greenberg, S. (1999). The effects of workspace aware-
are few if any benchmark tasks or results to use for                            ness support on the usability of real-time distributed group-
comparison studies. One promising research di-                                  ware. ACM Transactions on Computer-Human Interaction, 6(3),
                                                                                243–281.
rection involves fieldwork that identifies interesting                         Harrison, S., & Dourish, P. (1996). Re-placing space: The roles of place
collaboration scenarios; these are then scripted                                and space in collaborative systems. In Proceedings of the Conference
and simulated in a laboratory setting for more sys-                             on Computer Supported Cooperative Work: CSCW ’96 (pp. 67–76).
                                                                                New York: ACM Press.
tematic analysis.                                                            Hughes, J., King, V., Rodden, T., & Andersen, H. (1994). Moving out
    In the 1980s human-computer interaction                                     from the control room: Ethnography in system design. In
focused on solitary users finding and creating in-                              Proceedings of the Conference on Computer Supported Cooperative
formation using a personal computer. Today, the                                 Work: CSCW ’94 (pp. 429–439). New York: ACM Press.
                                                                             Hutchins, E. (1995). Distributed cognition. Cambridge, MA: MIT Press.
focus is on several to many people working together                          Malone, T. W., & Crowston, K. (1994). The interdisciplinary study
at a variety of times and in disparate places, relying                          of coordination. ACM Computing Surveys, 26(1), 87–119.
heavily on the Internet, and communicating and col-                          Markus, M. L. (1994). Finding a happy medium: Explaining the
                                                                                negative effects of electronic communication on social life at work.
laborating more or less continually. This is far more                           ACM Transactions on Information Systems, 12(2), 119–149.
than a transformation of human-computer inter-                               Nardi, B. A. (1993). A small matter of programming. Cambridge,
action; it is a transformation of human work and                                MA: MIT Press.
activity. It is still under way, and CSCW will continue                      Nardi, B. A. (Ed). (1996). Context and consciousness: Activity theory
                                                                                and human-computer interaction. Cambridge, MA: MIT Press.
to play a large role.                                                        Olson, G. M., & Olson, J. S. (2000). Distance matters. Human Computer
                                                                                Interaction, 15(2–3), 139–179.
                  Mary Beth Rosson and John M. Carroll                       Orlikowski, W. J. (1992). Learning from notes: Organizational issues
                                                                                in groupware implementation. In Proceedings of the Conference on
                                                                                Computer Supported Cooperative Work: CSCW ’92 (pp. 362–369).
See also Collaboratories; Ethnography; MUDs; Social                             New York: ACM Press.
Psychology and HCI                                                           Roseman, M., & Greenberg, S. (1996). Building real time groupware
                                                                                with Groupkit, a groupware toolkit. ACM Transactions on Computer
                                                                                Human Interaction, 3(1), 66–106.
                                                                             Sproull, L., & Kiesler, S. (1991). Connections: New ways of working in
FURTHER READING                                                                 the networked organization. Cambridge, MA: MIT Press.
                                                                             Streitz, N. A., Geißler, J., Haake, J., & Hol, J. (1994). DOLPHIN:
Ackerman, M. S. (2002). The intellectual challenge of CSCW: The gap             Integrated meeting support across local and remote desktop en-
   between social requirements and technical feasibility. In J. M. Carroll      vironments and liveboards. In Proceedings of the Conference on
   (Ed.), Human-computer interaction in the new millennium                      Computer Supported Cooperative Work: CSCW ’94 (pp. 345–358).
   (pp. 303–324). New York: ACM Press.                                          New York: ACM Press.
                                                                                              CONSTRAINT SATISFACTION ❚❙❘ 123



Suchman, L. (1987). Plans and situated actions: The problem of human-       straints specify that exactly four of the decision vari-
   machine communication. Cambridge, UK: Cambridge University               ables have value 1 (“queen in this square”) and
   Press.
Sun, C., & Chen, D. (2002). Consistency maintenance in real-time col-
                                                                            that there cannot be two queens in the same row, col-
   laborative graphics editing systems. ACM Transactions on Computer        umn, or diagonal. Because there are sixteen variables
   Human Interaction, 9(1), 1–41.                                           (one for each square) and each can take on two pos-
Tang, J., Yankelovich, N., Begole, J., Van Kleek, M., Li, F., & Bhalodia,   sible values, there are a total of 2¹⁶ (65,536) possible
   J. (2001). Connexus to awarenex: Extending awareness to mobile
   users. In Proceedings of Human Factors in Computing Systems: CHI         assignments of values to the decision variables. There
   2001 (pp. 221–228). New York: ACM Press.                                 are other ways of modeling the 4-queen problem
Winograd, T. (1987/1988). A language/action perspective on the de-          within the CSP framework. One alternative is to treat
   sign of cooperative work. Human-Computer Interaction, 3(1), 3–30.
                                                                            each row on the board as a decision variable. The
                                                                            values that can be taken by each variable are the four
                                                                            column positions in the row. This formulation yields
                                                                            4⁴ (256) possibilities. This example illustrates how
                CONSTRAINT                                                  the initial formulation or model affects the number
                                                                            of possibilities to be examined, and ultimately the
               SATISFACTION                                                 performance of problem solving.

Constraint satisfaction refers to a set of representa-
tion and processing techniques useful for model-                            CSP Representations
ing and solving combinatorial decision problems;                            A CSP is often represented as an undirected graph
this paradigm emerged from the artificial intelligence                       (or network), which is a set of nodes connected by
community in the early 1970s. A constraint satis-                           a set of edges. This representation opens up the
faction problem (CSP) is defined by three elements:                          opportunity to exploit the properties and algorithms
(1) a set of decisions to be made, (2) a set of choices                     developed in graph theory for processing and solv-
or alternatives for each decision, and (3) a set of con-                    ing CSPs. In a constraint graph, the nodes represent
straints that restrict the acceptable combinations                          the variables and are labeled with the domains of the
of choices for any two or more decisions. In general,                       variables. The edges represent the constraints and
the task of a CSP is to find a consistent solution—                          link the nodes corresponding to the variables to
that is, a choice for every decision such that all the                      which the constraints apply. The arity of a constraint
constraints are satisfied. More formally, each deci-                         designates the number of variables to which the con-
sion is called a variable, the set of alternative choices                   straint applies, and the set of these variables consti-
for a given variable is the set of values or domain of                      tutes the scope of the constraint. Constraints that apply
the variable, and the constraints are defined as the set                     to two variables are called binary constraints and are
of allowable combinations of assignments of values                          represented as edges in the graph. Constraints that
to variables. These combinations can be given in ex-                        apply to more than two variables are called non-
tension as the list of consistent tuples, or defined in                      binary constraints. While, early on, most research
intention as a predicate over the variables.                                has focused on solving binary CSPs, techniques for
                                                                            solving nonbinary CSPs are now being investigated.

The 4-Queen Problem
A familiar example of a CSP is the 4-queen problem.                         The Role of CSPs in Science
In this problem, the task is to place four queens on                        Beyond puzzles, CSPs have been used to model
a 4×4 chessboard in such a way that no two queens                           and solve many tasks (for example, temporal rea-
attack each other. One way to model the 4-queen                             soning, graphical user interfaces, and diagnosis) and
problem as a CSP is to define a decision variable for                       have been applied in many real-world settings (for
each square on the board. The square can be either                          example, scheduling, resource allocation, and prod-
empty (value 0) or have a queen (value 1). The con-                         uct configuration and design). They have been used
124 ❘❙❚ BERKSHIRE ENCYCLOPEDIA OF HUMAN-COMPUTER INTERACTION



in various areas of engineering, computer science,           domain of B is the interval [7, 11]), and B oc-
and management to handle decision problems.                  curred one hour after A (B-A ≥ 1). It is easy to in-
A natural extension of the CSP is the constrained            fer that the domains of A and B must be restricted
optimization problem (COP), where the task is to             to [8, 10] and [9, 11] respectively, because B can-
find an optimal solution to the problem given a set           not possibly occur before 9, or A after 10, without
of preferences and optimization criteria. The prob-          violating the constraint between A and B. This fil-
lems and issues studied in the constraint processing         tering operation considers every combination of two
(CP) community most obviously overlap with those             variables in a binary CSP. It is called 2-consistency.
investigated in operations research, satisfiability and       A number of formal properties have been proposed
theoretical computer science, databases, and pro-            to characterize the extent to which the alternative
gramming languages. The 1990s have witnessed a               combinations embedded in a problem description
sharp increase in the interactions and cross-fertilization   are likely to yield consistent solutions, as a mea-
among these areas.                                           sure of how “close is the problem to being solved.”
     A special emphasis is made in CP to maintain            These properties characterize the level of consistency
the expressiveness of the representation. Ideally, a         of the problem (for example, k-consistency, mini-
human user should be able to naturally express the           mality, and decomposability).
various relations governing the interactions among                Algorithms for achieving these properties, also
the entities of a given problem without having to re-        known as constraint propagation algorithms, remain
cast them in terms of complex mathematical mod-              the subject of intensive research. Although the cost
els and tools, as would be necessary in mathematical         of commonly used constraint propagation algorithms
programming. The area of constraint reformulation            is a polynomial function of the number of variables
is concerned with the task of transforming the prob-         of the CSP and the size of their domains, solving the
lem representation in order to improve the perfor-           CSP remains, in general, an exponential-cost process.
mance of problem solving or allow the use of available       An important research effort in CP is devoted to find-
solution techniques. Sometimes such transforma-              ing formal relations between the level of consistency
tions are truthful (that is, they preserve the essence       in a problem and the cost of the search process used
of the problem), but often they introduce some               for solving it. These relations often exploit the topol-
sufficient or necessary approximations, which may             ogy of the constraint graph or the semantic proper-
or may not be acceptable in a particular context.            ties of the constraint. For example, a tree-structured
                                                             constraint graph can be solved backtrack-free after
                                                             ensuring 2-consistency, and a network of constraints
Solution Methods                                             of bounded differences (typically used in temporal
The techniques used to solve a CSP can be divided            reasoning) is solved by ensuring 3-consistency.
into two categories: constraint propagation (or infer-
ence) and search. Further, search can be carried out         Systematic Search
as a systematic, constructive process (which is ex-          In systematic search, the set of consistent combi-
haustive) or as an iterative repair process (which of-       nations is explored in a tree-like structure starting
ten has a stochastic component).                             from a root node, where no variable has a value, and
                                                             considering the variables of the CSP in sequence.
Constraint Propagation                                       The tree is typically traversed in a depth-first man-
Constraint propagation consists in eliminating, from         ner. At a given depth of the tree, the variable under
the CSP, combinations of values for variables that           consideration (current variable) is assigned a value
cannot appear in any solution to the CSP. Consider           from its domain. This operation is called variable in-
for example two CSP variables A and B representing           stantiation. It is important that the value chosen for
two events. Assume that A occurred between                   the current variable be consistent with the instan-
8 a.m. and 12 p.m. (the domain of A is the interval          tiations of the past variables. The process of check-
[8, 12]), B occurred between 7 a.m. and 11 a.m. (the         ing the consistency of a value for the current variable
                                                                              CONSTRAINT SATISFACTION ❚❙❘ 125



with the assignments of past variables is called back-      state to another and attempting to find a state where
checking. It ensures that only instantiations that are      all constraints are satisfied. This move operator and
consistent (partial solutions) are explored. If a           the state evaluation function are two important com-
consistent value is found for the current variable,         ponents of an iterative-repair search. The move is
then this variable is added to the list of past variables   usually accomplished by changing the value of one
and a new current variable is chosen from among             variable (thus the name local search). However, a
the un-instantiated variables (future variables).           technique operating as a multiagent search allows
Otherwise (that is, no consistent value exists in the       any number of variables to change their values. The
domain of the current variable), backtracking is ap-        evaluation function measures the cost or quality of
plied. Backtracking undoes the assignment of the            a given state, usually in terms of the number of bro-
previously instantiated variable, which becomes the         ken constraints. Heuristics, such as the min-conflict
current variable, and the search process attempts           heuristic, are used to choose among the states reach-
to find another value in the domain of this variable.        able from the current state (neighboring states).
     The process is repeated until all variables have            The performance of iterative-repair techniques
been instantiated (thus yielding a solution) or back-       depends heavily on their ability to explore the so-
track has reached the root of the tree (thus proving        lution space. The performance is undermined by the
that the problem is not solvable). Various techniques       existence in this space of local optima, plateaux, and
for improving the search process itself have been pro-      other singularities caused by the nonconvexity of the
posed. For systematic search, these techniques in-          constraints. Heuristics are used to avoid falling
clude intelligent backtracking mechanisms such as           into these traps or to recover from them. One heuris-
backjumping and conflict-directed backjumping.              tic, a breakout strategy, consists of increasing the
These mechanisms attempt to remember the reasons            weight of the broken constraints until a state is
for failure and exploit them during search in order         reached that satisfies these constraints. Tabu search
to avoid exploring barren portions of the search            maintains a list of states to which search cannot move
space, commonly called thrashing. The choices of the        back. Other heuristics use stochastic noise such as
variable to be instantiated during search and that of       random walk and simulated annealing.
the value assigned to the variable are handled, re-
spectively, by variable and value ordering heuristics,      Blending Solution Techniques
which attempt to reduce the search effort. Such             Constraint propagation has been successfully com-
heuristics can be applied statically (that is, before the   bined with backtrack search to yield effective look-
search starts) or dynamically (that is, during the          ahead strategies such as forward checking. Combining
search process). The general principles that guide          constraint propagation with iterative-repair strate-
these selections are “the most constrained variable         gies is less common. On the other hand, randomiza-
first” and “the most promising value first.” Examples         tion, which has been for a long time utilized in
of the former include the least domain heuristic            local search, is now being successfully applied in back-
(where the variable with the smallest domain is cho-        track search.
sen for instantiation) and the minimal-width heuris-
tic (where the variables are considered in the ordering
of minimal width of the constraint graph).                  Research Directions
                                                            The use of constraint processing techniques is wide-
Iterative-Repair Search                                     spread due to the success of the constraint pro-
In iterative repair (or iterative improvement) search,      gramming paradigm and the increase of commercial
all the variables are instantiated (usually randomly)       tools and industrial achievements. While research on
regardless of whether or not the constraints are            the above topics remains active, investigations are also
satisfied. This set of complete instantiations, which        invested in the following directions: user interaction;
is not necessarily a solution, constitutes a state.         discovery and exploitation of symmetry relations;
Iterative-repair search operates by moving from one         propagation algorithms for high-arity constraints
126 ❘❙❚ BERKSHIRE ENCYCLOPEDIA OF HUMAN-COMPUTER INTERACTION



and for continuous domains; preference modeling                                 Montanari, U. (1974). Networks of constraints: Fundamental proper-
and processing; distributed search techniques; em-                                 ties and application to picture processing. Information Sciences, 7,
                                                                                   95–132.
pirical assessment of problem difficulty; and statis-                            Prosser, P. (1993). Hybrid algorithms for the constraint satisfaction
tical evaluation and comparison of algorithms.                                     problem. Computational Intelligence, 9(3), 268–299.
                                                                                Régin, J.-C. (1994). A filtering algorithm for constraints of difference
                                                 Berthe Y. Choueiry                in constraint satisfaction problems. In Proceedings from the National
                                                                                   Conference on Artificial Intelligence (AAAI 1994) (pp. 362–437).
                                                                                   Seattle, WA.
See also Artificial Intelligence; N-grams                                        Revesz, P. (2002). Introduction to constraint databases. New York:
                                                                                   Springer.
                                                                                Stuckey, K. M. (1998). Programming with constraints: An introduction.
                                                                                   Cambridge, MA: MIT Press.
FURTHER READING                                                                 Tsang, E. (1993). Foundations of constraint satisfaction. London, UK:
                                                                                   Academic Press.
                                                                                Yokoo, M. (1998). Distributed constraint satisfaction. New York:
Bistarelli, S., Montanari, U., & Rossi, F. (1997). Semiring-based con-
                                                                                   Springer.
   straint satisfaction and optimization. Journal of the ACM, 44(2),
   201–236.
Borning, A., & Duisberg, R. (1986). Constraint-based tools for build-
   ing user interfaces. ACM Transactions on Graphics, 5(4), 345–374.

                                                                                               CONVERGING
Cohen, P. R. (1995). Empirical methods for artificial intelligence.
   Cambridge, MA: MIT Press.
Dechter, R. (2003). Constraint processing. San Francisco: Morgan
   Kaufmann.
Ellman, T. (1993). Abstraction via approximate symmetry. In                                   TECHNOLOGIES
   Proceedings of the 13th IJCAI (pp. 916–921). Chambéry, France.
Freuder, E. C. (1982). A sufficient condition for backtrack-free search.         Human-computer interaction (HCI) is a multi-
   JACM, 29(1), 24–32.
Freuder, E. C. (1985). A sufficient condition for backtrack-bounded
                                                                                disciplinary field arising chiefly in the convergence
   search. JACM, 32(4), 755–761.                                                of computer science, electrical engineering, infor-
Freuder, E. C. (1991). Eliminating interchangeable values in constraint         mation technology, and cognitive science or psy-
   satisfaction problems. In Proceedings of AAAI-91 (pp. 227–233).              chology. In the future it is likely to be influenced
   Anaheim, CA.
Gashnig, J. (1979). Performance measurement and analysis of certain
                                                                                by broader convergences currently in progress, reach-
   search algorithms. Pittsburgh, PA: Carnegie-Mellon University.               ing out as far as biotechnology and nanotechnology.
Glaisher, J. W. L. (1874). On the problem of the eight queens.                  Together, these combined fields can take HCI to new
   Philosophical Magazine, 4(48), 457–467.                                      levels where it will unobtrusively but profoundly en-
Glover, F. (1989). Tabu Search—Part I. ORSA Journal on Computing,
   1(3), 190–206.                                                               hance human capabilities to perceive, to think, and
Gomes, C. P. (2004). Randomized backtrack search. In M. Milano                  to act with maximum effectiveness.
   (Ed.), Constraint and Integer Programming: Toward a Unified
   Methodology (pp. 233–291). Kluwer Academic Publishers.
Haralick, R. M., & Elliott, G. L. (1980). Increasing tree search efficiency
   for constraint satisfaction problems. Artificial Intelligence, 14, 263–313.   The Basis for Convergence
Hogg, T., Huberman, B. A., & Williams, C. P. (Eds.). (1996). Special            During the twentieth century a number of interdis-
   volume on frontiers in problem solving: Phase transitions and                ciplinary fields emerged, bridging the gaps be-
   complexity. Artificial Intelligence, 81(1–2). Burlington, MA: Elsevier
   Science.
                                                                                tween separate traditionally defined sciences. Notable
Hooker, J. (2000). Logic-based methods for optimization: Combining              examples are astrophysics (astronomy plus physics),
   optimization and constraint satisfaction. New York: Wiley.                   biochemistry (biology plus chemistry), and cogni-
Hoos, H. H., & Stützle, T. (2004). Stochastic local search. San Francisco:      tive science (psychology plus neurology plus com-
   Morgan Kaufmann.
Kirkpatrick, S., Gelatt, J. C. D., & Vecchi, M. P. (1983). Optimization         puter science). Many scientists and engineers believe
   by simulated annealing. Science, 220(4598), 671–680.                         that the twenty-first century will be marked by a
Liu, J., Jing, H., & Tang, Y. Y. (2002). Multi-agent oriented constraint        broader unification of all of the sciences, permitting
   satisfaction. Artificial Intelligence, 136(1), 101–144.
Minton, S., et al. (1992). Minimizing conflicts: A heuristic repair
                                                                                a vast array of practical breakthroughs—notably
   method for constraint satisfaction and scheduling problems.                  in the convergence of nanotechnology, biotechnology,
   Artificial Intelligence, 58, 161–205.                                         information technology, and cognitive technology—
                                                                           CONVERGING TECHNOLOGIES ❚❙❘ 127



based on the unification of nanoscience, biology,         seem remote from HCI because the human senses
information science, and cognitive science. HCI itself    operate at a much larger scale. However, we can al-
stands at the junction between the last two of these      ready identify a number of both direct and in-
four, and it has the potential to play a major role in    direct connections, and as work at the nanoscale
the emergence of converging technologies.                 promotes convergence between other fields it will
     A number of scientific workshops and conferences,     create new opportunities and challenges for HCI.
organized by scientists and engineers associated with          The largest single atoms, such as those of ura-
the U.S. National Science Foundation and building         nium, are just smaller than 1 nanometer. The struc-
upon the United States National Nanotechnology            tures of complex matter that are fundamental to
Initiative, have concluded that nanoscience and nano-     all sciences originate at the nanoscale. That is the
technology will be especially important in conver-        scale at which complex inorganic materials take on
gence. Nanoscience and nanotechnology concern             the characteristic mechanical, electrical, and
scientific research and engineering (respectively)        chemical properties they exhibit at larger scales. The
at the nanoscale, the size range of physical structures   nanoscale is where the fundamental structures of life
between about 1 nanometer and 100 nanometers in           arise inside biological cells, including the human
shortest dimension. A nanometer is 1 billionth of a       DNA (deoxyribonucleic acid) molecule itself. The
meter, or 1 millionth of a millimeter, and a millimeter   double helix of DNA has the proportions of a twisted
is about the thickness of a dime (the thinnest U.S.       piece of string, about 2.5 nanometers thick but as
coin). Superficially, nanoscience and nanotechnology       much as 4 centimeters (40 million nanometers) long




        The BRLESC-II, a solid-state digital computer introduced in 1967. It was designed to be
        200 times faster than the ORDVAC computer it replaced. Photo courtesy of the U.S. Army.
128 ❘❙❚ BERKSHIRE ENCYCLOPEDIA OF HUMAN-COMPUTER INTERACTION



if uncoiled. The synaptic gaps between neurons in          Wide Web, both hardware and software. Many of the
the human brain, and the structures that contain the       early applications have been new ways of accom-
neurotransmitter chemicals essential to their func-        plishing old tasks, for example, word processors, dig-
tioning, are on the order of 20 to 50 nanometers.          ital music and television, and more recently digital
     Nanotechnology and nanoscience are chiefly a           libraries. The integration of mobile computing with
partnership of physics, chemistry, and materials sci-      the Internet is expected to unleash a wave of radi-
ence (an interdisciplinary field at the intersection of     cally different innovations, many of which cannot
physics, chemistry, and engineering that deals with        even be imagined today, connected to ubiquitous
the properties of materials, including composite ma-       availability of information and of knowledge tools.
terials with complex structures). In the near term              Cognitive science is the study of intelligence,
nanotechnology offers engineering a host of new            whether human, nonhuman animal, or machine,
materials, including powders with nanoscale gran-          including perception, memory, decision, and un-
ules, thin coatings that transform the properties of       derstanding. It is itself a convergence of fields, draw-
surfaces, and composite materials having nanoscale         ing upon psychology, social psychology, cultural
structure that gives them greater strength, durabil-       anthropology, linguistics, economics, sociology,
ity, and other characteristics that can be precisely de-   neuroscience, artificial intelligence, and machine
signed for many specific uses. In the midterm to long       learning. The fundamental aim is a profound un-
term, nanotechnology is expected also to achieve           derstanding of the nature of the human mind. By
practical accomplishments with complex nano-               the beginning of the twenty-first century a new uni-
structures, including new kinds of electronic com-         verse of cognitive technologies clearly was opening
ponents and nanoscale machines.                            up, especially in partnerships between humans and
     Biotechnology applies discoveries in biology to       computers. The result could be technologies that
the invention and production of products that are          overcome breakdowns in human awareness, analy-
valuable for human health, nutrition, and economic         sis, planning, decision making, and communication.
well-being. The traditional application areas for               Each of these four fields is a fertile field of sci-
biotechnology are medicine and agriculture, in-            entific research and technological development, but
cluding the production of chemicals and construc-          in combination they can achieve progress much more
tion materials having organic origins. Biotechnology       rapidly and broadly than they can alone. Following
has a long history, extending back thousands of years      are examples of the science and engineering oppor-
to ancient industries such as fermentation of alco-        tunities in each of the six possible pairs.
hol, tanning of hides, dyeing of clothing, and bak-
ing of bread. The pace of innovation accelerated           Nanotechnology–Biotechnology
throughout the nineteenth and twentieth centuries,         Research at the nanoscale can reveal the detailed, dy-
leading to the latest developments in genomics (a          namic geometry of the tiny structures that carry out
branch of biotechnology concerned with applying            metabolism, movement, and reproduction inside the
the techniques of genetics and molecular biology to        living cell, thereby greatly expanding biological sci-
the genetic mapping and DNA sequencing of sets of          ence. Biology provides conceptual models and prac-
genes or the complete genomes of selected organ-           tical tools for building inorganic nanotechnology
isms) and a growing understanding of the structures        structures and machines of much greater complex-
and processes inside the living cell.                      ity than currently possible.
     Information technology is a creation of the
second half of the twentieth century, revolutioniz-        Nanotechnology–Information Technology
ing traditional communication technologies through         Nanoelectronic integrated circuits will provide the
the introduction of electronic computation. It             fast, efficient, highly capable hardware to support
comprises computers, information systems, and com-         new systems for collecting, managing, and distrib-
munication networks such as Internet and the World         uting information wherever and whenever it is
                                                                            CONVERGING TECHNOLOGIES ❚❙❘ 129



needed. Advances in information technology will be         ization tools to help people understand biology and
essential for the scientific analysis of nanoscale struc-   biotechnology correctly.
tures and processes and for the design and manu-
facture of nanotechnology products.                        Information Technology–Cognitive Technology
                                                           Experiments on human and nonhuman animal
Nanotechnology–Cognitive Technology                        behavior depend upon computerized devices for data
New research methods based on nanoscale sensor             collection and on information systems for data analy-
arrays will enable neuroscientists to study the fine        sis, and progress can be accelerated by sharing in-
details of neural networks in the brain, including the     formation widely among scientists. Discoveries by
dynamic patterns of interaction that are the basis of      cognitive scientists about the ways the human mind
human thought. Cognitive science will help nano-           carries out a variety of judgments provide models
scientists and educators develop the most readily          for how machines could do the same work, for ex-
intelligible models of nanoscale structures and the        ample, to sift needed information from a vast as-
innovative curriculum needed for students to un-           sembly of undigested data.
derstand the world as a complex hierarchy of sys-
tems built up from the nanoscale.
                                                           HCI Contributions to Convergence
Biotechnology–Information Technology                       Attempting to combine two scientific disciplines
Principles from evolutionary biology can be applied        would be futile unless they have actually moved into
to the study of human culture, and biologically in-        adjacent intellectual territories and proper means
spired computational methods such as genetic al-           can be developed to bridge between them. Disciplines
gorithms (procedures for solving a mathematical            typically develop their own distinctive assumptions,
problem in a finite number of steps that frequently         terminologies, and methodologies. Even under the
involve repetition of an operation) can find mean-          most favorable conditions, transforming tools are
ingful patterns in vast collections of information.        needed, such as new concepts that can connect the
Bioinformatics, which consists of biologically ori-        disparate assumptions of different disciplines,
ented databases with lexicons for translating from         ontologies—category schemes and lexicons of
one to another, is essential for managing the huge         concepts in a particular domain—that translate lan-
trove of data from genome (the genetic material of         guage across the cultural barriers between disciplines,
an organism) sequencing, ecological surveys,               and research instrumentation or mathematical analy-
large-scale medical and agricultural experiments,          sis techniques that can be applied equally well in
and systematic comparisons of evolutionary con-            either discipline. Because many of these trans-
nections among thousands of species.                       forming tools are likely to be computerized, human-
                                                           computer interaction research will be essential for
Biotechnology–Cognitive Technology                         scientific and technological convergence.
Research techniques and instruments developed in                One of the key ways of developing fresh scientific
biotechnology are indispensable tools for research         conceptualizations, including models and metaphors
on the nature and dynamics of the nervous system,          that communicate successfully across disciplinary
in both humans and nonhuman animals, under-                barriers, is computer visualizations. For example,
stood as the products of millions of years of bio-         three-dimensional graphic simulations can help stu-
logical evolution. Human beings seem to have great         dents and researchers alike understand the structures
difficulty thinking of themselves as parts of com-          of complex molecules at the nanoscale, thus bridg-
plex ecological systems and as the products of             ing between nanoscience and molecular biology,
evolution by natural selection from random evolu-          including genomics and the study of the structures
tion, so advances will be needed to design fresh           inside the living cell. In trying to understand the
approaches to scientific education and new visual-          behavior of protein molecules, virtual reality (VR)
130 ❘❙❚ BERKSHIRE ENCYCLOPEDIA OF HUMAN-COMPUTER INTERACTION



may incorporate sonification (the use of sounds to             Researchers in many diverse sciences already have
represent data and information) in which a                established a shared data infrastructure, such as in-
buzzing sound represents ionization (the dissocia-        ternational protein structure and genomics databases
tion of electrons from atoms and molecules, thus giv-     and the online archives that store thousands of so-
ing them an electric charge), and haptics (relating to    cial and behavioral science questionnaire datasets.
the sense of touch) may be used to represent the at-      The development of digital libraries has expanded
traction between atoms by providing a counteract-         the range of media and the kinds of content that can
ing force when a VR user tries to pull them apart. For    be provided to scholars, scientists, and engineers over
data that do not have a natural sensory representa-       the Internet. Grid computing, which initially served
tion, a combination of psychology and user-centered       the supercomputing community by connecting geo-
design, focusing on the needs and habitual thought        graphically distributed “heavy iron” machines, is ma-
patterns of scientists, will identify the most success-   turing into a vast, interconnected environment of
ful forms of data visualization, such as information      shared scientific resources, including data collection
spaces that map across the conceptual territories of      instrumentation, information storage facilities, and
adjacent sciences.                                        major storehouses of analytic tools. As more and
     HCI is relevant not only for analyzing statistical   more research traditions join the grid world, they
or other data that have already been collected and        will come to understand each other better and find
computerized, but also for operating scientific in-        progressively more areas of mutual interest. This con-
struments in real time. Practically every kind of sci-    vergence will be greatly facilitated by advances in
entific research uses computerized instruments today.      human-computer interaction research.
Even amateur astronomical telescopes costing under
$500 have guidance computers built into them. In
the future expensive computerized instruments used        Implications for Computing
in nanoscience, such as atomic force microscopes          Because HCI already involves unification of infor-
(tools for imaging individual atoms on a surface, al-     mation and cognitive technologies, distinctive effects
lowing one to see the actual atoms), may provide          of convergence will primarily occur in unification
haptic feedback and three-dimensional graphics to         with the two other realms: nanotechnology and bio-
let a user virtually feel and see individual atoms when   technology. Nanotechnology is likely to be especially
manipulating them, as if they have been magnified          crucial because it offers the promise of continued
10 million times.                                         improvement in the performance of computer com-
     In any branch of science and engineering, HCI-       ponents. Already a nanoscale phenomenon called
optimized augmented cognition and augmented re-           the “giant magnetoresistance” (GMR) effect has been
ality may play a useful role, and after scientists and    used to increase the data density on mass produc-
engineers in different fields become accustomed to         tion computer hard disks, giving them much greater
the same computer methods for enhancing their abil-       capacity at only slight cost. The two key components
ities, they may find it easier to communicate and thus     of a computer hard disk are a rotatable magnetic disk
collaborate with each other. For example, primate         and a read-and-write head that can move along
cognitive scientists, studying the behavior of baboons,   the radius of the disk to sense the weak magnetism
may collaborate with artificial-intelligence re-          of specific tiny areas on the disk, each of which
searchers, and both can employ augmented reality          represents one bit (a unit of computer information
to compare the behavior of a troop of real animals        equivalent to the result of a choice between two al-
with a multiagent system designed to simulate them.       ternatives) of data. Making the active tip of the read-
Internet-based scientific collaboratories can not only     and-write head of precisely engineered materials
provide a research team at one location with a va-        constructed in thin (nanoscale) layers significantly
riety of transforming tools, but also let researchers     increases its sensitivity. This sensitivity, in turn, al-
from all around the world become members of the           lows the disk to be formatted into a larger number
team through telepresence.                                of smaller areas, thereby increasing its capacity.
                                                                             CONVERGING TECHNOLOGIES ❚❙❘ 131



    Since the beginning of the human-computer in-           that perform logical operations) and carbon nano-
teraction field, progress in HCI has depended not            tube transistors (transistors made of nanoscale tubes
only on the achievements of its researchers, but also       composed of carbon).
on the general progress in computer hardware. For                If successful, these radically new approaches
example, early in the 1970s the Altair computer pio-        require development of an entire complex of fresh
neered the kind of graphic user interface employed          technologies and supporting industries; thus, the
by essentially all personal computers at the end of         cost of shifting over to them may be huge. Only a
the twentieth century, but its memory chips were too        host of new applications could justify the massive
expensive, and its central processing unit was              investments, by both government and industry, that
too slow. A decade later the chips had evolved to the       will be required. Already people in the computer in-
point where Apple could just barely market the              dustry talk of “performance overhang,” the possi-
Macintosh, the first commercially successful com-            bility that technical capabilities have already
puter using such an interface. Today many areas             outstripped the needs of desirable applications. Thus,
of HCI are only marginally successful, and along with       a potential great benefit for HCI becomes also a great
HCI research and development, increased power and           challenge. If HCI workers can demonstrate that a
speed of computers are essential to perfect such ap-        range of valuable applications is just beyond the reach
proaches as virtual reality, real-time speech recogni-      of the best computers that the old technology can
tion, augmented cognition, and mobile computing.            produce, then perhaps people will have sufficient
    Since the mid-1960s the density of transistors on       motivation to build the entire new industries that
computer chips has been doubling roughly every eight-       will be required. Otherwise, all of computer sci-
een months, and the cost of a transistor has been drop-     ence and engineering may stall.
ping by half. So long as this trend continues, HCI can           During the twentieth century several major tech-
count on increasingly capable hardware. At some             nologies essentially reached maturity or ran into
point, possibly before 2010, manufacturers will no          social, political, or economic barriers to progress.
longer be able to achieve progress by cramming more         Aircraft and automobiles have changed little in recent
and more components onto a chip of the tradi-               years, and they were certainly no faster in 2000 than
tional kind. HCI progress will not stop the next day,       in 1960. The introduction of high-definition tele-
of course, because a relatively long pipeline of research   vision has been painfully slow, and applications of
and development exists and cannot be fully exploited        haptics and multimodal augmented reality outside
before several more years pass. Progress in other areas,    the laboratory move at a snail’s pace. Space flight
such as parallel processing and wireless networking,        technology has apparently stalled at about the tech-
will still be possible. However, HCI would benefit          nical level of the 1970s. Nuclear technology has either
greatly if electronic components continued to become        been halted by technical barriers or blocked by po-
smaller and smaller because this miniaturization            litical opposition, depending on how one prefers
means they will continue to get faster, use progres-        to analyze the situation. In medicine the rate of in-
sively less power, and possibly also be cheaper.            troduction of new drugs has slowed, and the great
    Here is where nanotechnology comes in. Actually,        potential of genetic engineering is threatened by in-
the transistors on computer chips have already shrunk       creasing popular hostility. In short, technological
into the nanoscale, and some of them are less than          civilization faces the danger of stasis or decline un-
50 nanometers across. However, small size is only           less something can rejuvenate progress.
one of the important benefits of nanotechnology.                  Technological convergence, coupled with ag-
Equally important are the entirely new phenomena,           gressive research at the intersections of technical
such as GMR, that do not even exist at larger scales.       fields, may be the answer. Because HCI is a conver-
Nanotechnologists have begun exploring alterna-             gent field itself and because it can both benefit from
tives to the conventional microelectronics that we          and promote convergence, HCI can play a central
have been using for decades, notably molecular logic        role. In addition to sustaining progress as tradition-
gates (components made of individual molecules              ally defined, convergence enables entirely new
132 ❘❙❚ BERKSHIRE ENCYCLOPEDIA OF HUMAN-COMPUTER INTERACTION



applications. For example, nanotechnology provides         explanations—statements about how and why re-
the prospect of developing sensors that can instantly      wards may be obtained and costs are incurred. In
identify a range of chemicals or microorganisms in         the language of computer science, such explana-
the environment, and nano-enabled microscale sen-          tions are algorithms.
sor nets can be spread across the human body, a rain            Some algorithms are very specific and apply only
forest, and the wing of an experimental aircraft to        under certain narrowly defined circumstances. If one
monitor their complex systems of behavior.                 wants meat, one takes a big stick from the forest, goes
                                                           into the meadow, and clobbers one of the sheep graz-
                                                           ing there. If one wants water, one goes to the brook
Paradigm Transformation                                    at the bottom of the valley. These are rather spe-
Convergence is not just a matter of hiring a multi-        cific explanations, assuming that only one meadow,
disciplinary team of scientists and engineers and          one kind of animal, one brook, and one valley exist.
telling them to work together. To do so they need ef-      As the human mind evolved, it became capable of
fective tools, including intellectual tools such as com-   working out much more general algorithms that ap-
prehensive theories, mathematical techniques for           plied to a range of situations. If one wants meat, one
analyzing dynamic systems, methods for visualizing         takes a club, goes to any meadow, and sees what one
complex phenomena, and well-defined technical              can clobber there. If one wants water, the bottoms
words with which to talk about them.                       of deep valleys are good places to look. In the terms
     Decades ago historian Thomas Kuhn described           of artificial intelligence, the challenge for human in-
the history of science as a battle between old ways of     telligence was how to generalize, from a vast com-
thought and new paradigms (frameworks) that may            plexity of experience, by reasoning from particular
be objectively better but inevitably undergo oppo-         cases to develop rules for solving particular broad
sition from the old-guard defenders of the prevail-        kinds of problems.
ing paradigm. His chief example was the so-called               Stark and Bainbridge noted how difficult it is for
Copernican Revolution in astronomy, when the no-           human beings to invent, test, and perfect very gen-
tion that the Earth is the center of the universe was      eral explanations about the nature of the universe
displaced by a new notion that the sun is the cen-         and thereby to find empirically good algorithms for
ter of the solar system and of a vast, centerless uni-     solving the problems faced by our species. In other
verse far beyond. The problem today is that many           words, science and technology are difficult enter-
paradigms exist across all branches of science and         prises that could emerge only after ten thousand years
engineering. Some may be equivalent to each other,         of civilization and that cannot be completed for many
after their terms are properly translated. Others may      decades to come. In the absence of a desired reward,
be parts of a larger intellectual system that needs        people often will accept algorithms that posit at-
to be assembled from them. However, in many areas          tainment of the reward in the distant future or in
inferior paradigms that dominate a particular dis-         some other non-verifiable context. Thus, first sim-
cipline will need to be abandoned in favor of one          ple magic and then complex religious doctrines
that originated in another discipline, and this process    emerged early in human history, long before humans
is likely to be a hard-fought and painful one taking       had accurate explanations for disease and other dis-
many years.                                                asters, let alone effective ways of dealing with
     The human intellectual adventure extends back         them. If the full convergence of all the sciences and
tens of thousands of years. In their research and          technologies actually occurs, as it may during the
theoretical work on the origins of religion, Rodney        twenty-first century, one can wonder what will be-
Stark and William Sims Bainbridge observed that            come not only of religion but of all other forms of
human beings seek rewards and try to avoid costs—          unscientific human creativity, what are generally
a commonplace assumption in economics and other            called the “humanities.”
branches of social science. To solve the problems               The U.S. entomologist and sociobiologist Edward
they faced every day, ancient humans sought                O. Wilson has written about the convergence that
                                                                                         CYBERCOMMUNITIES ❚❙❘ 133



is far advanced among the natural sciences, calling       human minds of all ages. Possibly no such com-
it “consilience,” and has wondered whether the hu-        prehensive explanation of reality (an algorithm for
manities and religion will eventually join in to be-      controlling nature) is possible. Or perhaps the in-
come part of a unified global culture. Here again         tellectuals and investors who must build this fu-
human-computer interaction may have a crucial role        ture world may not be equal to the task. Thus,
to play because HCI thrives exactly at the boundary       whether it succeeds or fails, the technological con-
between humans and technology.                            vergence movement presents a huge challenge for
     During the first sixty years of their existence,     the field of human-computer interaction, testing
computers evolved from a handful of massive ma-           how well we can learn to design machines and in-
chines devoted to quantitative problems of engi-          formation systems that help humans achieve their
neering and a few physical sciences to hundreds of        maximum potential.
millions of personal tools, found in every school
or library, most prosperous people’s homes, and                                               William Sims Bainbridge
many people’s pockets. Many people listen to music
or watch movies on their computers, and thousands         See also Augmented Cognition; Collaboratories
of works of literature are available over the Internet.
A remarkable number of digital libraries are devoted
to the humanities, and the U.S. National Endowment        FURTHER READING
for the Humanities was one of the partner agencies
in the Digital Library Initiative led by the U.S.         Atkins, D. E., Drogemeier, K. K., Feldman, S. I., Garcia-Molina, H.,
                                                             Klein, M. L., Messerschmitt, D. G., Messina, P., Ostriker, J. P., &
National Science Foundation.                                 Wright, M. H. (2003). Revolutionizing science and engineering
     The same HCI methods that are used to help sci-         through cyberinfrastructure. Arlington, VA: National Science
entists visualize complex patterns in nature can             Foundation.
become new ways of comprehending schools of art,          Kuhn, T. (1962). The structure of scientific revolutions. Chicago:
                                                             University of Chicago Press.
tools for finding a desired scholarly reference, or even   Roco, M. C., & Bainbridge, W. S. (2001). Societal implications of
new ways of creating the twenty-second-century               nanoscience and nanotechnology. Dordrecht, Netherlands: Kluwer.
equivalents of paintings, sculptures, or symphonies.      Roco, M. C., & Bainbridge, W. S. (2003). Converging technologies for
                                                             improving human performance. Dordrecht, Netherlands: Kluwer.
The same virtual reality systems that will help sci-      Roco, M. C., & Montemagno, C. D. (Eds.). (2004). The coevolution of
entists collaborate across great distances can become        human potential and converging technologies. Annals of the New
a new electronic medium, replacing television, in            York Academy of Sciences, 1013. New York: New York Academy of
which participants act out roles in a drama while            Sciences.
                                                          Stark, R., & Bainbridge, W. S. (1987). A theory of religion. New York:
simultaneously experiencing it as theater. Cyber-            Toronto/Lang.
infrastructure resources such as geographic infor-        Wilson, E. O. (1998). Consilience: The unity of knowledge. New York:
mation systems, automatic language translation               Knopf.
machines, and online recommender systems can
be used in the humanities as easily as in the sciences.
     The conferences and growing body of publica-
tions devoted to converging technologies offer a pic-          CYBERCOMMUNITIES
ture of the world a decade or two in the future when
information resources of all kinds are available at all   For many people, the primary reason for interacting
times and places, organized in a unified but malleable     with computers is the ability, through computers, to
ontology, and presented through interfaces tai-           communicate with other people. People form cyber-
lored to the needs and abilities of individual users.     communities by interacting with one another
Ideally, education from kindergarten through grad-        through computers. These cybercommunities are
uate school will be organized around a coherent           conceived of as existing in cyberspace, a concep-
set of concepts capable of structuring reality in ways    tual realm created through the networking and in-
that are simultaneously accurate and congenial to         terconnection that computers make possible.
134 ❘❙❚ BERKSHIRE ENCYCLOPEDIA OF HUMAN-COMPUTER INTERACTION



Cybercommunity Definition and History                      able on the Internet first made available in 1991 and
The prefix cyber first appeared in the word cyber-        given a graphical interface in 1993). People also
netics, popularized by Norbert Weiner (1894–1964)         began using the Internet in the early 1980s to run
in the 1940s to refer to the science of “control and      bulletin board services such as Usenet, which, unlike
communication in the animal and the machine” (the         the earlier local BBSs, could now be distributed to
subtitle of Weiner’s 1948 book Cybernetics). Since        a much larger group of people and accessed by people
that time, cyber has prefixed many other words to          in widely dispersed geographical locations. Usenet
create new terms for various interconnections be-         expanded to include many different cybercommu-
tween computers and humans. One of the terms, cy-         nities, most based around a common interest such
berspace, has become a popular metaphor for the           as Linux programming or soap operas.
perceived location of online interactions. Coined by
William Gibson in his 1984 novel Neuromancer,             Existing Cybercommunities
cyberspace originally referred to a graphical repre-      Cybercommunities have risen in number with the
sentation of computerized data to which people con-       increasing availability and popularity of the Internet
nected through direct electrical links to the brain.      and the World Wide Web. Even within the short over-
Since then, the term has come to mean any virtual         all history of cybercommunities, some cyber-
forum in which people communicate through com-            communities have been short-lived. However, there
puters, whether the form of communication involves        are several, begun in the early days of computer net-
text, graphics, audio, or combinations of those.          working, that still exist online and therefore present
    Cybercommunities predate widespread use of            a useful view of factors involved in the formation
the Internet, with the first forming in localized sys-     and maintenance of online communities.
tems called bulletin board services (BBSs). BBSs              One of the oldest still-extant cybercommunities
usually ran on a single computer, and participants        is The WELL, which began in 1985 as a local BBS
connected through modems and a local phone line.          in the San Francisco Bay Area in California. Laurence
This meant that most participants lived within a lim-     Brilliant, a physician with an interest in computer
ited geographical area. Thus many BBSs were able to       conferencing, and Stewart Brand, editor of the Whole
hold occasional face-to-face get-togethers, enhanc-       Earth Review and related publications, founded The
ing community relationships. Communication on             WELL with the explicit goal of forming a virtual
BBSs was usually asynchronous; that is, people logged     community. One savvy method the founders used
on at different times and posted messages in various      to attract participants was to give free accounts to
topical forums for others to read and respond to later.   local journalists, many of whom later wrote about
(E-mail and similar bulletin boards now available         their participation, generating further interest and
on the World Wide Web are also asynchronous forms         publicity. In the early years, when most participants
of online communication, while the various types of       lived in the same geographical area, The WELL held
online chat and instant messaging are considered          monthly face-to-face meetings. Currently owned by
to be synchronous forms of communication, since           Salon.com, The WELL is now accessible through the
participants are present on a forum simultane-            World Wide Web.
ously and can spontaneously respond to each other’s           Another venerable cybercommunity, Lambda-
communications.)                                          MOO, also began as an experiment in online com-
    From the earliest days of the Internet and its        munity. In contrast to The WELL, LambdaMOO
military-funded precursor, the Arpanet (established       provided a forum for synchronous communica-
in 1969), online participants began forming cyber-        tion and allowed people to create a virtual envi-
communities. E-mail immediately emerged as the            ronment within which to interact. LambdaMOO
largest single use of the Internet, and remained so       is an example of a type of program called a MUD,
until 2002, when it was matched by use of the World       (for multiuser dimension or multiuser dungeon).
Wide Web (an information-exchange service avail-          MUDs are similar to online chatrooms, but also allow
                                                                                    CYBERCOMMUNITIES ❚❙❘ 135



participants to create their own virtual spaces and
objects as additions to the program, with which they            Welcome to LamdaMOO
and others can then interact. These objects enhance

                                                         B
                                                         elow is an introduction to the cybercommunity
the feel of being in a virtual reality. Created by the
                                                         LamdaMOO, as presented on www.lamdamoo.info:
computer scientist Pavel Curtis as a research project
for Xerox, LambdaMOO opened in 1990. A 1994 ar-              LambdaMOO is sort of like a chat room. It’s a
ticle about it in Wired magazine led to a significant         text-only based virtual community of thousands
increase in interest in it and to dramatic population        of people from all over the world. It’s comprised
growth. Pavel Curtis has moved on to other projects,         of literally thousands of “rooms” that have
and LambdaMOO is no longer associated with Xerox.            been created by the users of LambdaMOO, and
                                                             you endlessly navigate (walk around) north,
But although it has undergone considerable social
                                                             south, etc. from room to room, investigating, and
changes over the years, it still attracts hundreds of        meeting people that you can interact with to your
participants.                                                hearts content.
     MUDs began as interactive text-based role-                    You get there not thru an HTML browser like
playing games inspired by similar face-to-face               Netscape or IE but through another program
roleplaying games such as Dungeons and Dragons               called TELNET (search). Your computer most
                                                             likely has Telnet but enhanced versions can be
(hence dungeon in one expansion of the acronym).
                                                             found. (Telnet address: telnet://lambda.moo
More recently, similar online games have become              .mud.org:8888/). You can try the Lambda button
available with the enhancement of a graphical in-            at the top of this page to see if all goes well. If so,
terface. People have used MMORPGs (massively                 a window will open and you’ll be able to log in.
multiplayer online role-playing games) such as Ever-               When you get the hang of it, you can create
quest as forums for socializing as well as gaming, and       a character who has a name and a physical de-
                                                             scription, and who can be “seen” by all who meet
cybercommunities are forming amongst online
                                                             you. As you walk around from room to room you
gamers.                                                      are given a description of the room and a list of
     Web logs, or blogs, are a relatively new and in-        contents (including other people). You can “look”
creasingly popular platform for cybercommunities.            at each person to get a more detailed description
Blogs are online journals in which one can post one’s        and when you do, they see a message stating that
thoughts, commentary, or reflections, sometimes              you just checked them out. You can talk to
                                                             them and they see your words in quotes, like read-
also allowing others to post comments or reactions
                                                             ing spoken words in a book. You can also
to these entries. Many blogs provide a forum for             emote (communicate with body language) using
amateur (or, in some cases, professional) journal-           gestures such as a smile or a nod of the head. In
ism, however others more closely resemble online             time you’ll learn to create your own rooms or
personal diaries. There are many different programs          other objects, which are limited only by your
available for blogging, and some give specific at-           imagination.
                                                                   There are many people to meet up with
tention to community formation. LiveJournal, for
                                                             and build “cyber-friendships” with. When you
instance, enables each participant to easily gather          first get there you’ll be asked to log in. First timers
other journals onto a single page, making it easy to         can sign in as a guest. After that you can apply for
keep up with friends’ journals. Links between par-           a permanent character name and password. Give
ticipants are also displayable, enabling people to see       it a try and if you see me around say hi.
who their friends’ friends are and to easily form and            Felis~Rex
expand social networks.                                          Status: Programmer/108/33%Fogy/PC
                                                                 Parent: Psychotic Class of Players
Community Networks                                               Seniority: 1320/4093, (33%)
                                                                 MOO-age: 108 months. (1995 January 28,
Some cybercommunities grow out of existing offline
                                                                     Saturday)
communities. In particular, some local municipali-
ties have sought to increase citizen participation in
136 ❘❙❚ BERKSHIRE ENCYCLOPEDIA OF HUMAN-COMPUTER INTERACTION



the local government and community by forming                As more and more online sites use multimedia
community computer networks that allow people            and bandwidth-intensive enhancements (that is,
access to government officials and provide forums         enhancements that can only be successfully trans-
for community discussion. The first of these, the        mitted across a wide range—or band—of electro-
Public Electronic Network (PEN), started in Santa        magnetic frequencies), speed of access has also become
Monica, California, in 1989. It was particularly         a crucial issue. People with older equipment—slower
successful in providing access to low-income citizens    modems and computer processors—are disadvan-
who might not otherwise have had access to com-          taged in their ability to access online materials, es-
puters or computer networks.                             pecially at multimedia sites. Some governments,
     In more recent years, some offline communities       notably in South Korea and Japan, have sought to
have gone beyond providing an online forum specif-       address that problem by subsidizing the develop-
ically related to the community and have also sought     ment of broadband networks, enabling widespread
to promote computer use and connectivity in gen-         relatively inexpensive access in those countries to
eral. For instance, the town of Blacksburg, Virginia,    high-speed Internet connections.
with the help of Virginia Polytechnic Institute and          In addition to access to equipment and networks,
State University (known as Virginia Tech and located     people need the skills that enable them to use that
in town) and local businesses, is attempting to          access. Research has also shown that people are
provide the infrastructure necessary to bring Internet   unlikely to take advantage of the availability of
connectivity to every household in town and in the       computers and the Internet if they do not consider
surrounding rural area. This project, called the         computer-related activities useful and do not have
Blacksburg Electronic Village (BEV) has had several      social support for such activities from people they
goals, including expanding the local economy             know, especially their peers. This is particularly
through the promotion of high-tech industry, in-         apparent in wealthier nations such as the United
creasing citizen access to online resources, and         States, where the usefulness of and accessibility to
promoting a stronger sense of community. Recent          online resources is taken for granted by more af-
evaluations by project leaders indicate that BEV         fluent members of society but where such online re-
has been more successful in the first two areas than      sources are less likely to be perceived as desirable
in the third.                                            by members of less affluent communities. To address
                                                         that problem, several nonprofit groups in the United
Access Issues                                            States have set up community computing centers in
As the BEV project recognized, in order to partici-      poorer neighborhoods, where they provide both train-
pate in cybercommunities, people need access to          ing in necessary computer skills and a community-
computers and to computer networks, especially           based context for valuing such skills.
to the Internet and the World Wide Web. Although             Another approach to broadening community ac-
such access has been expanding rapidly, people in        cess to the Internet has been to integrate Internet
poorer nations and disadvantaged populations in          connections into the construction of new build-
more affluent countries still have limited access to      ings or entire neighborhoods. However, these
the Internet, if they have it at all.                    types of developments also benefit only those who
    This issue has been particularly salient for com-    can afford to buy into them.
munity networks, which are often created with the
specific goal of making it possible for disadvantaged     Interfaces
groups to access and influence their local govern-        The direct-brain interfaces envisioned by Gibson,
mental structures. Thus many community net-              if possible at all, are likely many years in the future
works, in addition to setting up websites, have          (although there have been some promising early ex-
provided publicly accessible terminals for the use       periments in recent years, including one in which a
of those who do not have access to computers at          blind person was given partial sight through a video
home or at work.                                         feed wired to the optical nerve). Most people cur-
                                                                                  CYBERCOMMUNITIES ❚❙❘ 137



rently access and participate in cybercommunity            dimensional virtual spaces, has slowed the develop-
through personal computers. Usually, these com-            ment of cybercommunities using three-dimensional
puters are connected to the Internet by a modem or         spaces and avatars. One such community, Active
other wired connection to an Internet service pro-         Worlds (introduced in 1995), provides a three-
vider. However, wireless services are increasing, and      dimensional view of the environment similar to those
in some countries, most notably Japan, cell phones         first used in first-person shooter computer games
are commonly used to access the Internet and to            (games in which you see on the screen what your
communicate textually with others. In other coun-          character sees, rather than watching your character
tries, including the United States, people are also be-    move about) such as Doom and Quake. Technical
ginning to use cell phones and personal digital            considerations, including the simple problem of the
assistants (PDAs) for these purposes.                      amount of “real estate” available on a computer
     Most communication in cybercommunities oc-            screen, meant that participants in the early years of
curs through text, although some forums use graph-         Active Worlds could see only the twelve closest
ics or voice communication, often supplemented by          avatars. This contrasts with text-only interactive
text. Some of the oldest existing cybercommunities         forums such as MUDs and chat, in which thirty to
are still text-only and therefore require a high level     fifty participants can be simultaneously involved
of literacy as well as comfort with computers. Early       in overlapping textual conversations. Graphical in-
text-based forums were not always particularly easy        terfaces provide both limitations and enhancements
to use, either. The WELL’s original interface was no-      to online communications.
toriously difficult to work with. This meant that only
those with an understanding of computers and a             Identity in Cybercommunities
strong interest in the possibilities of cybercommu-        Aside from the more technical aspects of interface
nity had the motivation and ability to participate.        design, cybercommunities have also had to grapple
Currently, The WELL has a much more accessible             with the question of self-representation. How do par-
Web interface and a concomitantly more diverse pop-        ticipants appear to one another? What can they know
ulation of users.                                          about one another at the outset, and what can they
     As available Internet bandwidth and computer          find out? How accountable are cybercommunity
processing speeds have increased, cybercommuni-            members for their words and behavior within the
ties are able to use graphical representations of people   virtual space?
and objects within the cyberspace. One of the ear-              In purely text-based systems such as chat forums
liest examples, from 1985, was Habitat, a role-playing     or MUDs, participants are generally expected to pro-
game and socializing space that emulated an of-            vide some sort of description or personal informa-
fline community. Habitat featured a local economy           tion, although on some systems it is understood that
(based on points rather than real money) and such          this information may be fanciful. On LamdaMOO,
social structures as a church and sheriff ’s office.       for instance, many participants describe themselves
Habitat used two-dimensional cartoon-like draw-            as wizards, animals, or creatures of light. However,
ings to represent people and objects within the            each participant is required to choose a gender for
forum. Many current graphical worlds also use flat          their character, partly in order to provide pronoun
cartoon-type representations. Habitat originated the       choice for text generated by the MUD program, but
use of the term avatar to refer to the representa-         also indicating the assumed importance of this
tion of people in such graphical worlds, and that term     aspect of identity. In a divergence from real life,
has persisted in most such systems.                        LambdaMOO provides ten choices for gender iden-
     The technical difficulties inherent in rendering       tification. Despite this, most participants choose
three-dimensional spaces through which characters          either male or female. LambdaMOO participants
can move and in which people can manipulate vir-           choose what other personal information they wish
tual objects, along with the high level of computer        to reveal. On other MUDs, especially those intended
processing power required to make possible three-          as professional spaces or as forums for discussions
138 ❘❙❚ BERKSHIRE ENCYCLOPEDIA OF HUMAN-COMPUTER INTERACTION



relating to “real life” (offline life), participants may   face that can recover well from erroneous input.
be required to provide e-mail addresses or real names.    Names and accountability are another crucial issue
     In graphical forums, participants are represented    for community forums. PEN found that anonymity
both by the textual information they provide              tended to facilitate and perhaps even encourage “flam-
about themselves and by their avatar. Design choices      ing” (caustic criticism or verbal abuse) and other an-
involved in avatar creation in different virtual spaces   tisocial behavior, disrupting and in some cases
often reveal important underlying social assump-          destroying the usefulness of the forums for others.
tions, as well as technical limitations. In the early
years on Active Worlds, for instance, participants        Conflict Management and Issues of Trust
were required to choose from a limited number of          Cybercommunities, like other types of communi-
existing predesigned avatars. In part this stemmed        ties, must find ways to resolve interpersonal conflicts
from the difficulties of rendering even nominally         and handle group governance. In the early years of
human-seeming avatars in the three-dimensional            the Internet, users of the Internet were primarily
space. However, the particular avatars available also     white, male, young, and highly educated; most were
revealed biases and assumptions of the designers. In      connected to academic, government, or military in-
contrast to MUDs such as LambdaMOO, all avatars           stitutions, or to computing-related businesses.
were human. At one point, participants exploited a        However, in the mid-1990s the Internet experienced
programming loophole to use other objects, such as        a great increase in participation, especially from
trees and walls, as personal representations, but this    groups who had previously been on private systems
loophole was quickly repaired by the designers, who       not connected to the Internet, notably America
felt strongly that human representations promoted         Online (AOL). This sudden change in population
better social interaction. Active Worlds’ avatars         and increase in diversity of participants created ten-
also displayed a very limited range of human vari-        sions in some existing cybercommunities.
ation. Most were white, and the few non-white avatars          In one now-famous Usenet episode in 1993, par-
available tended to display stereotypical aspects.        ticipants in a Usenet newsgroup called alt.tasteless,
For instance, the single Asian avatar, a male, used       a forum for tasteless humor frequented primarily by
kung-fu moves, the female avatars were all identifi-       young men, decided to stage an “invasion” of another
able by their short skirts, and the single black male     newsgroup, rec.pets.cats, whose participants, atypi-
avatar sported dreadlocks. Since then, programming        cally for Usenet newsgroups at the time, were largely
improvements and feedback from users has enabled          women, older than Usenet participants in general,
Active Worlds to improve their graphics (the avatars      and in many cases relatively new to the Internet. The
now have distinct facial features) and expand their       alt.tasteless participants flooded rec.pets.cats with
representational offerings.                               gross stories of cat mutilation and abuse, disrupting
     In two-dimensional graphical environments such       the usual discussions of cat care and useful infor-
as Worlds Away (introduced in 1995), variation            mation about cats. Some of the more computer-savvy
tended to be greater from the beginning, and par-         participants on rec.pets.cats attempted to deal
ticipants were given the ability to construct avatars     with the disruption through technical fixes such as
from components. They could even change avatar            kill files (which enable a participant to automatically
appearance at will by (for example) buying new heads      eliminate from their reading queue messages posted
from the “head shop.” In some systems, participants       by particular people), but this was difficult for par-
can also import their own graphics to further cus-        ticipants with less understanding of the somewhat
tomize their online self-representation.                  arcane Usenet system commands. The invaders,
     Cybercommunities with greater ties to offline        meanwhile, found ways around those fixes. The con-
communities also have to deal with interface and rep-     flict eventually spread to people’s offline lives, with
resentations issues. In order to provide community        some rec.pets.cats participants receiving physical
access to as wide a range of townspeople as possi-        threats, and at least one alt.tasteless participant hav-
ble, networks such as PEN need an easy-to-use inter-      ing their Internet access terminated for abusive be-
                                                                                   CYBERCOMMUNITIES ❚❙❘ 139



havior. Eventually, the invaders tired of their sport      present—something not really possible offline. As a
and rec.pets.cats returned to normal. Some news-           positive contribution, this command can allow com-
groups have sought to avoid similar problems by es-        munity members to discuss approaches to dealing
tablishing a moderator, a single person who must           with a disruptive participant. However, the com-
approve all contributions before they are posted to        mand can also have negative consequences.
the group. In high traffic groups, however, the task            The use of avatars in graphical forums presents
of moderation can be prohibitively time-consuming.         another set of potential conflicts. In the two-
     LambdaMOO experienced a dramatic popula-              dimensional space of Worlds Away, participants
tion surge in the late 1990s, causing not only social      found that they could cause another participant to
tensions, but also technical problems as the               completely disappear from view by placing their own
LambdaMOO computer program attempted to                    avatar directly on top of the other’s. With no avail-
process the increasing numbers of commands.                able technical fix for this problem, users had to
LambdaMOO community members had to come up                 counter with difficult-to-enforce social sanctions
with social agreements for slowing growth and for          against offenders.
limiting commands that were particularly taxing on
the server. For instance, they instituted a limit on the   Trust
numbers of new participants that could be added            The potential for conflicts in cybercommunities is
each day, started deleting (“reaping”) the characters      probably no greater than that in offline communi-
and other information of participants who had been         ties. On the one hand, physical violence is not pos-
inactive for several months, and set limits on the         sible online (although in theory escalating online
number of new virtual objects and spaces that par-         conflicts can lead to offline violence). On the other
ticipants could build. This created some tension as        hand, the difficulty in completely barring offend-
community members attempted to find fair ways to            ers from a site (since people can easily reappear us-
determine who would be allowed to build and how            ing a different e-mail address) and the inability to
much. The solution, achieved through vote by par-          otherwise physically enforce community standards
ticipants, was to create a review board elected by the     has increased cybercommunities’ vulnerability to
community that would reject or approve proposed            disruption. In some cases, the greater potential for
projects. Designers of cybercommunity forums have          anonymity or at least pseudonymity online has
also had to consider what types of capabilities to give    also facilitated antisocial behavior. Many cyber-
participants and what the social effects of those ca-      communities have therefore tried to find ways to en-
pabilities might be. For instance, Active Worlds orig-     hance trust between community members.
inally did not allow participants to have private               Some have sought to increase accountability by
conversations that were not visible to all other par-      making participants’ e-mail addresses or real life
ticipants in the same virtual space. The designers felt    names available to other participants. Others have
that such conversations were antisocial and might          set rules for behavior with the ultimate sanction be-
lead to conflicts. However, participants continued to       ing the barring of an individual from the forum
request a command that enabled such “whispered”            (sometimes technologically tricky to implement).
conversations, and also implemented other programs,        LambdaMOO, for instance, posts a set of rules for
such as instant messaging, in order to work around         polite behavior. Because it is also one of the most fa-
the forum’s limitations. The designers eventually ac-      mous (and most documented) cybercommunities,
quiesced and added a whisper command.                      LambdaMOO’s opening screen also displays rules of
     Similarly, some MUDs have a command known             conduct for journalists and academic researchers vis-
as mutter. Rather than letting you talk only to one        iting the site.
other person, as is the case with whisper, mutter lets          LiveJournal requires potential participants to ac-
you talk to everyone else in the virtual room except       quire a code from an existing user in order to become
a designated person; in other words, it enables you        a member, which it is hoped ensures that at least one
to talk behind a person’s back while that person is        person currently a member of the community
140 ❘❙❚ BERKSHIRE ENCYCLOPEDIA OF HUMAN-COMPUTER INTERACTION



vouches for the new member. LiveJournal is con-            vious prejudices might be left behind, enabling a
sidering abandoning this practice in favor of a            utopian meeting of minds and ideas.
complex system of interpersonal recommendations                So far, it appears that cybercommunities tend to
that give each participant a trust rating, theoretically   augment rather than supplant people’s other social
an indication of their trustworthiness and status          connections. They appear to contain many of the
within the community.                                      same positive and negative social aspects present
    Although perhaps not as complex, similar sys-          in offline communities. Further, many cybercom-
tems are in use at other online forums. Slashdot, a        munities emerge from existing offline groups, also
bulletin board service focusing primarily on com-          include an offline component (including face-to-
puter-related topics, allows participants to rank post-    face contact between at least some participants), or
ings and then to filter what they read by aggregated        utilize other technologies such as the telephone to
rank. A participant can, for instance, decide to           enhance connections. Whatever form cybercommu-
read only messages that achieve the highest aver-          nities take in the future, their presence and popu-
age rating, as averaged from the responses of other        larity from the earliest days of computer networks
participants.                                              makes it clear that such interconnections will con-
    The online auction site eBay has a feedback            tinue to be a significant part of human-computer
system through which buyers and sellers rate one an-       interaction.
other’s performance after each transaction, result-
ing in a numerical score for each registered member.                                                                Lori Kendall
Each instance of positive feedback bestows a point,
and each instance of negative feedback deletes one.        See also Avatars; Digital Divide; MUDs
A recent change in the way auctions are displayed
now lists a percentage of positive feedback for each
seller. Users can also read the brief feedback mes-        FURTHER READING
sages left for other users. These features are intended
to allow users to evaluate a person’s trustworthiness      Baym, N. K. (2000). Tune in, log on: Soaps, fandom, and online com-
                                                              munity. Thousand Oaks, CA: Sage.
prior to engaging in transactions with that person.        Belson, K., & Richtel, M. (2003, May 5). America’s broadband
    The degree to which these types of trust-                 dream is alive in Korea. The New York Times, p. C1.
promotion systems work to foster and enhance               Benedikt, M. (Ed.). (1992). Cyberspace: First steps. Cambridge, MA:
community is unclear. Participants in various cyber-          MIT Press.
                                                           Blackburg Electronic Village. (n.d.) About BEV. Retrieved August
communities continue to consider issues of trust and          12, 2003, from http://www.bev.net/about/index.php
to work on technological enhancements to the virtual       Cherny, L. (1999). Conversation and community: Chat in a virtual
environment that will help suppress antisocial be-            world. Stanford, CA: CSLI Publications.
                                                           Damer, B. (1998). Avatars! Berkeley, CA: Peachpit Press.
havior and promote greater community solidarity.           Dibbell, J. (1998). My tiny life: Crime and passion in a virtual world.
                                                              New York: Henry Holt and Company.
Future Directions                                          Gibson, W. (1984). Neuromancer. New York: Ace Books.
As cybercommunities first developed, mainstream             Hafner, K. (2001). The Well: A Story of love, death & real life in the sem-
                                                              inal online community. Berkeley, CA: Carroll & Graf.
media commentary discussed a variety of hyperbolic         Hampton, K. (2001). Living the wired life in the wired suburb: Netville,
fears and hopes. People feared that cybercommu-               glocalization and civil society. Unpublished doctoral dissertation,
nities would replace and supplant other forms of              University of Toronto, Ontario, Canada.
community and that cybercommunities were less              Herring, S. C., with D. Johnson & T. DiBenedetto. (1995). “This dis-
                                                              cussion is going too far!” Male resistance to female participation
civilized, with greater potential for rude and anti-          on the Internet. In M. Bucholtz & K. Hall (Eds.), Gender articu-
social behavior. On the other hand, people also hoped         lated: Language and the socially constructed self (pp. 67–96). New
that cybercommunities might provide forms of                  York: Routledge.
                                                           Jones, S. (Ed.). (1995). Cybersociety: Computer-mediated communica-
interconnectedness that had otherwise been lost in            tion and community. Thousand Oaks, CA: Sage.
modern life. Some people also suggested that cyber-        Jones, S. (Ed.). (1997). Virtual culture: Identity and communication in
communities could provide a forum in which pre-               cybersociety. London: Sage.
                                                                                                                 CYBERSEX ❚❙❘ 141



Kavanaugh, A., & Cohill, A. (1999). BEV research studies, 1995–               Internet, visiting sex-related websites, masturbating
   1998. Retrieved August 12, 2003, from http://www.bev.net/about/            to sexual media from the Internet, engaging in sex-
   research/digital_library/docs/BEVrsrch.pdf
Kendall, L. (2002). Hanging out in the virtual pub. Berkeley, CA:
                                                                              ualized videoconferencing activities, creating sexual
   University of California Press.                                            materials for use/distribution on the Internet, and
Kiesler, S. (1997). Culture of the Internet. Mahwah, NJ: Lawrence             using the Internet to obtain/enhance offline sexual
   Erlbaum Associates.                                                        behaviors.
McDonough, J. (1999). Designer selves: Construction of technologically-
   mediated identity within graphical, multi-user virtual environ-                A broader term used to describe Internet sex-
   ments. Journal of the American Society for Information Science, 50(10),    ual behavior is “online sexual activity” (OSA), which
   855–869.                                                                   includes using the Internet for any sexual purpose,
McDonough, J. (2000). Under construction. Unpublished doctoral dis-
   sertation, University of California at Berkeley.
                                                                              including recreation, entertainment, exploration, or
Morningstar, C., & Farmer, F. R. (1991). The lessons of Lucasfilm’s            education. Examples of OSA are using online ser-
   Habitat. In M. Benedikt (Ed.), Cyberspace: First steps (pp. 273–302).      vices to meet individuals for sexual /romantic pur-
   Cambridge, MA: The MIT Press.                                              poses, seeking sexual information on the Internet
Porter, D. (1997). Internet culture. New York: Routledge.
Renninger, K. A., & Shumar, W. (Eds.). (2002). Building virtual com-
                                                                              (for instance, about contraception and STDs), and
   munities. Cambridge, UK: Cambridge University Press.                       purchasing sexual toys/paraphernalia online. What
Rheingold, H. (1993). The virtual community: Homesteading on the              distinguishes cybersex from OSA is that cybersex
   electronic frontier. Reading, MA: Addison-Wesley.                          involves online behaviors that result in sexual arousal
Smith, M., & Kollock, P. (Eds.). (1999). Communities and cyber-
   space. New York: Routledge.                                                or gratification, while other online sexual activities
Taylor, T. L. (2002). Living digitally: Embodiment in virtual worlds.         may lead to offline sexual arousal and gratifica-
   In R. Schroeder (Ed.), The social life of avatars: Presence and inter-     tion. Sexual arousal from cybersex is more immedi-
   action in shared virtual environments. London: Springer Verlag.
Turkle, S. (1995). Life on the screen: Identity in the age of the Internet.
                                                                              ate and is due solely to the online behavior.
   New York: Simon & Schuster.
Wellman, B. (2001). The persistence and transformation of community:
   From neighbourhood groups to social networks. Report to the Law            Venues
   Commission of Canada. Retrieved August 12, 2003, from http://              Many people assume that the World Wide Web is the
   www.chass.utoronto.ca/~wellman/publications/lawcomm/                       main venue for cybersex. In fact, the Web repre-
   lawcomm7.htm
Wellman, B., & Haythornthwaite, C. (Eds.). (2002). The Internet in
                                                                              sents only a small portion of the places where cyber-
   everyday life. Oxford, UK: Blackwell.                                      sex activities can occur. Other areas of the Internet
Wellman, B., Boase, J., & Chen, W. (2002). The networked nature of            where cybersex may take place include the following:
   community online and offline. IT & Society, 1(1), 151–165.
Weiner, N. (1948). Cybernetics, or control and communication in the             ■ Newsgroups— This area serves as a bulletin
   animal and the machine. Cambridge, MA: MIT Press.                              board where individuals can post text or multi-
WELL, The. (2002). About the WELL. Retrieved August, 2003, from
   http://www.well.com/aboutwell.html
                                                                                  media messages, such as sexual text, pictures,
                                                                                  sounds, and videos;
                                                                                ■ E-mail—E-mail can be used for direct com-
                                                                                  munication with other individuals or groups
                                                                                  of individuals. In the case of cybersex, the mes-
                      CYBERSEX                                                    sage may be a sexual conversation, story, picture,
                                                                                  sound, or video;
The term cybersex is a catch-all word used to describe                          ■ Chatrooms—Both sexualized conversation
various sexual behaviors and activities performed                                 and multimedia can be exchanged in chatrooms.
while on the Internet. The term does not indicate                                 Casual users are familiar with Web-based chat-
that a particular behavior is good or bad, only that                              ting such as Yahoo Chat or America Online
the sexual behavior occurred in the context of the                                (AOL) Chat. Most Web-based chat areas have
Internet. Examples of behaviors or activities that may                            sections dedicated to sexual chats. However,
be considered cybersex include sexual conversations                               the largest chat-based system is the Internet Relay
in Internet chatrooms, retrieving sexual media (for                               Chat (IRC), an area largely unfamiliar to most
example, photographs, stories, or videos) via the                                 casual users. In addition to text-based chatting,
142 ❘❙❚ BERKSHIRE ENCYCLOPEDIA OF HUMAN-COMPUTER INTERACTION



    IRC contains a number of chatrooms specifically         that from 11 to 17 percent of individuals who en-
    dedicated to the exchange of pornography               gaged in cybersex report some consequences in their
    through “file servers”;                                 life and score moderately high on measures of gen-
  ■ Videoconferencing/Voice Chatting—The use of            eral sexual compulsivity. In addition, approximately
    these areas is rapidly increasing. As technology       6 percent report feeling out of control with their
    improves and connection speeds increase, the           Internet sexual behavior and scored high on mea-
    use of the Internet for “live” cybersex sessions       sures of sexual compulsivity.
    will become commonplace. Videoconferencing
    combined with voice chat constitutes a high-tech
    version of a peep show mixed with an obscene           Healthy Versus
    phone call; and
  ■ Peer-to-Peer File Sharing—Software packages            Problematic Cybersex
    such as Napster and Kazaa have made file shar-          One of the difficulties in defining cybersex as
    ing a popular hobby. Casual users of this soft-        either healthy or problematic is the fact that there
    ware know its use for exchanging music files, but       are few agreed-upon definitions about what con-
    any file can be shared on the network, including        stitutes sexually healthy behavior. Society has clearly
    sexual images, sounds, and videos.                     delineated some behaviors as unhealthy, for exam-
                                                           ple, sex with children or other non-consenting part-
                                                           ners. However, people disagree about whether
Statistics                                                 masturbation, multiple affairs, bondage, and fetishes
Although the term cybersex often has negative con-         are healthy or unhealthy. In the world of cybersex,
notations, research in this area suggests that             these same gray areas exist between healthy and un-
nearly 80 percent of individuals who engage in             healthy and are often even more difficult to define
Internet sex report no significant problems in their        since the behavior does not include actual sexual
lives associated with their online sexual activities.      contact.
Although this may be an underestimate since the re-             It is also important not to assume that frequency
search relied on the self-reports of respondents, it       is the key factor in determining whether an indi-
is safe to assume that the majority of individuals         vidual is engaged in unhealthy cybersex. Some indi-
who engage in cybersex behavior report this activ-         viduals engage in cybersex at a high frequency and
ity to be enjoyable and pleasurable, with few nega-        have few problems, while others who engage in it
tive consequences.                                         only a few hours a week have significant negative
     However, there are individuals who engage in          consequences.
cybersex who do report significant negative conse-               Physician and researcher Jennifer Schneider pro-
quences as a result of their online sexual behavior.       posed three criteria to help determine if some-
These individuals often report that their occupa-          one’s behavior has become compulsive—that is,
tional, social, or educational life areas have been neg-   whether the person has crossed the line from a
atively impacted or are in jeopardy as a result of their   “recreational” to a “problematic” user of cyber-
sexual use of the Internet. Often these individuals        sex. The three criteria are (1) loss of freedom to
report a sense of being out of control or compulsive       choose whether to stop the behavior; (2) negative
in their sexual use of the Internet and often compare      consequences as a result of the behavior; and (3) ob-
it to addictions like gambling, eating, shopping, or       sessive thinking about engaging in the behavior. The
working.                                                   Internet Sex Screening Test (ISS) described by coun-
     Several large-scale studies estimate the per-         seling professor David Delmonico and professor of
centage of individuals who are negatively impacted         school psychology Jeffrey Miller can be used to con-
by cybersex behaviors. While exact numbers are im-         duct initial screening of whether an individual has
possible given the size of the Internet, estimates are     a problem with cybersex.
                                                                                                                 CYBERSEX ❚❙❘ 143



The Appeal of the Internet
With an estimated 94 million users accessing it                                  Cybersex Addiction
regularly, it is difficult to dispute the Internet’s wide-

                                                                 T
                                                                        he Center for Online and Internet Addiction (www
spread appeal. In 2001 Delmonico, Moriarity, and                        .netaddiction.com) offers the following test to help
marriage and family therapist Elizabeth Griffin, pro-                    diagnose cybersex addiction:
posed a model called “the Cyberhex” for under-
                                                                  1. Do you routinely spend significant amounts of time
standing why the Internet is so attractive to its users.
                                                                     in chat rooms and private messaging with the sole
Their model lists the following six characteristics:
                                                                     purpose of finding cybersex?
     Integral: The Internet is nearly impossible to avoid.
                                                                  2. Do you feel preoccupied with using the Internet to
Even if a cybersex user decided to never use the Internet
                                                                     find on-line sexual partners?
again, the integral nature of the Internet would make
                                                                  3. Do you frequently use anonymous communica-
that boundary nearly impossible, since many need the
                                                                     tion to engage in sexual fantasies not typically car-
Internet for work, or to access bank information, and
                                                                     ried out in real-life?
so on. In addition, public availability, the use of e-mail,
                                                                  4. Do you anticipate your next on-line session with the
and other activities like shopping and research make
                                                                     expectation that you will find sexual arousal or
the Internet a way of life that is integrated into our daily
                                                                     gratification?
routines.
                                                                  5. Do you find that you frequently move from cyber-
     Imposing: The Internet provides an endless sup-
                                                                     sex to phone sex (or even real-life meetings)?
ply of sexual material 7 days a week, 365 days a year.
                                                                  6. Do you hide your on-line interactions from your sig-
The amount of information and the imposing nature
                                                                     nificant other?
of marketing sexual information on the Internet
                                                                  7. Do you feel guilt or shame from your on-line use?
contributes to the seductiveness of the world of cybersex.
                                                                  8. Did you accidentally become aroused by cybersex at
     Inexpensive: For a relatively small fee, twenty to
                                                                     first, and now find that you actively seek it out when
forty dollars per month, a user can access an intoxi-
                                                                     you log on-line?
cating amount of sexual material on the Internet. In
                                                                  9. Do you masturbate while on-line while engaged in
the offline world such excursions can be cost-prohibitive
                                                                     erotic chat?
to many.
                                                                 10. Do you provide less investment with your real-life
     Isolating: Cybersex is an isolating activity. Even
                                                                     sexual partner only to prefer cybersex as a primary
though interpersonal contact may be made during the
                                                                     form of sexual gratification?
course of cybersex, these relationships do not require
                                                                 Source: Are you addicted to cybersex. Center for Online and Internet
the same level of social skills or interactions that offline      Addiction. Retrieved March 23, 2004, from http://www.netaddiction.com/
behaviors require. The Internet becomes a world in it-           resources/cybersexual_addiction_test.htm

self, where it is easy to lose track of time, consequences,
and real-life relationships. The isolation of cybersex of-
ten provides an escape from the real world, and while
everyone takes short escapes, cybersex often becomes                Intoxicating: This is what happens when the pre-
the drug of choice to anesthetize any negative feelings        ceding five elements are added together. This com-
associated with real-life relationships.                       bination makes for an incredibly intoxicating
     Interactive: While isolating in nature, the Internet      experience that is difficult for many to resist. The
also hooks individuals into pseudorelationships. These         intoxication of the Internet is multiplied when cyber-
pseudorelationships often approximate reality with-            sex is involved since behaviors are reinforced with
out running the risks of real relationships—like emo-          one of the most powerful rewards, sex.
tional and physical vulnerability and intimacy. This                Any single aspect of the Internet can be pow-
close approximation to reality can be fuel for the fan-        erful enough to entice a cybersex user. However,
tasy life of those who experience problems with their          it is typically a combination of these six factors that
cybersex behaviors.                                            draws problematic cybersex users into their rituals
144 ❘❙❚ BERKSHIRE ENCYCLOPEDIA OF HUMAN-COMPUTER INTERACTION



and leads to their loss of control over their cyber-         be useful in educating children and adolescents about
sex use.                                                     sexuality, it can also be a dangerous venue for the de-
                                                             velopment of compulsive behavior and victimiza-
                                                             tion by online predators. Although the effect of
Special Populations                                          hardcore, explicit pornography on the sexual devel-
                                                             opment of children and adolescents has yet to be re-
Engaged in Cybersex                                          searched, early exposure to such pornography may
The following subgroups of cybersex users have been          impact their moral and sexual development.
studied in some detail:                                          Physically or Developmentally Challenged
     Males and Females: In the early to mid-1990s            People: Only recently have questions been raised
there were three times as many males online as fe-           about the appropriate use of the Internet for sex-
males. Recent research shows that the gap has closed         ual and relational purposes among physically chal-
and that the split between male and female Internet          lenged individuals. This area warrants more research
users is nearly fifty-fifty. As a result, research on cyber-   and exploration, but initial writings in this area sug-
sex behavior has also included a significant number           gest that the Internet can confer a tremendous ben-
of females who engage in cybersex. Most of this re-          efit for sexual and relationship exploration for
search suggests that men tend to engage in more vi-          persons with disabilities.
sual sex (for example, sexual media exchange), while             While sex on the Internet can be a positive ex-
women tend to engage in more relational sex (for             perience for these subpopulations, it can also intro-
example, chatrooms and e-mail). Females may                  duce the people in these groups to the same problems
find the Internet an avenue to sexual exploration and         associated with cybersex that other groups report.
freedom without fear of judgment or reprisal from
society. In this way, the Internet can have genuine
benefits.                                                     Implications
     Gays and Lesbians: Researchers have reported            Cybersex is changing sexuality in our culture. The
that homosexuals tend to engage in cybersex at higher        positive side is that sexual behavior is becoming more
levels than heterosexuals, which may be because they         open and varied, and better understood. The nega-
don’t have to fear negative cultural responses or even       tive implications are that sexuality may become ca-
physical harm when they explore sexual behaviors             sual, trivial, and less relational.
and relationships on the Internet. Some homosexu-                The pornography industry continues to take ad-
als report that cybersex is a way to engage in sexual        vantage of the new technologies with the primary
behavior without fear of HIV or other sexually trans-        goal of profit, and these new technologies will allow
mitted diseases. By offering homosexuals a safe              for faster communication to support better video
way to explore and experience their sexuality, the           and voice exchanges. The eventual development of
Internet gives them freedom from the stigma often            virtual reality technologies online will further en-
placed on them by society.                                   hance the online sexual experience, and perhaps make
     Children and Adolescents: Studies conducted by          the sexual fantasy experience more pleasurable than
AOL and Roper Starch revealed that children use the          real life. These technological advances will continue
Internet not only to explore their own sexuality and         to alter the way we interact and form relationships
relationships, but also to gather accurate sexual health     with others.
information. Since many young adults have grown                  Researchers are just starting to realize the impli-
up with the Internet, they often see it through a            cations of sex on the Internet. Theories like Cyberhex
different lens than adults. Children, adolescents, and       are helpful in understanding why people engage in
young adults use the Internet to seek answers to a           cybersex, but the best methods for helping those strug-
multitude of developmental questions, including              gling with cybersex have yet to be discovered. However,
sexuality, which they may be afraid to address di-           society will continue to be impacted by the Internet
rectly with other adults. Although the Internet can          and cybersex. Parents, teachers, and others who
                                                                                                                CYBORGS ❚❙❘ 145



have not grown up with the Internet will fail future
generations if they discount the significant impact it                                       CYBORGS
can have on social and sexual development. Continued
research and education will be necessary to help in-                       A cyborg is a technologically enhanced human be-
dividuals navigate the Internet and the world of cyber-                    ing. The word means cybernetic organism. Because
sex more safely.                                                           many people use the term cybernetics for computer
                                                                           science and engineering, a cyborg could be the fu-
           David L. Delmonico and Elizabeth J. Griffin                      sion of a person and a computer. Strictly speaking,
                                                                           however, cybernetics is the science of control pro-
See also Chatrooms; Cybercommunities                                       cesses, whether they are electronic, mechanical, or
                                                                           biological in nature. Thus, a cyborg is a person, some
                                                                           of whose biological functions have come under tech-
FURTHER READING                                                            nological control, by whatever means. The standard
                                                                           term for a computer-simulated person is an avatar,
Carnes, P. J. (1983). Out of the shadows. Minneapolis, MN: CompCare.       but when an avatar is a realistic copy of a specific real
Carnes, P. J., Delmonico, D. L., Griffin, E., & Moriarity, J. (2001). In
   the shadows of the Net: Breaking free of compulsive online behavior.
                                                                           person, the term cyclone is sometimes used, a cyber-
   Center City, MN: Hazelden Educational Materials.                        netic clone or virtual cyborg.
Cooper, A. (Ed.). (2000). Sexual addiction & compulsivity: The jour-
   nal of treatment and prevention. New York: Brunner-Routledge.
Cooper, A. (Ed.). (2002). Sex and the Internet: A guidebook for clini-
   cians. New York: Brunner-Routledge.                                     Imaginary Cyborgs
Cooper, A., Delmonico, D., & Burg, R. (2000). Cybersex users, abusers,     The earliest widely known cyborg in literature,
   and compulsives: New findings and implications. Sexual Addiction         dating from the year 1900, is the Tin Woodman in
   and Compulsivity: Journal of Treatment and Prevention, 7, 5–29.
Cooper, A., Scherer, C., Boies, S. C., & Gordon, B. (1999). Sexuality on
                                                                           The Wonderful Wizard of Oz by L. Frank Baum.
   the Internet: From sexual exploration to pathological expres-           Originally he was a man who earned his living chop-
   sion. Professional Psychology: Research and Practice, 30(2), 154–164.   ping wood in the forest. He wanted to marry a beau-
Delmonico, D. L. (1997). Internet sex screening test. Retrieved            tiful Munchkin girl, but the old woman with whom
   August 25, 2003, from http://www.sexhelp.com/
Delmonico, D. L., Griffin, E. J., & Moriarity, J. (2001). Cybersex un-
                                                                           the girl lived did not want to lose her labor and pre-
   hooked: A workbook for breaking free of compulsive online behavior.     vailed upon the Wicked Witch of the East to enchant
   Wickenburg, AZ: Gentle Path Press.                                      his axe. The next time he went to chop wood, the axe
Delmonico, D. L., & Miller, J. A. (2003). The Internet sex screening       chopped off his left leg instead. Finding it incon-
   test: A comparison of sexual compulsives versus non-sexual com-
   pulsives. Sexual and Relationship Therapy, 18(3), 261–276.              venient to get around without one of his legs, he went
Robert Starch Worldwide, Inc. (1999). The America Online/Roper             to a tinsmith who made a new one for him. The
   Starch Youth Cyberstudy. Author. Retrieved on December 24, 2003,        axe then chopped off his right leg, which was then
   from http://www.corp.aol.com/press/roper.html/
Schneider, J. P. (1994). Sex addiction: Controversy within mainstream
                                                                           also replaced by one made of tin. This morbid process
   addiction medicine, diagnosis based on the DSM-III-R and physi-         continued until there was nothing left of the origi-
   cian case histories. Sexual Addiction & Compulsivity: The Journal       nal man but his heart, and when that finally was
   of Treatment and Prevention, 1(1), 19–44.                               chopped out, he lost his love for the Munchkin girl.
Schneider, J. P., & Weiss, R. (2001). Cybersex exposed: Recognizing the
   obsession. Center City, MN: Hazelden Educational Materials.
                                                                           Still, he missed the human emotions that witchcraft
Tepper, M. S., & Owens, A. (2002). Access to pleasure: Onramp to           and technology had stolen from him, and was ready
   specific information on disability, illness, and other expected         to join Dorothy on her journey to the Emerald City,
   changes throughout the lifespan. In A. Cooper (Ed.), Sex and            on a quest for a new heart.
   the Internet: A guidebook for clinicians. New York: Brunner-
   Routledge.                                                                   This story introduces one of the primary themes
Young, K. S. (1998). Caught in the Net. New York: Wiley.                   associated with cyborgs: the idea that a person ac-
Young, K. S. (2001). Tangled in the web: Understanding cybersex from       cepts the technology to overcome a disability. That
   fantasy to addiction. Bloomington, IN: 1st Books Library.
                                                                           is, the person is already less than complete, and the
                                                                           technology is a substitute for full humanity, albeit an
                                                                           inferior one. This is quite different from assimilating
146 ❘❙❚ BERKSHIRE ENCYCLOPEDIA OF HUMAN-COMPUTER INTERACTION



new technology in order to become more than hu-               become its enemy. This reflects a second persistent
man, a motive severely criticized by the President’s          literary theme associated with cyborgs: They re-
Council on Bioethics in 2003. A very different                flect the evils of an oppressive society in which tech-
viewpoint on what it means to be “disabled” has been          nology has become a tool by which the masters
expressed by Gregor Wolbring, a professor at the              enslave the majority.
University of Calgary. Who decides the meanings                   By far the most extensive treatment of the idea
of disability and normality is largely a political issue,     that cyborg technology is wicked can be found in the
and Wolbring argues that people should generally              Dalek menace from the long-running BBC televi-
have the power to decide for themselves. He notes the         sion series, Dr. Who. Sometimes mistaken for robots,
example of children who are born without legs be-             Daleks are metal-clad beings that resemble huge salt
cause their mothers took thalidomide during preg-             shakers, wheeled trash cans, or British post boxes.
nancy, then forced to use poorly designed artificial           They became extremely popular villains since their
legs because that makes them look more normal,                first appearance in 1963. Two low-budget feature
when some other technology would have given them              films that retold the first TV serials added to their
far better mobility.                                          fame, Dr. Who and the Daleks (1965) and Dr. Who:
                                                              Daleks Invasion Earth 2150 A.D. (1966). Inside a
                                                              Dalek’s metal shell lurks a helpless, sluggish creature
  C makes it easy to shoot yourself in the foot. C++ makes    with vestigial claws, yet the combination of biol-
  it harder, but when you do, it blows away your whole leg.   ogy and technology gave it the possibility of con-
                                          Bjarne Stroustrup   quering the universe. Their motto describes how they
                                                              treat all other living creatures: “Exterminate.” The
                                                              secret of their origins is revealed in the 1975 serial,
     A common variation on the disability theme is            “Genesis of the Daleks.” The protagonist of Dr. Who,
the hero who suffers a terrible accident, is rebuilt,         The Doctor, lands his time machine on the battle-
and becomes a cyborg superhero. A well-known ex-              scarred planet Skaro, just as the nuclear war between
ample is The Six Million Dollar Man, a television se-         the Thals and the Kaleds reaches its climax. Davros,
ries that aired 1973–1978 and was based on the 1972           the evil (and disabled) Kaled scientist, recognizes
novel Cyborg by Martin Caidin. Test pilot Steve Austin        that chemical weapons are causing his people to mu-
is severely injured in a plane crash, then rebuilt with       tate horribly, and rather than resist this trend, he
bionic (biological plus electronic) technology. A spin-       accelerates it, transforming humans into the vile
off series, The Bionic Woman (1976–1978) focuses              Dalek cyborgs.
on tennis player Jaime Sommers who is similarly dis-
abled in a parachute accident. Both become super-
hero special agents, perhaps to justify the heavy             Real Cyborg Research
investment required to insert and maintain their              Since human beings began wearing clothing, the
bionics. An especially striking example is the motion         boundary between ourselves and our technology has
picture Robocop (1987). Policeman Alex Murphy lives           blurred. Arguably, everybody who wears a wristwatch
in a depressing future Detroit, dominated by a single,        or carries a cell phone is already a cyborg. But the
exploitative corporation. To control the increasingly         usual definition implies that a human body has been
violent population, the corporation develops robot            modified, typically by insertion of some nonbiological
police possessing overwhelming firepower but lack-             technology. In the early years of the twentieth cen-
ing the judgment to interact successfully with human          tury, when surgeons first gained technological
beings. Thus, when Murphy is blown to pieces by               control over pain and infection, many brave or irre-
criminals, the corporation transforms him into a cy-          sponsible doctors began experimenting with im-
borg that combines human judgment with machine                provements to their patients. Sir William Arbuthnot
power. The corporation denies Murphy the right                Lane, the British royal physician, theorized that many
to be considered human, thereby forcing him to                illnesses were caused by a sluggish movement of food
                                                                                              CYBORGS ❚❙❘ 147



through the bowels that supposedly flooded the sys-       safely, and how to interface active or sensory com-
tem with poisonous toxins. Diagnosing this chronic       ponents to the human nervous system. Several re-
intestinal stasis in many cases, Lane performed          searchers, such as Miguel Nicolelis of Duke
surgery to remove bands and adhesions, and free the      University, have been experimenting with brain im-
intestines to do their job. Some of his colleagues op-   plants in monkeys that allow them to operate artifi-
erated on neurotic patients, believing that moving       cial arms, with the hope that this approach could be
the abdominal organs into their proper places could      applied therapeutically to human beings in the near
alleviate mental disorders. Later generations of doc-    future.
tors abandoned these dangerous and useless pro-
cedures, but one of Lane’s innovations has persisted.
He was the first to “plate” a bone—that is to screw       Visions of the Future
a supportive metal plate onto a broken bone. Today       Kevin Warwick, professor of Cybernetics at Reading
many thousands of people benefit from artificial hip       University in Britain, is so convinced of the near-term
and knee joints.                                         prospects for cyborg technology, that he has experi-
    In World War I, even before the introduction         mented on his own body. In 1998, he had surgeons
of antibiotics, rigorous scientific techniques were      implant a transponder in his left arm so a computer
sufficiently effective to prevent death from infection    could monitor his movements. His first implant
in most wounded cases, thereby vastly increasing the     merely consisted of a coil that picked up power from
number of people who survived with horrendous            a transmitter and reflected it back, letting the com-
war-caused disabilities. The Carrel-Dakin technique      puter know where he was so it could turn on lights
was especially impressive, employing an antiseptic       when he entered a room. In 2002 he had neuro-
solution of sodium hypochlorite in amazingly rig-        surgeons connect his nervous system temporarily
orous procedures. Suppose a soldier’s leg had been       to a computer for some very modest experiments, but
badly torn by an artillery shell. The large and irreg-   in the future he imagines that implants interfacing
ular wound would be entirely opened up and cleaned.      between computers and the human nervous system
Then tubes would be placed carefully in all parts of     will allow people to store, playback, and even share
the wound to drip the solution very slowly, for          experiences. He plans someday to experiment with
days and even for weeks. Daily, a technician takes       the stored perceptions associated with drinking wine,
samples from every part of the wound, examining          to see if playing them back really makes him feel
them under the microscope, until no more microbes        drunk. His wife, Irena, has agreed that someday
are seen and the wound can be sewed up. Restorative      they both will receive implants to share feelings
plastic surgery and prosthetics could often help the     such as happiness, sexual arousal, and even pain. In
survivors live decent lives.                             the long run, Warwick believes, people will join
    In the second half of the twentieth century, much    with their computers to become superhuman cyborgs.
progress was achieved with transplants of living         In so doing, they will adopt a radically new concep-
tissue—such as kidneys from donors and coronary          tion of themselves, including previously unknown
artery bypass grafts using material from the patient.    understandings, perceptions, and desires.
Inorganic components were also successfully in-               Natasha Vita-More, an artist and futurist, has
troduced, from heart values to tooth implants.           sketched designs for the cyborg posthuman she
Pacemakers to steady the rhythm of the heart and         calls Primo, based on aesthetics and general techno-
cochlear transplants to overcome deafness are among      logical trends. Although she is not building proto-
the relatively routine electronic components inserted    types or experimenting with components at the
into human bodies, and experiments are being             present time, she believes that her general vision could
carried out with retina chips to allow the blind to      be achieved within this century. Primo would be age-
see. There are many difficult technical challenges,       less rather than mortal, capable of upgrades when-
notably how to power artificial limbs, how to connect     ever an organ wore out or was made obsolete by
large components to the structure of the human body      technical progress, and able to change gender
148 ❘❙❚ BERKSHIRE ENCYCLOPEDIA OF HUMAN-COMPUTER INTERACTION



whenever (s)he desires. Nanotechnology would give          technology takes me above and beyond myself. A per-
Primo 1,000 times the brainpower of a current hu-          son who is thoroughly plugged in experiences radi-
man, and thus capable of running multiple viewpoints       cally transformed consciousness: “I construct, and I
in parallel rather than being locked into one narrow       am constructed, in a mutually recursive process
frame of awareness. Primo’s senses would cover a           that continually engages my fluid, permeable bound-
vastly wider bandwidth, with sonar mapping onto the        aries and my endlessly ramifying networks. I am a
visual field at will, an internal grid for navigating and   spatially extended cyborg” (Mitchell 2003, 39).
moving anywhere like an acrobatic dancer with
perfect sense of direction, and a nervous system                                                Williams Sims Bainbridge
that can transmit information from any area of the
body to any other instantly. Primo’s nose could iden-
tify any chemical or biological substance in the en-       FURTHER READING
vironment, and smart skin will not only protect the
                                                           Bainbridge, W. S. (1919). Report on medical and surgical developments
body, but provide vastly enhanced sensations. Instead         of the war. Washington, DC: Government Printing Office.
of the depression and envy that oppress modern             Barnes, B. A. (1977). Discarded operations: Surgical innovation by
humans, (s)he would be filled with ecstatic yet real-          trial and error. In J. P. Bunker, B. A. Barnes, & F. Mosteller (Eds.),
istic optimism. The old fashioned body’s need to elim-        Costs, risks, and benefits of surgery (pp. 109–123). New York: Oxford
                                                              University Press.
inate messy wastes will be transcended by Primo’s          Baum, L. F. (1900). The wonderful wizard of Oz. Chicago: G. M. Hill.
ability to recycle and purify.                             Bentham, J. (1986). Doctor Who: The early years. London: W. H. Allen.
     William J. Mitchell, the director of Media Arts and   Caidin, M. (1972). Cyborg. New York: Arbor House.
                                                           Haining, P. (Ed.). (1983). Doctor Who: A celebration. London:
Sciences at the Massachusetts Institute of Technology,        W. H. Allen.
argues that we have already evolved beyond tradi-          Mitchell, W. J. (2003). ME++: The cyborg self and the networked city.
tional homo sapiens by become embedded in a                   Cambridge, MA: MIT Press.
ubiquitous communication network. The title of             Nicolelis, M. A. L., & Srinivasan, M. A. (2003). Human-machine
                                                              interaction: Potential impact of nanotechnology in the design of
his book, ME++: The Cyborg Self and the Networked             neuroprosthetic devices aimed at restoring or augmenting human
City (2003), offers a nice metaphor derived from              performance. In M. C. Roco & W. S. Bainbridge (Eds.), Converging
the C language for programming computers. C was               technologies for improving human performance (pp. 251–255).
originally developed by a telephone company (Bell             Dordrecht, Netherlands: Kluwer.
                                                           President’s Council on Bioethics. (2003). Beyond therapy: Biotechnology
Labs) and has become possibly the most influential             and the pursuit of happiness. Washington, DC: President’s Council
language among professional programmers, especially           on Bioethics.
in the modular version called C++. In C (and in the        Warwick, K. (2000). Cyborg 1.0. Wired, 8(2), 144–151.
                                                           Wolbring, G. (2003). Science and technology and the triple D (Disease,
Java language as well), “++” means to increment a             Disability, Defect). In M. C. Roco & W. S. Bainbridge (Eds.),
number by adding 1 to it. Thus, C++ is one level more         Converging technologies for improving human performance
than C, and ME++ is one level more than me, in which          (pp. 232–243). Dordrecht, Netherlands: Kluwer.
                                                                                            DATA MINING

                                                                                     DATA VISUALIZATION

                                                                                              DEEP BLUE

                                                                               DENIAL-OF-SERVICE ATTACK

                                                                                     DESKTOP METAPHOR




            D
                                                                                        DIALOG SYSTEMS

                                                                                           DIGITAL CASH

                                                                                          DIGITAL DIVIDE

                                                                                    DIGITAL GOVERNMENT

                                                                                       DIGITAL LIBRARIES

                                                                                    DRAWING AND DESIGN




                                                           future data samples. Only then can the rules and mod-
            DATA MINING                                    els obtained be considered meaningful. The discov-
                                                           ered patterns should also be novel, that is, not already
Data mining is the process of automatic discovery          known to experts; otherwise, they would yield very
of valid, novel, useful, and understandable patterns,      little new understanding. Finally, the discoveries
associations, changes, anomalies, and statistically sig-   should be useful as well as understandable.
nificant structures from large amounts of data. It is            Typically data mining has two high-level goals:
an interdisciplinary field merging ideas from sta-         prediction and description. The former answers the
tistics, machine learning, database systems and data-      question of what and the latter the question of
warehousing, and high-performance computing, as            why. For prediction, the key criterion is the accuracy
well as from visualization and human-computer in-          of the model in making future predictions; how
teraction. It was engendered by the economic and           the prediction decision is arrived at may not be
scientific need to extract useful information from          important. For description, the key criterion is the
the data that has grown phenomenally in all spheres        clarity and simplicity of the model describing the
of human endeavor.                                         data in understandable terms. There is sometimes
     It is crucial that the patterns, rules, and models    a dichotomy between these two aspects of data min-
that are discovered be valid and generalizable not only    ing in that the most accurate prediction model for
in the data samples already examined, but also in          a problem may not be easily understandable, and the

                                                                                                              149
150 ❘❙❚ BERKSHIRE ENCYCLOPEDIA OF HUMAN-COMPUTER INTERACTION



most easily understandable model may not be highly         mining algorithms—association rule discovery, se-
accurate in its predictions.                               quence mining, classification tree induction, clus-
                                                           tering, and so on—to analyze the data.
                                                               Interpret, evaluate, and visualize patterns: After
Steps in Data Mining                                       the algorithms have produced their output, it is still
Data mining refers to the overall process of discov-       necessary to examine the output in order to inter-
ering new patterns or building models from a               pret and evaluate the extracted patterns, rules, and
given dataset. There are many steps involved in the        models. It is only by this interpretation and evalu-
mining enterprise. These include data selection, data      ation process that new insights on the problem be-
cleaning and preprocessing, data transformation and        ing analyzed can be derived.
reduction, data mining task and algorithm selection,
and finally, postprocessing and the interpretation
of discovered knowledge. Here are the most im-             Data Mining Tasks
portant steps:                                             In verification-driven data analysis the user postu-
     Understand the application domain: A proper           lates a hypothesis, and the system tries to validate it.
understanding of the application domain is neces-          Common verification-driven operations include
sary to appreciate the data mining outcomes desired        querying and reporting, multidimensional analysis,
by the user. It is also important to assimilate and take   and statistical analysis. Data mining, on the other
advantage of available prior knowledge to maximize         hand, is discovery driven—that is, it automatically
the chance of success.                                     extracts new hypotheses from data. The typical data
     Collect and create the target dataset: Data min-      mining tasks include the following:
ing relies on the availability of suitable data that            Association rules: Given a database of trans-
reflects the underlying diversity, order, and structure     actions, where each transaction consists of a set of
of the problem being analyzed. Therefore, it is cru-       items, association discovery finds all the item sets
cial to collect a dataset that captures all the possi-     that frequently occur together, and also the rules
ble situations relevant to the problem being analyzed.     among them. For example, 90 percent of people
     Clean and transform the target dataset: Raw data      who buy cookies also buy milk (60 percent of gro-
contain many errors and inconsistencies, such as           cery shoppers buy both).
noise, outliers, and missing values. An important el-           Sequence mining: The sequence-mining task is
ement of this process is the unduplication of data         to discover sequences of events that commonly oc-
records to produce a nonredundant dataset. Another         cur together. For example, 70 percent of the people
important element of this process is the normal-           who buy Jane Austen’s Pride and Prejudice also buy
ization of data records to deal with the kind of           Emma within a month.
pollution caused by the lack of domain consistency.             Similarity search: An example is the problem
     Select features and reduce dimensions: Even           where a person is given a database of objects and a
after the data have been cleaned up in terms of elim-      “query” object, and is then required to find those
inating duplicates, inconsistencies, missing values,       objects in the database that are similar to the query
and so on, there may still be noise that is irrelevant     object. Another example is the problem where a per-
to the problem being analyzed. These noise attrib-         son is given a database of objects, and is then required
utes may confuse subsequent data mining steps, pro-        to find all pairs of objects in the databases that are
duce irrelevant rules and associations, and increase       within some distance of each other.
computational cost. It is therefore wise to perform             Deviation detection: Given a database of objects,
a dimension-reduction or feature-selection step to         find those objects that are the most different from
separate those attributes that are pertinent from those    the other objects in the database—that is, the out-
that are irrelevant.                                       liers. These objects may be thrown away as noise, or
     Apply data mining algorithms: After perform-          they may be the “interesting’’ ones, depending on the
ing the preprocessing steps, apply appropriate data        specific application scenario.
                                                                                                 DATA MINING ❚❙❘ 151



     Classification and regression: This is also called     stream. Extracted knowledge thus needs to be con-
supervised learning. In the case of classification, some-   stantly updated.
one is given a database of objects that are labeled with        Database integration: The various steps of the
predefined categories or classes. They are required         mining process, along with the core data mining
to develop from these objects a model that separates       methods, need to be integrated with a database
them into the predefined categories or classes.            system to provide common representation, storage,
Then, given a new object, the learned model is applied     and retrieval. Moreover, enormous gains are possi-
to assign this new object to one of the classes. In the    ble when these are combined with parallel data-
more general situation of regression, instead of pre-      base servers.
dicting classes, real-valued fields have to be predicted.        Privacy and security issues in mining: Privacy-
     Clustering: This is also called unsupervised learn-   preserving data mining techniques are invaluable in
ing. Here, given a database of objects that are usu-       cases where one may not look at the detailed data,
ally without any predefined categories or classes, the      but one is allowed to infer high-level information.
individual is required to partition the objects into       This also has relevance for the use of mining for na-
subsets or groups such that elements of a group share      tional security applications.
a common set of properties. Moreover, the partition             Human interaction: While a data mining algo-
should be such that the similarity between members         rithm and its output may be readily handled by a
of the same group is high and the similarity between       computer scientist, it is important to realize that the
members of different groups is low.                        ultimate user is often not the developer. In order for
                                                           a data mining tool to be directly usable by the ulti-
                                                           mate user, issues of automation—especially in the
Challenges in Data Mining                                  sense of ease of use—must be addressed. Even for
Many existing data mining techniques are usually ad        computer scientists, the use and incorporation of
hoc; however, as the field matures, solutions are          prior knowledge into a data mining algorithm is often
being proposed for crucial problems like the incor-        a challenge; they too would appreciate data mining
poration of prior knowledge, handling missing data,        algorithms that can be modularized in a way that fa-
adding visualization, improving understandability,         cilitates the exploitation of prior knowledge.
and other research challenges. These challenges in-             Data mining is ultimately motivated by the
clude the following:                                       need to analyze data from a variety of practical
     Scalability: How does a data mining algorithm         applications— from business domains such as
perform if the dataset has increased in volume and         finance, marketing, telecommunications, and man-
in dimensions? This may call for some innovations          ufacturing, or from scientific fields such as biology,
based on efficient and sufficient sampling, or on a          geology, astronomy, and medicine. Identifying new
trade-off between in-memory and disk-based pro-            application domains that can benefit from data min-
cessing, or on an approach based on high-perfor-           ing will lead to the refinement of existing techniques,
mance distributed or parallel computing.                   and also to the development of new methods where
     New data formats: To date, most data mining re-       current tools are inadequate.
search has focused on structured data, because it is
the simplest and most amenable to mining. However,                                                  Mohammed J. Zaki
support for other data types is crucial. Examples in-
clude unstructured or semistructured (hyper)text,
temporal, spatial, and multimedia databases. Mining        FURTHER READING
these is fraught with challenges, but it is necessary
because multimedia content and digital libraries pro-      Association for Computing Machinery’s special interest group on
                                                              knowledge discovery and data mining. Retrieved August 21, 2003,
liferate at astounding rates.                                 from http://www.acm.org/sigkdd.
     Handling data streams: In many domains the            Dunham, M. H. (2002). Data mining: Introductory and advanced top-
data changes over time and/or arrives in a constant           ics. Upper Saddle River, NJ: Prentice Hall.
152 ❘❙❚ BERKSHIRE ENCYCLOPEDIA OF HUMAN-COMPUTER INTERACTION



Han, J., & Kamber, M. (2000). Data mining: Concepts and techniques.        ping, rendering, and viewing. The data-generation
   San Francisco: Morgan Kaufman.                                          step can be a numerical simulation, a laboratory
Hand, D. J., Mannila, H., & Smyth, P. (2001). Principles of data mining.
   Cambridge, MA: MIT Press.
                                                                           experiment, a collection of sensors, an image scan-
Kantardzic, M. (2002). Data mining: Concepts, models, methods, and         ner, or a recording of Web-based business trans-
   algorithms. Somerset, NJ: Wiley-IEEE Press.                             actions. Filtering removes noise, extracts and
Witten, I. H., & Frank, E. (1999). Data mining: Practical machine learn-   enhances features, or rescales data. Mapping derives
   ing tools and techniques with Java implementations. San Francisco:
   Morgan Kaufmann.                                                        appropriate representations of data for the render-
                                                                           ing step. The representations can be composed of
                                                                           geometric primitives such as point, lines, poly-
                                                                           gons, and surfaces, supplemented with properties
     DATA VISUALIZATION                                                    such as colors, transparency, textures. Whereas the
                                                                           visualization of a computerized tomography (CT)
Data visualization is a new discipline that uses com-                      scan of a fractured bone should result in an image
puters to make pictures that elucidate a concept,                          of a bone, plenty of room for creativity exists when
phenomenon, relationship, or trend hidden in a                             making a visual depiction of the trend of a stock mar-
large quantity of data. By using interactive three-                        ket or the chemical reaction in a furnace.
dimensional (3D) graphics, data visualization goes                             Rendering generates two-dimensional or
beyond making static illustrations or graphs and em-                       three-dimensional images based on the mapping re-
phasizes interactive exploration.                                          sults and other rendering parameters, such as the
    The pervasive use of computers in all fields of                         lighting model, viewing position, and so forth. Finally,
science, engineering, medicine, and commerce has                           the resulting images are displayed for viewing. Both
resulted in an explosive growth of data, presenting                        photorealistic and nonphotorealistic rendering tech-
people with unprecedented challenges in under-                             niques exist for different purposes of visual com-
standing data. Data visualization transforms raw data                      munication. Nonphotorealistic rendering, which
into pictures that exploit the superior visual pro-                        mimics how artists use brushes, strokes, texture, color,
cessing capability of the human brain to detect                            layout, and so forth, is usually used to increase the
patterns and draw inferences, revealing insights hid-                      clarity of the spatial relationship between objects,
den in the data. For example, data visualization al-                       improve the perception of an object’s shape and size,
lows us to capture trends, structures, and anomalies                       or give a particular type of media presentation.
in the behavior of a physical process being modeled                            Note that the filtering and mapping steps are
or in vast amounts of Internet data. Furthermore, it                       largely application dependent and often require do-
provides us with a visual and remote means to com-                         main knowledge to perform. For example, the fil-
municate our findings to others.                                            tering and mapping steps for the visualization of
    Since publication of a report on visualization in                      website structure or browsing patterns would be
scientific computing by the U.S. National Science                          quite different from those of brain tumors or
Foundation in 1987, both government and industry                           bone fractures.
have invested tremendous research and development                              A data-visualization process is inherently itera-
in data-visualization technology, resulting in advances                    tive. That is, after visualization is made, the user
in visualization and interactive techniques that have                      should be able to go back to any previous steps,
helped lead to many scientific discoveries, better en-                      including the data-generation step, which consists
gineering designs, and more timely and accurate                            o f a nu m e r i c a l o r p hy s i c a l m o d e l , to m a ke
medical diagnoses.                                                         changes such that more information can be obtained
                                                                           from the revised visualization. The changes may be
                                                                           made in a systematic way or by trial and error.
Visualization Process                                                      The goal is to improve the model and understand-
A typical data-visualization process involves mul-                         ing of the corresponding problem via this visual
tiple steps, including data generation, filtering, map-                     feedback process.
                                                                                     DATA VISUALIZATION ❚❙❘ 153



Computational Steering                                      visualization can offer better 3D spatial acuity than
Data visualization should not be performed in iso-          humans have, and computer-assisted surgery is more
lation. It is an integral part of data analysis and the     reliable and reproducible. However, computer-
scientific discovery process. Appropriate visualiza-         assisted surgery has several challenges. First, the en-
tion tools integrated into a modeling process can           tire visualization process—consisting of 3D
greatly enhance scientists’ productivity, improve the       reconstruction, segmentation, rendering, and image
efficiency of hardware utilization, and lead to scien-       transport and display—must be an integrated part
tific breakthroughs. The use of visualization to drive       of the end-to-end surgical planning and procedure.
scientific discovery processes has become a trend.           Second, the visualization must be in real time, that
However, we still lack adequate methods to achieve          is, flicker free. Delayed visual response could lead to
computational steering—the process of inter-                dangerous outcomes for a surgery patient. Most im-
acting with as well as changing states, parameters, or      portant, the visualization must attain the required
resolution of a numerical simulation—and to be able         accuracy and incorporate quantitative measuring
to see the effect immediately, without stopping or          mechanisms.
restarting the simulation. Consequently, the key to              Telesurgery, which allows surgeons at remote sites
successful data visualization is interactivity, the abil-   to participate in surgery, will be one of the major ap-
ity to effect change while watching the changes             plications of virtual reality and augmented reality
take effect in real time on the screen.                     (where the virtual world and real world are allowed
    If all the steps in the modeling and visualization      to coexist and a superimposed view is presented to
processes can be performed in a highly interactive          the user). Due to the distance between a patient and
fashion, steering can be achieved. The ability to steer     surgeons, telesurgery has much higher data visual-
a numerical model makes the visualization process           ization, hardware, and network requirements.
a closed loop, becoming a scientific discovery process       Nevertheless, fast-improving technology and de-
that is self-contained. Students can benefit from such       creasing costs will make such surgery increasingly
a process because they can more easily move from            appealing. Whether and how much stereoscopic
concepts to solutions. Researchers can become much          viewing can benefit surgeons remains to be investi-
more productive because they can make changes ac-           gated. The most needed advance, however, is in inter-
cording to the interactive graphical interpretation of      face software and hardware.
the simulation states without restarting the simu-
lation every time.
    Computational steering has been attempted in            User Interfaces for Data Visualization
only a few fields. An example is the SCIRun system           Most data visualization systems supply the user with
used in computational medicine. To adopt compu-             a suit of visualization tools that requires the user
tational steering, researchers likely must redesign the     to be familiar with both the corresponding user in-
computational model that is required to incorporate         terfaces and a large visualization parameter space (a
feedback and changes needed in a steering process.          multidimensional space which consists of those
More research is thus needed to make computational          input variables used by the visualization program).
steering feasible in general.                               Intuitive and intelligent user interfaces can greatly
                                                            assist the user in the process of data exploration. First,
                                                            the visual representation of the process of data ex-
Computer-Assisted Surgery                                   ploration and results can be incorporated into the
During the past ten years significant advances have          user interface of a visualization system. Such an inter-
been made in rendering software and hardware tech-          face can help the user to keep track of the visualiza-
nologies, resulting in higher fidelity and real-time         tion experience, use it to generate new visualizations,
visualization. Computer-assisted surgery is an              and share it with others. Consequently, the interface
application of such advanced visualization tech-            needs to display not only the visualizations but
nologies with a direct societal impact. Computer            also the visualization process to the user. Second, the
154 ❘❙❚ BERKSHIRE ENCYCLOPEDIA OF HUMAN-COMPUTER INTERACTION



task of exploring large and complex data and visu-        by visual perception study, and user studies for mea-
alization parameter space during the mapping step         suring the usability of visualization tools and the suc-
can be delegated to an intelligent system such as a       cess of visualizations.
neural network. One example is to turn the 3D
segmentation problem into a simple 2D painting                                                               Kwan-Liu Ma
process for the user, leaving the neural network to
classify the multidimensional data. As a result, the      See also Information Spaces; Sonification
user can focus on the visualizations rather than on
the user interface widgets (e.g., a color editor, plot-
ting area, or layout selector) for browsing through       FURTHER READING
the multidimensional parameter space. Such next-
generation user interfaces can enhance data un-           Johnson, C., & Parker, S. (1995). Applications in computational med-
                                                             icine using SCIRun: A computational steering programming
derstanding while reducing the cost of visualization         environment. The 10th International Supercomputer Conference
by eliminating the iterative trial-and-error process         (pp. 2–19).
of parameter selection. For routine analysis of large-    Ma, K. L. (2000). Visualizing visualizations: Visualization viewpoints.
scale data sets, the saving can be tremendous.               IEEE Computer Graphics and Applications, 20(5), 16–19.
                                                          Ma, K.-L. (2004). Visualization—A quickly emerging field. Computer
                                                             Graphics, 38(1), 4–7.
                                                          McCormick, B., DeFanti, T., & Brown, M. (1987). Visualization in sci-
Research Directions                                          entific computing. Computer Graphics, 21(6)
The pervasiveness of the World Wide Web in aver-
age people’s lives has led to a data explosion. Some
data are relevant to some people’s needs, but most
are not. Nevertheless, many people do their every-                            DEEP BLUE
day jobs by searching huge databases of information
distributed in locations all over the world. A large      In 1997, the chess machine Deep Blue fulfilled a
number of computer services repeatedly operate on         long-standing challenge in computer science by de-
these databases. Information visualization, a branch      feating the human world chess champion, Garry
of visualization, uses visual-based analysis of data      Kasparov, in a six-game match. The idea that a com-
with no spatial references, such as a large amount of     puter could defeat the best humanity had to offer in
text and document. A data mining step (the proce-         an intellectual game such as chess brought many
dure to reduce the size, dimensionality, and/or com-      important questions to the forefront: Are com-
plexity of a data set), which may be considered as the    puters intelligent? Do computers need to be intelli-
filtering step, usually precedes the picture-making        gent in order to solve difficult or interesting
step of visualization. The mapping step often con-        problems? How can the unique strengths of humans
verts reduced relations into graphs or charts. Most       and computers best be exploited?
information visualizations are thus about display-
ing and navigating 2D or 3D graphs. People need
new reduction, mapping, and navigation methods            Early History
so that they can manage, comprehend, and use the          Even before the existence of electronic computers,
fast-growing information on the World Wide Web.           there was a fascination with the idea of machines
    Other important research directions in data           that could play games. The Turk was a chess-playing
visualization include improving the clarity of visu-      machine that toured the world in the eighteenth and
alizations, multidimensional and multivariate data        nineteenth centuries, to much fanfare. Of course the
(a data set with a large number of dependent vari-        technology in the Turk was mainly concerned with
ables) visualization, interaction mechanisms for large    concealing the diminutive human chess master hid-
and shared display space, visualization designs guided    den inside the machine.
                                                                                             DEEP BLUE ❚❙❘ 155



     In 1949, the influential mathematician Claude            ■ a complex evaluation function to assess the good-
Shannon (1916–2001) proposed chess as an ideal                 ness of a chess position, and
domain for exploring the potential of the then-new           ■ a strong emphasis on intelligent exploration
electronic computer. This idea was firmly grasped               (selective search) of the possible move sequences.
by those studying artificial intelligence (AI), who
                                                               The first two factors allowed the full Deep Blue
viewed games as providing an excellent test bed for
                                                           system to examine 100–200 million chess positions
exploring many types of AI research. In fact, chess
                                                           per second while selecting a move, and the complex
has often been said to play the same role in the field
                                                           evaluation function allowed Deep Blue to make more
of artificial intelligence that the fruit fly plays in ge-
                                                           informed decisions. However, a naive brute-force ap-
netic research. Although breeding fruit flies has
                                                           plication of Deep Blue’s computational power would
no great practical value, they are excellent subjects
                                                           have been insufficient to defeat the top human chess
for genetic research: They breed quickly, have suffi-
                                                           players. It was essential to combine the computer’s
cient variation, and it is cheap to maintain a large
                                                           power with a method to focus the search on move
population. Similarly, chess avoids some aspects
                                                           sequences that were “important.” Deep Blue’s selective
of complex real-world domains that have proven
                                                           search allowed it to search much more deeply on the
difficult, such as natural-language understanding,
                                                           critical move sequences.
vision, and robotics, while having sufficient com-
                                                               Deep Blue first played against world champion
plexity to allow an automated problem solver to
                                                           Garry Kasparov in 1996, with Kasparov winning the
focus on core AI issues such as search and knowl-
                                                           six-game match by a score of 4-2. A revamped Deep
edge representation.
                                                           Blue, with improved evaluation and more compu-
     Chess programs made steady progress in the
                                                           tational power, won the 1997 rematch by a score of
following decades, particularly after researchers aban-
                                                           3.5–2.5.
doned the attempt to emulate human thought
processes and instead focused on doing a more thor-
ough and exhaustive exploration of possible move
sequences. It was soon observed that the playing           Human and Computer
strength of such “brute-force” chess programs cor-
related strongly with the speed of the underlying          Approaches to Chess
computer, and chess programs gained in strength            It is clear that systems like Deep Blue choose moves
both from more sophisticated programs and faster           using methods radically different from those em-
computers.                                                 ployed by human experts. These differences result
                                                           in certain characteristic strengths of the two types
                                                           of players. Computers tend to excel at the short-
Deep Blue                                                  range tactical aspects of a game, mainly due to an
The Deep Blue computer chess system was devel-             extremely thorough investigation of possible move
oped in 1989–1997 by a team (Murray Campbell, A.           sequences. Human players can only explore perhaps
Joseph Hoane, Jr., Feng-hsiung Hsu) from IBM’s             a few dozen positions while selecting a move, but
T. J. Watson Research Center. Deep Blue was a leap         can assess the long-term strategic implications of
ahead of the chess-playing computers that had gone         these moves in a way that has proven difficult for
before it. This leap resulted from a number of fac-        a computer.
tors, including:                                                The combination of human and computer play-
  ■ a computer chip designed specifically for high-         ers has proven to be very powerful. High-level chess
    speed chess calculations,                              players routinely use computers as part of their prepa-
  ■ a large-scale parallel processing system, with more    ration. One typical form of interaction would have the
    than five hundred processors cooperating to se-        human player suggest strategically promising moves
    lect a move,                                           that are validated tactically by the computer player.
156 ❘❙❚ BERKSHIRE ENCYCLOPEDIA OF HUMAN-COMPUTER INTERACTION



Research Areas                                                           Newborn, M. (2003). Deep Blue: An artificial intelligence milestone.
While computer programs such as Deep Blue achieve                           New York: Springer-Verlag.
                                                                         Schaeffer, J. (2001). A gamut of games. AI Magazine, 22(3), 29–46.
a very high level of play, most of the knowledge in                      Schaeffer, J., & van den Herik, J. (Eds.). (2002). Chips challenging cham-
such systems is carefully crafted by human experts.                         pions: Games, computers, and artificial intelligence. New York:
Much research is needed to understand how to have                           Elsevier.
future systems learn the knowledge necessary to play                     Shannon, C. (1950). Programming a computer for playing chess.
                                                                            Philosophical Magazine, 41, 256–275.
the game without extensive human intervention.                           Standage, T. (2002). The Turk: The life and times of the famous
TD-gammon, a neural-network system that plays                               eighteenth-century chess-playing machine. New York: Walker &
world-class backgammon, was an early leader in                              Company.
this area.
    Determining the best mode for humans and
computers to interact in a more cooperative man-
ner (for example, with one or the other acting as as-                          DENIAL-OF-SERVICE
sistant, trainer, or coach) is another area worthy of
further research. Game-playing programs that are                                    ATTACK
based on large-scale searches have problems in trans-
lating the search results into forms that humans can                     A denial-of-service (DoS) attack causes the con-
deal with easily.                                                        sumption of a computing system’s resources—
    Some types of games, such as the Chinese                             typically with malicious intent—on such a scale as
game Go, have too many possible moves to allow a                         to compromise the ability of other users to interact
straightforward application of the methods used for                      with that system. Virgil Gligor coined the term denial-
chess, and pose significant challenges. Games with                        of-service attack in reference to attacks on operating
hidden information and randomness, such as poker                         systems (OS) and network protocols. Recently the
or bridge, also require new and interesting ap-                          term has been used specifically in reference to attacks
proaches. Interactive games, which employ com-                           executed over the Internet.
puter-generated characters in simulated worlds, can                          As governments and businesses increasingly rely
be more realistic and entertaining if the characters                     on the Internet, the damage that a DoS attack can
can behave in intelligent ways. Providing such                           cause by the disruption of computer systems has
intelligent characters is a key goal for future AI                       provided incentive for attackers to launch such at-
researchers.                                                             tacks and for system operators to defend against such
                                                                         attacks.
                                              Murray Campbell

See also Artificial Intelligence                                          Evolution of
                                                                         Denial-of-Service Attacks
FURTHER READING                                                          DoS vulnerabilities occur when a poor resources-allo-
                                                                         cation policy allows a malicious user to allocate so
Campbell, M., Hoane, A. J., & Hsu, F. (2002). Deep Blue. Artificial      many resources that insufficient resources are left for
   Intelligence, 134(1–2), 57–83.                                        legitimate users. Early DoS attacks on multiuser
Frey, P. W. (Ed.). (1983). Chess skill in man and machine. New York:     operating systems involved one user spawning a large
   Springer-Verlag.
Hsu, F. (2002). Behind Deep Blue: Building the computer that             number of processes or allocating a large amount of
   defeated the world chess champion. Princeton, NJ: Princeton           memory, which would exhaust the memory avail-
   University Press.                                                     able and result in operating system overload.
Laird, J. E., & van Lent, M. (2000). Human-level AI’s killer applica-
   tion: Interactive computer games. AI Magazine, 22(2), 15–26.
                                                                             Early network DoS attacks took advantage of the
Marsland, T. A., & Schaeffer, J. (Eds.). (1990). Computers, chess, and   fact that the early Internet was designed with implicit
   cognition. New York: Springer-Verlag.                                 trust in the computers connected to it. The unin-
                                                                         DENIAL-OF-SERVICE ATTACK ❚❙❘ 157



tended result of this trust was that users paid little   and the DDoS variant was that attackers were be-
attention to handling packets (the fundamental unit      ginning to use multiple computers in each attack,
of data transferred between computers on the             thus amplifying the attack. Internet software imple-
Internet) that did not conform to standard Internet      mentation and protocol attacks did not require mul-
protocols. When a computer received a malformed          tiple attackers to be successful, and effective defenses
packet that its software was not equipped to handle,     were designed (in some cases) against them. A DDoS
it might crash, thus denying service to other users.     attack, however, did not require a software imple-
These early DoS attacks were relatively simple and       mentation or protocol flaw to be present. Rather, a
could be defended against by upgrading the OS soft-      DDoS attack would consist of an attacker using mul-
ware that would identify and reject malformed pack-      tiple computers (hundreds to tens of thousands)
ets. However, network DoS attacks rapidly increased      to send traffic at the maximum rate to a victim’s com-
in complexity over time.                                 puter. The resulting flood of packets was sometimes
     A more serious threat emerged from the implicit     enough to either overload the victim’s computer
trust in the Internet’s design: The protocols in the     (causing it to slow to a crawl or crash) or overload
Internet themselves could be exploited to execute a      the communication line from the Internet to that
DoS attack. The difference between exploiting an         computer. The DDoS attacker would subvert con-
Internet software implementation (as did the pre-        trol of other people’s computers for use in the attack,
vious class of DoS attacks) and exploiting an Internet   often using flaws in the computer’s control code sim-
protocol itself was that the former were easy to iden-   ilar to Internet software implementation DoS attacks
tify (malformed packets typically did not occur out-     or simply attaching the attack control codes in an
side of an attack) and once identified could be          e-mail virus or Internet worm. The presence of at-
defended against, whereas protocol-based attacks         tacking computers on many portions of the Internet
could simply look like normal traffic and were dif-       gave this class of attacks its name.
ficult to defend against without affecting legiti-
mate users as well. An example of a protocol attack
was TCP SYN flooding. This attack exploited the fact      Defense against DoS Attacks
that much of the communication between comput-           Defending against DoS attacks is often challenging
ers over the Internet was initiated by a TCP hand-       because the very design of the Internet allows
shake where the communicating computers                  them to occur. The Internet’s size requires that even
exchanged specialized packets known as “SYN pack-        the smallest change to one of its fundamental pro-
ets.” By completing only half of the handshake, an       tocols be compatible with legacy systems that do not
attacker could leave the victim computer waiting for     implement the change. However, users can deploy
the handshake to complete. Because computers could       effective defenses without redesigning the entire
accept only a limited number of connections at           Internet. For example, the defense against Internet
one time, by repeating the half-handshake many           software implementation DoS attacks is as simple as
times an attacker could fill the victim computer’s       updating the software on a potential victim’s com-
connection capacity, causing it to reject new con-       puter; because the packets in this type of attack are
nections by legitimate users or, even worse, causing     usually malformed, the compatibility restriction is
the OS to crash. A key component of this attack was      easy to meet.
that an attacker was able to hide the origin of his           Defending against a protocol-level attack is more
or her packets and pose as different computer            difficult because of the similarity of the attack it-
users so the victim had difficulty knowing which         self to legitimate traffic. Experts have proposed several
handshakes were initiated by the attacker and which      mechanisms, which mostly center on the concept of
were initiated by legitimate users.                      forcing all computers initiating a handshake to show
     Then another class of DoS attack began to ap-       that they have performed some amount of “work”
pear: distributed denial-of-service (DDoS) attacks.      during the handshake. The expectation is that an
The difference between traditional DoS attacks           attacker will not have the computing power to
158 ❘❙❚ BERKSHIRE ENCYCLOPEDIA OF HUMAN-COMPUTER INTERACTION



impersonate multiple computers making handshake            We must wait to see whether DoS attacks will further
requests. Unfortunately, this class of defenses requires   threaten the Internet, provoking the acceptance of
a change in the Internet protocol that must be im-         radical defense proposals, or will simply fade into
plemented by all computers wanting to contact the          the background and become accepted as a regular
potential victim’s computer. Moreover, more                aspect of the Internet.
protocol-compliant solutions involve placing be-
tween the victim’s computer and the Internet spe-                                     Adrian Perrig and Abraham Yaar
cialized devices that are designed to perform many
handshakes at once and to pass only completed hand-        See also Security; Spamming
shakes to the victim.
     The DDoS variant of DoS attacks is the most dif-
ficult to defend against because the attack simply         FURTHER READING
overwhelms the victim’s computer with too many
packets or, worse, saturates the victim’s connection       Aura, T., Nikander, P., & Leiwo, J. (2000). DoS-resistant authentica-
                                                              tion with client puzzles. Security Protocols—8th International
to the Internet so that many packets are dropped be-          Workshop.
fore ever reaching the victim’s computer or network.       Gligor, V. D. (1983). A note on the denial of service problem. Proceedings
Some businesses rely on overprovisioning, which               of 1983 Symposium on Security and Privacy (pp. 139–149).
is the practice of buying computer resources far in        Gligor, V. D. (1986). On denial of service in computer networks.
                                                              Proceedings of International Conference on Data Engineering
excess of expected use, to mitigate DDoS attacks; this        (pp. 608–617).
practice is expensive but raises the severity of an        Gligor, V. D. (2003). Guaranteeing access in spite of service-flooding
attack that is necessary to disable a victim. Proposed        attacks. Proceedings of the Security Protocols Workshop.
                                                           Savage, S., Wetherall, D., Karlin, A., & Anderson, T. (2000). Practical
defenses against this type of attack—more so than             network support for IP traceback. Proceedings of ACM SIGCOMM
proposed defenses against other types of attacks—             2000 (pp. 295–306).
have focused on changing Internet protocols. Many          Wang, X., & Reiter, M. K. (2003). Defending against denial-of-service
proposals favor some type of traceback mechanism,             attacks with puzzle auctions. Proceedings of the 2003 IEEE
                                                              Symposium on Security and Privacy (pp. 78–92).
which allows the victim of an attack to determine          Yaar, A., Perrig, A., & Song, D. (2003). Pi: A path identification mech-
the identity and location of the attacking comput-            anism to defend against DDoS attacks. IEEE Symposium on Security
ers, in the hope that filters can be installed in the         and Privacy (pp. 93–107).
Internet to minimize the flood of traffic while
leaving legitimate traffic unaffected. At the time of
this writing, no DDoS defense proposal has been ac-
cepted by the Internet community.                               DESKTOP METAPHOR
                                                           The desktop metaphor is being used when the inter-
The Future                                                 face of an interactive software system is designed such
DoS attacks are likely to trouble the Internet for         that its objects and actions resemble objects and
the foreseeable future. These attacks, much like ur-       actions in a traditional office environment. For ex-
ban graffiti, are perpetrated by anonymous attack-          ample, an operating system designed using the desk-
ers and require a substantial investment to defend         top metaphor represents directories as labeled folders
against, possibly requiring a fundamental change in        and text documents as files. In graphical user inter-
the Internet’s protocols. Although several DoS at-         faces (GUIs), the bitmap display and pointing devices
tacks have succeeded in bringing down websites of          such as a mouse, a trackball, or a light pen are used
well-known businesses, most attacks are not as wildly      to create the metaphor: The bitmap display pres-
successful, nor have all businesses that have been         ents a virtual desk, where documents can be created,
victimized reported attacks for fear of publicizing        stored, retrieved, reviewed, edited, and discarded.
exactly how weak their computing infrastructure is.        Files, folders, the trash can (or recycle bin) and so
                                                                                  DESKTOP METAPHOR ❚❙❘ 159



forth are represented on the virtual desktop by graph-     Historical Overview
ical symbols called icons. Users manipulate these icons    In the 1960s and 1970s, several innovative con-
using the pointing devices. With pointing devices, the     cepts in the area of HCI were originated and im-
user can select, open, move, or delete the files or fold-   plemented using interactive time-shared computers,
ers represented by icons on the desktop.                   graphics screens, and pointing devices.
     Users can retrieve information and read it on the
desktop just as they would read actual paper docu-         Sketchpad
ments at a physical desk. The electronic document          Sketchpad was a pioneering achievement that opened
files can be stored and organized in electronic            the field of interactive computer graphics. In 1963,
folders just as physical documents are saved and man-      Ivan Sutherland used a light pen to create engineer-
aged in folders in physical file cabinets. Some of         ing drawings directly on the computer screen for his
the accessories one finds in an office are also present      Ph.D. thesis, Sketchpad: A Man-Machine Graphical
on the virtual desktop; these include the trash can,       Communications System. His thesis initiated a totally
a clipboard, a calendar, a calculator, a clock, a          new way to use computers. Sketchpad was exe-
notepad, telecommunication tools, and so on.               cuted on the Lincoln TX-2 computer at MIT. A light
     The metaphor of the window is used for the            pen and a bank of switches were the user interface
graphical boxes that let users look into informa-          for this first interactive computer graphics system.
tion in the computer. Multiple windows can be open         Sketchpad also pioneered new concepts of mem-
on the desktop at once, allowing workers to alternate      ory structures for storing graphical objects, rubber-
quickly between multiple computer applications (for        banding of lines (stretching lines as long as a user
example, a worker may have a word processing ap-           wants) on the screen, the ability to zoom in and
plication, a spreadsheet application, and an Internet      out on the screen, and the ability to make perfect
browser open simultaneously, each in its own win-          lines, corners, and joints.
dow). Computer users can execute, hold, and resume
their tasks through multiple windows.                      NLS
                                                           The scientist and inventor Doug Engelbart and his
                                                           colleagues at Stanford University introduced the oN
  BITMAP An array of pixels, in a data file or structure,   Line System (NLS) to the public in 1968. They also
  which correspond bit for bit with an image.              invented the computer mouse. The NLS was
                                                           equipped with a mouse and a pointer cursor for the
                                                           first time; it also was the first system to make use
     Beginning in the late 1970s, as personal computers    of hypertext. Among other features, the system pro-
and workstations became popular among knowledge            vided multiple windows, an online context-sensitive
workers (people whose work involves developing and         help system, outline editors for idea development,
using knowledge—engineers, researchers, and teach-         two-way video conferencing with shared workspace,
ers, for example), the usability of the computers and      word processing, and e-mail.
the productivity of those using them became im-
portant issues. The desktop metaphor was invented          Smalltalk
in order to make computers more usable, with the           In the early 1970s, Alan Kay and his colleagues at
understanding that more-usable computers would             Xerox’s Palo Alto Research Center (PARC) invented
increase users’ productivity. The desktop metaphor         an object-oriented programming language called
enabled users to work with computers in a more fa-         Smalltalk. It was the first integrated programming
miliar, more comfortable manner and to spend less          environment, and its user interface was designed us-
time learning how to use them. The invention of the        ing the desktop metaphor. It was designed not only
desktop metaphor greatly enhanced the quality of           for expert programmers of complex software, but
human-computer interaction.                                also for the novice users, including children: Its
160 ❘❙❚ BERKSHIRE ENCYCLOPEDIA OF HUMAN-COMPUTER INTERACTION



designers intended it to be an environment in which            Personalized interaction: A user could set up
users learned by doing.                                    attributes of the system in order to customize the
     In 1981 Xerox PARC integrated the innovations         interaction with the system. The personalized inter-
in the fields of human-computer symbiosis, personal         action did not interfere with the standard interaction
computing, objected-oriented programming lan-              methods.
guages, and local-area networks and arrived at the             Multiple tasks: Because workers in office envi-
Xerox 8010 Star information system. It was the             ronments perform many tasks simultaneously and
first commercial computer system that implemented           are often interrupted in their work, Lisa was designed
the desktop metaphor, using a mouse, a bitmap dis-         to be able to hold the current work while users at-
play, and a GUI. In interacting with the system, users     tended to those interruptions and to other business.
made use of windows, icons, menus, and pointing            The idea was that the user should be able to switch
devices (WIMP). Most of the workstations and per-          from one task to another freely and instantly.
sonal computers that were developed subse-                     Apple’s Lisa provided knowledge workers with a
quently, including Apple Computer’s Lisa (1983) and        virtual desktop environment complete with manip-
Macintosh (1984) and Microsoft’s Windows (1985),           ulable documents, file folders, calculators, electronic
were inspired by Star; like the Star system, they too      paper clips, wastebasket, and other handy tools. The
adopted the desktop metaphor.                              documents and other office-based objects were rep-
     Apple’s Lisa was designed to be a high-quality,       resented by naturalistic icons. The actions defined
easy-to-use computer for knowledge workers such            for the icons, such as selecting, activating, moving,
as secretaries, managers, and professionals in gen-        and copying, were implemented by means of mouse
eral office environments. Its design goals were:            operations such as clicking, double-clicking, and
     User friendliness: The developers of the Lisa         dragging. Lisa users did not have to memorize com-
wanted users to use the computer not only because          mands such as “delete” (“del”), “remove” (“rm”),
doing so was part of their job, but also because it was    or “erase” in order to interact with the system.
fun to use. The users were expected to feel comfort-           The first version of Microsoft Windows was in-
able because the user interface resembled their work-      troduced in 1985. It provided an interactive software
ing environment.                                           environment that used a bitmap display and a mouse.
     Standard method of interaction: A user was pro-       The product included a set of desktop applica-
vided with consistent look and feel in the system and      tions, including a calendar, a card file, a notepad, a
all applications, which meant that learning time could     calculator, a clock, and telecommunications pro-
be dramatically decreased and training costs lowered.      grams. In 1990 Windows 3.0 was introduced, the first
     Gradual and intuitive learning: A user should be      real GUI-based system running on IBM-compatible
able to complete important tasks easily with mini-         PCs. It became widely popular. The Windows oper-
mal training. The user should not be concerned with        ating system evolved through many more incarna-
more sophisticated features until they are necessary.      tions, and in the early 2000s was the most popular
Interaction with the computer should be intuitive;         operating system in the world.
that is, the user should be able to figure out what
he or she needs to do.
     Error protection: A user should be protected from     Research Directions
obvious errors. For example, Lisa allowed users to         The desktop metaphor, implemented through a
choose from a collection of possible operations that       graphical user interface, has been the dominate
were proper for the occasion and the object. By lim-       metaphor for human-computer interfaces since the
iting the choices, fatal errors and obvious errors could   1980s. What will happen to the human-computer
be avoided. Any error from a user should be pro-           interaction paradigm in the future? Will the desktop
cessed in a helpful manner by generating a warn-           metaphor continue to dominate? It is extremely dif-
ing message or providing a way of recovering from          ficult to predict the future in the computer world.
the error.                                                 However, there are several pioneering researchers
                                                                                      DESKTOP METAPHOR ❚❙❘ 161



exploring new interaction paradigms that could               mensional views of the MIT geographical model.
replace the desktop-based GUI.                               Users could physically control the phicons by
                                                             grasping and placing them so that a two dimensional
                                                             map of the MIT campus appears on the desk surface
  OPERATING SYSTEM Software (e.g., Windows 98, UNIX,         beneath the phicons. The locations of the Dome and
  or DOS) that enables a computer to accept input and        the Media Lab buildings on the map should match
  produce output to peripheral devices such as disk drives   with the physical locations of the phicons on the desk.
  and printers.
                                                             Ubiquitous Computing
                                                             In 1988 Mark Weiser at Xerox PARC introduced a
A Tangible User Interface                                    computing paradigm called ubiquitous computing.
At present, interactions between human and com-              The main idea was to enable users to access com-
puters are confined to a display, a keyboard, and a           puting services wherever they might go and when-
pointing device. The tangible user interface (TUI)           ever they might need them. Another requirement
proposed by Hiroshi Ishii and Brygg Ullmer at MIT’s          was that the computers be invisible to the users, so
Media Lab in 1997 bridges the space between human            the users would not be conscious of them. The users
and computers in the opposite direction. The user            do what they normally do and the computers in the
interface of a TUI-based system can be embodied in           background recognize the intention of the users and
a real desk and other real objects in an office envi-         provide the best services for them. It means that the
ronment. Real office objects such as actual papers            users do not have to learn how to operate comput-
and pens could become meaningful objects for the             ers, how to type a keyboard, how to access the
user interface of the system. Real actions on real ob-       Internet, etc. Therefore, the paradigm requires
jects can be recognized and interpreted as operations        that new types of computing services and computer
applied to the objects in the computer world, so that,       systems be created. New technologies such as con-
for example, putting a piece of paper in a waste-            text awareness, sensors, and intelligent distributed
basket could signal the computer to delete a docu-           processing. are required. Their interaction methods
ment. This project attempts to bridge the gap between        must be based on diverse technologies such as face
the computer world and the physical office envi-             recognition, character recognition, gesture recogni-
ronment by making digital information tangible.              tion, and voice recognition.
While the desktop metaphor provides the users with
a virtual office environment, in a TUI the physical
office environment, including the real desktop, be-             OPEN-SOURCE SOFTWARE Open-source software per-
comes the user interface.                                      mits sharing of a program’s original source code with
    Ishii and Ullmer designed and implemented                  users, so that the software can be modified and redis-
a prototype TUI called metaDESK for Tangible                   tributed to other users.
Geospace, a physical model of landmarks such as the
MIT campus. The metaDESK was embodied in real-
world objects and regarded as a counterpart of the                As new computing services and technologies are
virtual desktop. The windows, icons, and other graph-        introduced, new types of computing environments and
ical objects in the virtual desktop corresponded to          new interaction paradigms will emerge. The desktop
physical objects such as activeLENS (a physically em-        metaphor will also evolve to keep pace with techno-
bodied window), phicon (a physically embodied                logical advances. However, the design goals of the user
icon—in this case, models of MIT buildings such as           interfaces will not change much. They should be de-
the Great Dome and the Media Lab building), and so           signed to make users more comfortable, more effec-
forth. In the prototype system, the activeLENS was           tive, and more productive in using their computers.
equivalent to a window of the virtual desktop and
was used in navigating and examining the three di-                                                        Jee-In Kim
162 ❘❙❚ BERKSHIRE ENCYCLOPEDIA OF HUMAN-COMPUTER INTERACTION



See also Alto; Augmented Reality; Graphical User                         are more resource intensive; however, despite the extra
Interface                                                                costs, the need to reach a broader community of users
                                                                         and the desire to create systems that can perform tasks
                                                                         that require collaboration with users has led to a shift
FURTHER READING                                                          toward more dialog-based systems.
                                                                             Speech is a good modality for remote database
Goldberg, A. (1984). Smalltalk-80: The interactive programming envi-     access systems, such as telephone information ser-
   ronment. Reading, MA: Addison-Wesley.
Ishii, H., & Ullmer, B. (1997). Tangible bits: Towards seamless inter-
                                                                         vices, which would otherwise require a human op-
   faces between people, bits and atoms. In Proceedings of CHI ’97       erator or a tedious sequence of telephone keystrokes.
   (pp. 234–241), New York: ACM Press.                                   Spoken interaction is also useful when a user’s hands
Kay, A. (1993) The early history of Smalltalk. ACM SIGPLAN Notices,      are busy with other tasks, such as operating me-
   28(3), 69–95.
Kay, A., & Goldberg, A (1977). Personal dynamic media. IEEE              chanical controls, or for tasks for which the sound
   Computer, 10(3), 31–42.                                               of the user’s speech is important, such as tutoring
Myers, B., Ioannidis, Y., Hollan, J., Cruz, I., Bryson, S., Bulterman,   speakers in oral reading. Speech interaction can also
   D., et al. (1996). Strategic directions in human computer inter-
   action. ACM Computing Survey, 28(4), 794–809.
                                                                         make computers more accessible to people with
Perkins, R., Keller, D., & Ludolph, F. (1997). Inventing the Lisa user   vision impairments.
   interface. Interactions, 4(1), 40–53.                                     Speech dialog systems have been used for tutor-
Shneiderman, B. (1998). Designing the user interface: Strategies for     ing in oral reading; for providing information about
   effective human-computer interaction (3rd ed.). Reading, MA:
   Addison-Wesley.
                                                                         public transportation, train schedules, hotels, and
Weiser, M. (1991). The computer for the 21st century. Scientific         sight seeing; for making restaurant and real estate
   American, 256(3), 94–104.                                             recommendations; for helping people diagnose fail-
                                                                         ures in electronic circuits; and for making travel reser-
                                                                         vations. Spoken dialog is most successful when the
                                                                         scope of the task is well-defined and narrow, such as
          DIALOG SYSTEMS                                                 providing airline reservations or train schedules, be-
                                                                         cause the task creates expectations of what people
Speech dialog systems are dialog systems that use                        will say—and the more limited the scope, the more
speech recognition and speech generation to allow                        limited the expectations. These expectations are
a human being to converse with a computer, usually                       needed for the system to interpret what has been said;
to perform some well-defined task such as making                          in most cases the same group of speech sounds
travel reservations over the telephone.                                  will have several different possible interpretations,
     A dialog is a two-way interaction between two                       but the task for which the dialog system is used makes
agents that communicate. Dialogs are incremental                         one of the interpretations by far the most likely.
and can be adapted dynamically to improve the ef-
fectiveness of the communication.
     While people communicate efficiently and effec-                      The Architecture of
tively using dialog, computers do not typically engage
in dialogs with people. More common are presen-                          Speech Dialog Systems
tation systems, which are concerned with the effec-                      Speech dialog systems include the following
tive presentation of a fixed content, subject to a limited                components or processes: speech recognition,
number of constraints. Unlike dialogs, presentations                     natural-language parsing, dialog management,
are planned and displayed in their entirety (with-                       natural-language generation, and speech synthesis.
out intermediate feedback from the user) and thus                        There is also an application or database that provides
do not allow the system to monitor the effective-                        the core functionality of the system (such as book-
ness of the presentation or allow the user to interrupt                  ing a travel reservation) and a user interface to trans-
and request clarification. Dialog systems have been                       mit inputs from the microphone or telephone to
less common than presentation systems because they                       the speech-recognition component.
                                                                                        DIALOG SYSTEMS ❚❙❘ 163



Speech Recognition                                         its style of dialog interaction. In either case, a vo-
Understanding speech involves taking the sound in-         cabulary and a language model can be obtained that
put and mapping it onto a command, request, or             may only require a few thousand (and possibly only
statement of fact to which the application can re-         a few hundred) words.
spond. Speech recognition is the first step, which in-
volves mapping the audio signal into words in the          Natural-Language Parsing
target language. Early approaches, such as HEARSAY II      Natural-language parsing maps the sequence of
and HARPY, which were developed in the 1970s, were         words produced by the speech recognizer onto com-
based on rule-based artificial intelligence. They were      mands, queries, or propositions that will be meaning-
not very successful. Current approaches are based          ful to the application. There are a variety of approaches
on statistical models of language that select the most     to parsing; some try to identify general-purpose lin-
probable interpretation of each sound unit given im-       guistic patterns, following a so-called syntactic gram-
mediately preceding or following ones and the con-         mar, while others look for patterns that are specific
text of the task (which determines the vocabulary).        to the domain, such as a semantic grammar. Some
    For spoken-dialog systems, the level of speech         systems use simpler approaches, such as word
recognition quality that is desired is known as tele-      spotting, pattern matching, or phrase spotting.
phone quality, spontaneous speech (TQSS). This level       The output of the parser will typically be a slot-and-
of recognition is necessary if spoken-dialog applica-      filler-based structure called a case frame, in which
tions such as reservation services are to be success-      phrases in the input are mapped to slots correspon-
ful over a telephone line. TQSS is more difficult to        ding to functions or parameters of the application.
understand than face-to-face speech because over                A key requirement of parsing for spoken dialog
the telephone the audio signal normally includes           is that the parser be able to handle utterances that
background and channel noise, acoustical echo and          do not form a complete sentence or that contain the
channel variations, and degradation due to band-           occasional grammatical mistake. Such parsers are
width constraints. Moreover, spontaneous speech            termed robust. Syntactic and semantic parsers work
includes pauses, disfluencies (such as repetitions and      best when the input is well formed structurally.
incomplete or ill-formed sentences), pronunciation         Simpler methods such as pattern matching and
variations due to dialects, as well as context-dependent   phrase spotting can be more flexible about structural
formulations and interruptions or overlapping speech       ill-formedness, but may miss important syntactic
(known as “barge-in”).                                     variations such as negations, passives, and topical-
    One way that speech recognition is made more           izations. Also, the simpler approaches have little
accurate is by limiting the vocabulary that the sys-       information about how to choose between two
tem allows. To determine a sublanguage that will be        different close matches. To be useful in practice, an-
sufficiently expressive to allow people to use the ap-      other requirement is that the parser be fast enough
plication effectively and comfortably, two techniques      to work in real time, which is usually only possible
are generally used: One approach is to observe or          if the analysis is expectation driven. By the late 1990s,
stage examples of two people engaged in the domain         most spoken-dialog systems still focused on get-
task; the other approach is to construct a simulated       ting the key components to work together and had
man-machine dialog (known as a Wizard of Oz                not achieved real-time behavior (interpretation and
[WOZ] simulation) in which users try to solve the          generation of speech)—with only a few systems
domain task. In a WOZ simulation, users are led to         using real-time behavior.
believe they are communicating with a function-
ing system, while in reality the output is generated       Dialog Management
by a person who simulates the intended functional-         Dialog management involves interpreting the rep-
ity of the system. This approach allows the design-        resentations created by the natural-language
ers to see what language people will use in response       parser and deciding what action to take. This process
to the limited vocabulary of the proposed system and       is often the central one and drives the rest of the
164 ❘❙❚ BERKSHIRE ENCYCLOPEDIA OF HUMAN-COMPUTER INTERACTION



system. Dialog management may involve following                  Historically, natural-language generation com-
a fixed pattern of action defined by a grammar (for           ponents have not run in real time, with the realiza-
example, answering a question, or it may involve rea-       tion component being an important bottleneck.
soning about the users’ or the system’s current knowl-      These systems can be slow if they follow an approach
edge and goals to determine the most appropriate            that is essentially the inverse of parsing—taking a
next step. In this second instance, then, dialog man-       structural description of a sentence, searching for
agement may also keep track of the possible goals of        grammar rules that match the description, and then
the users and their strategies (plans) for achieving        applying each of the rules to produce a sequence of
them. It may also try to identify and resolve break-        words. As a result, many spoken-dialog systems have
downs in communication caused by lack of under-             relied on preformulated answers (canned text). More
standing, misunderstanding, or disagreement.                recently, real-time approaches to text generation have
     One factor that distinguishes dialog managers is       been developed that make use of fixed patterns or
the distribution of control between the system and          templates that an application can select and thereby
the user. This has been referred to as initiative, or the   bypass the need to perform a search within the
mode of communication, with the mode being con-             generation grammar.
sidered from the perspective of the computer sys-
tem. When the computer has complete control, it             Speech Synthesis
is responsible for issuing queries to the user, col-        Speech synthesis allows the computer to respond to
lecting answers, and formulating a response. This           the user in spoken language. This may involve se-
has been called directive mode. At the opposite ex-         lecting and concatenating pieces of prerecorded
treme, some systems allow the user to have complete         speech or generating speech two sounds at a time, a
control, telling the system what the user wants to do       method known as diphone-based synthesis. (Diphone
and asking the system to provide answers to specific         refers to pairs of sounds.) Databases of utterances to
queries. This is known as passive mode. In the              be prerecorded for a domain can be determined by
middle are systems that share initiative with the user.     analyzing the utterances produced by a human per-
The system may begin by issuing a query to the user         forming the same task as the information system and
(or receiving a query from the user) but control may        then selecting the most frequent utterances. Diphone-
shift if either party wishes to request clarification or     based synthesis also requires a database of prerecorded
to obtain information needed for a response. Control        sound; however instead of complete utterances the
may also shift if one party identifies a possible break-     database will contain a set of nonsense words (that
down in communication or if one party disagrees             have examples of all pairs of sounds), containing all
with information provided by the other. Dialogs have        phone-phone transitions for the target output lan-
been shown to more efficient if control can shift to         guage. Then when the synthesizer wants to gener-
the party with the most information about the               ate a pair of sounds, it selects a word that contains
current state of the task.                                  the sound-pair (diphone) and uses the correspon-
                                                            ding portion of the recording.
Natural-Language Generation                                     Although these basic components of speech
Natural-language generation is used to generate an-         dialog systems can be combined in a number of
swers to the user’s queries or to formulate queries         ways, there are three general approaches: pipelined
for the user in order to obtain the information needed      architectures, agent-based architectures, and hub-
to perform a given task. Natural-language genera-           and-spoke-based architectures. In a pipelined archi-
tion involves three core tasks: content selection (de-      tecture, each component in the sequence processes its
ciding what to say), sentence planning (deciding how        input and initiates the next component in the
to organize what to say into units), and realization        sequence. Thus, the audio interface would call the
(mapping the planned response onto a grammati-              speech recognizer, which would call the natural-
cally correct sequence of words).                           language parser, and so on, until the speech synthesis
                                                                                      DIALOG SYSTEMS ❚❙❘ 165



component is executed. In an agent-based approach,        infrastructure and is distributed with a working
a centralized component (typically the dialog man-        implementation for a travel-planning domain. It can
ager) initiates individual components and deter-          be downloaded freely from the Internet. The group
mines what parameters to provide them. This may           also maintains a telephone number connected to
involve some reasoning about the results provided         their telephone-based travel-planning system that
by the components. In a hub-and-spoke architecture        anyone can try.
there is a simple centralized component (the hub)             The Center for Spoken Language Research (CSLR)
which brokers communication among the other               at the University of Colorado in Boulder distributes
components, but performs no reasoning. Since 1994,        the Conversational Agent Toolkit. This toolkit in-
a hub-and-spoke architecture called Galaxy                cludes modules that provide most of the function-
Communicator has been under development. It has           ality needed to build a spoken-dialog system,
been proposed as a standard reference architec-           although code must be written for the application
ture that will allow software developers to combine       itself. As a model, CSLR distributes their toolkit
“plug-and-play”-style components from a variety           with a sample (open-source) application for the travel
of research groups or commercial vendors. The             domain that can be used as a template; it is based on
Galaxy Communicator effort also includes an open-         the Galaxy Communicator hub architecture.
source software infrastructure.                               TRINDIKIT is a toolkit for building and exper-
                                                          imenting with dialog move engines (mechanisms for
                                                          updating what a dialog system knows, based on di-
Dialog System Toolkits                                    alog moves (single communicative actions such as
Creating speech dialog systems is a major under-          “giving positive feedback”) and information states
taking because of the number and complexity of the        (information stored by the dialog system). It has been
components involved. This difficulty is mitigated by       developed in the TRINDI and SIRIDUS projects, two
the availability of a number of software toolkits that    European research projects that investigate human-
include many, if not all, of the components needed        machine communication using natural language.
to create a new spoken-dialog system. Currently such      TRINDIKIT specifies formats for defining informa-
toolkits are available both from commercial vendors       tion states, rules for updating the information state,
(such as IBM, which markets ViaVoice toolkits for         types of dialog moves, and associated algorithms.
speech recognition and speech synthesis) and aca-
demic institutions. Academic institutions generally
distribute their software free for noncommercial use,     The Evaluation of
but sell licenses for commercial applications. Below,
we consider a few of the major (academically avail-       Speech Dialog Systems
able) speech dialog toolkits. In addition to these        Prior to 1990, methods of evaluating speech dialog
toolkits, a number of institutions distribute indi-       systems concentrated on the number of words that
vidual components useful in building speech dialog        the speech recognizer identified correctly. In the early
systems. For example, the Festival Speech Synthesis       1990s, there was a shift to looking at the quality of
System developed by the University of Edinburgh           the responses provided by spoken-dialog systems.
has been used in a number of applications.                For example, in 1991, the U.S. Defense Advanced
    The Communicator Spoken Dialog Toolkit,               Research Projects Agency community introduced a
developed by researchers at Carnegie Mellon Uni-          metric that evaluates systems based on the number
versity, is an open-source toolkit that provides a com-   of correct and incorrect answers given by the system.
plete set of software components for building and         Systems are rewarded for correct answers and
deploying spoken-language dialog systems for              penalized for bad answers, normalized by the total
both desktop and telephone applications. It is built      number of answers given. (The effect is that it is bet-
on top of the Galaxy Communicator software                ter to give a nonanswer such as “I do not understand”
166 ❘❙❚ BERKSHIRE ENCYCLOPEDIA OF HUMAN-COMPUTER INTERACTION



or “please rephrase your request” than to give an in-     and government. Among its goals are promoting,
correct answer.) This approach relies on the exis-        developing, and distributing reusable discourse-
tence of a test database with a number of sample          processing components; encouraging empirical meth-
sentences from the domain along with the correct          ods in research; sharing resources and data among
answer, as well as a set of answers from the system       the international community; exploring techniques
to be evaluated.                                          for evaluating dialog systems; promoting stan-
    Starting in the late 1990s, approaches to evalu-      dards for discourse transcription, segmentation, and
ating dialog success have looked at other measures,       annotation; facilitating collaboration between de-
such as task-completion rates and user satisfaction       velopers of various system components; and en-
(as determined by subjective questionnaires).             couraging student participation in the discourse and
Subjective factors include perceived system-response      dialog community.
accuracy, likeability, cognitive demand (how much
effort is needed to understand the system), habit-                                                         Susan W. McRoy
ability (how comfortable or natural the system is
to use), and speed. There has also been success in        See also Natural-Language Processing, Open Source
predicting user satisfaction or task completion on        Software, Speech Recognition, Speech Synthesis
the basis of objectively observable features of the di-
alog, such as task duration, the number of system
words per turn, the number of user words per              FURTHER READING
turn, the number of overlapping turns, sentence er-
ror rates, and perceived task completion. Statistical     Allen, J. F., Schubert, L. K., Ferguson, G., Heeman, P., Hwang, C. H.,
methods such as multiple regression models and clas-         Kato, T., et al. (1995). The TRAINS project: A case study in
                                                             building a conversational planning agent. Journal of Experimental
sification trees are then used to predict user satis-         and Theoretical Artificial Intelligence, 7, 7–48.
faction and task-completion scores.                       Bernsen, N. O., Dybkjaer, H., & Dybkjaer, L. (1998). Designing inter-
                                                             active speech systems: From first ideas to user testing. New York:
                                                             Springer Verlag.
                                                          Fraser, N. (1997). Assessment of interactive systems. In D. Gibbon,
The Research Community for                                   R. Moore, and R. Winski (Eds.), Handbook of standards and re-
                                                             sources for spoken language systems (pp. 564–614). New York:
Speech Dialog Systems                                        Mouton de Gruyter.
Research on speech dialog systems is interdiscipli-       Grosz, B. J., & Sidner, C. (1986). Attention, intention, and the struc-
                                                             ture of discourse. Computational Linguistics, 12(3), 175–204.
nary, bringing together work in computer science,         Haller, S., Kobsa, A., & McRoy, S. (Eds.). (1999). Computational mod-
engineering, linguistics, and psychology. There are          els for mixed-initiative interaction. Dordrect, Netherlands: Kluwer
a number of journals, conferences, and workshops             Academic Press.
                                                          Huang X. D., Alleva, F., Hon, H. W., Hwang, M. Y., Lee, K. F., and
through which researchers and developers of spoken-          Rosenfeld, R. (1993). The Sphinx II Speech Recognition System:
dialog systems disseminate their work. Important             An overview. Computer Speech and Language, 7(9), 137–148.
journals include Computer Speech and Language and         Jurafsky, D., & Martin, J. (2000). Speech and language processing: An
Natural Language Engineering. Conferences most fo-           introduction to natural language processing, computational linguis-
                                                             tics, and speech recognition. Upper Saddle River, NJ: Prentice-Hall.
cused on such systems include Eurospeech and              Larsson, S., & Traum, D. (2000). Information state and dialogue man-
Interspeech (the International Conference on Spoken          agement in the TRINDI Dialogue Move Engine Toolkit [Special
Language Processing). In addition, the Special Interest      issue on best practice in spoken dialogue systems]. Natural Language
Group on Discourse and Dialog (SIGdial) organizes            Engineering, 6(3–4), 323–340.
                                                          Luperfoy, S. (Ed.). (1998). Automated spoken dialog systems. Cambridge,
an annual workshop. SIGdial is a Special Interest            MA: MIT Press.
Group (SIG) of both the Association for Com-              McRoy, S. W. (Ed.). (1998). Detecting, repairing, and preventing
putational Linguistics and the International Speech          human-machine miscommunication [Special issue]. International
                                                             Journal of Human-Computer Studies, 48(5).
Communication Association (ISCA). SIGdial is an           McRoy, S. W., Channarukul, S., & Ali, S. S. (2001). Creating natural
international, nonprofit cooperative organization             language output for real-time applications intelligence. Intelligence:
that includes researchers from academia, industry,           New Visions of AI in Practice, 12(2), 21–34.
                                                                                                             DIGITAL CASH ❚❙❘ 167



McRoy, S. W., Channarukul, S., & Ali, S. S. (2003). An augmented               1. Token-based systems store funds as tokens that
   template-based approach to text realization. Natural Language                  can be exchanged between parties. Traditional
   Engineering, 9(2), 1–40.
Minker, W., Bühler, D., & Dybkjær, L. (2004). Spoken multimodal
                                                                                  currency falls in this category, as do many types
   human-computer dialog in mobile environments. Dordrect,                        of stored-value payment systems, such as sub-
   Netherlands: Kluwer.                                                           way fare cards, bridge and highway toll systems
Mostow, J., Roth, S. F., Hauptmann, A., & Kane, M. (1994). A proto-               in large metropolitan areas (e.g., FastPass,
   type reading coach that listens. In Proceedings of the Twelfth National
   Conference on Artificial Intelligence (AAAI-94) (pp. 785–792). Seattle,         EasyPass), and electronic postage meters. These
   WA: AAAI Press.                                                                systems store value in the form of tokens, either
Pellom, B., Ward, W., Hansen, J., Hacioglu, K., Zhang, J., Yu, X., &              a physical token, such as a dollar bill, or an elec-
   Pradhan, S. (2001, March). University of Colorado dialog systems
   for travel and navigation. Paper presented at the Human Language
                                                                                  tronic register value, such as is stored by a sub-
   Technology Conference (HLT-2001), San Diego, CA.                               way fare card. During an exchange, if the full
Roe, D. B., & Wilpon, J. G. (Eds.). (1995). Voice communication be-               value of a token is not used, then the remain-
   tween humans and machines. Washington, D.C.: National Academy                  der is returned (analogous to change in a cur-
   Press.
Seneff, S., Hurley, E., Lau, R. Pau, C., Schmid, P., & Zue, V. (1998).
                                                                                  rency transaction)—either as a set of smaller
   Galaxy II: A reference architecture for conversational system de-              tokens or as a decremented register value.
   velopment. Proceedings of the 5th International Conference on Spoken           Generally, if tokens are lost (e.g., if one’s wallet
   Language Processing, 931–934.                                                  is stolen or one loses a subway card), the tokens
Smith, R., & Hipp, D. R. (1995). Spoken natural language dialog sys-
   tems: A practical approach. New York: Oxford University Press.                 cannot be recovered.
Smith, R., & van Kuppevelt, J. (Eds.). (2003). Current and new direc-          2. Account-based systems charge transactions to
   tions in discourse and dialogue. Dordrect, Netherlands: Kluwer.                an account. Either the account number or a ref-
van Kuppevelt, J, Heid, U., & Kamp, H. (Eds.). (2000). Best practice
   in spoken dialog systems [Special issue]. Natural Language
                                                                                  erence to the account is used to make payment.
   Engineering, 6(3–4).                                                           Examples include checking accounts, credit card
Walker, M., Litman, D., Kamm, C., & Abella, A. (1998). Evaluating                 accounts, and telephone calling cards. In some
   spoken dialogue agents with PARADISE: Two case studies.                        instances, the account is initially funded and
   Computer Speech and Language, 12(3), 317–347.
Walker, M. A., Kamm, C. A., & Litman, D. J. (2000). Towards devel-
                                                                                  then spent down (e.g., checking accounts); in
   oping general models of usability with PARADISE [Special issue                 other instances, debt is increased and periodi-
   on best practice in spoken dialogue systems]. Natural Language                 cally must be paid (e.g., credit cards). In most
   Engineering, 6(3–4).                                                           account-based systems, funds (or debt) are
Wilks, Y. (Ed.). (1999). Machine conversations. Dordrect, Netherlands:
   Kluwer.                                                                        recorded by a trusted third party, such as a bank.
                                                                                  The account can be turned off or renumbered
                                                                                  if the account number is lost.
                                                                                 The more complex an electronic payment sys-
                                                                             tem is, the less likely consumers are to use it. (As
                                                                             an example, a rule of thumb is that merchants of-
                DIGITAL CASH                                                 fering “one-click ordering” for online purchases en-
                                                                             joy twice the order rate of merchants requiring
                                                                             that payment data be repeatedly entered with each
The use of digital cash has increased in parallel with
                                                                             purchase.)
the use of electronic commerce; as we purchase
items online, we need to have ways to pay for them
electronically. Many systems of electronic payment                           Electronic Payment
exist.
                                                                             Using Credit Cards
                                                                             The most common form of electronic payment on
Types of Money                                                               the Internet today is credit card payment. Credit cards
Most systems of handling money fall into one of two                          are account based. They are issued by financial
categories:                                                                  institutions to consumers and in some cases to
168 ❘❙❚ BERKSHIRE ENCYCLOPEDIA OF HUMAN-COMPUTER INTERACTION



organizations. A consumer presents the credit card         auctions. As a result, a market has opened for third-
number to a merchant to pay for a transaction. On          party payment processors. Today, the largest third-
the World Wide Web credit card account numbers             party payment processor is PayPal, owned by the
are typically encrypted using the Secure Socket Layer      eBay auction service. Third-party payment proces-
(SSL) protocol built into most Web browsers. The           sor systems are account based.
merchant often attempts to verify the card holder by           Consumers can pay for third-party purchases in
performing address verification (checking numbers           three ways: by paying from an account maintained
appearing in an address) or by using a special veri-       with the third party, by paying from a credit card
fication code (typically printed on the reverse side        account, and by paying from a checking account.
of the credit card). In the United States credit card      Merchants’ rates for accepting funds from a credit card
users typically enjoy strong rights and can reverse        account are slightly higher than their rates for ac-
fraudulent transactions.                                   cepting funds from a conventional credit card account.
     Although the SSL protocol (in typical configu-             Third-party payment accounts are convenient
rations) provides strong encryption preventing third       because they are simple to use and provide consumers
parties from observing the transaction, risks still ex-    with protection against being overcharged. However,
ist for the credit card holder. Many merchants apply       they tend not to provide the same degree of pro-
inadequate security to their database of purchases,        tection that a credit card-funded purchase provides.
and attackers have gained access to large numbers of       Because third-party payment accounts are widely
credit cards stored online. Moreover, some merchants       used with auction systems, where fraud rates are un-
charge incorrect amounts (or charge multiple times)        usually high, the degree of protection is a serious
for credit card transactions. Although fraudulent          consideration.
transactions are generally reversible for U.S. resi-
dents, time and effort are required to check and
amend such transactions. In some instances, crim-          Smartcards and
inals engage in identity theft to apply for addi-
tional credit by using the identity of the victim.         Other Stored-Value Systems
     To reduce these risks, some experts have proposed     Stored-value systems store value on a card that is
a system that uses third parties (such as the bank that    used as needed. Smartcards are a token-based pay-
issued the card) to perform credit card transactions.      ment system. Many smartcards use an integrated cir-
A notable example of this type of system is Verified        cuit to pay for purchases. They are widely used in
by Visa. However, the additional work required to con-     Europe for phone cards and in the GSM cellular tele-
figure the system has deterred some consumers, and          phone system. Mondex is a consumer-based system
as a result Verified by Visa and similar systems remain     for point-of-sale purchases using smartcards. Use of
largely unused. The most elaborate of these systems        smartcards is limited in Asia and largely unused in
was the Secure Electronic Transactions (SET) pro-          North America. (In North America only one major
tocol proposed by MasterCard International and Visa        vendor, American Express, has issued smartcards to
International; however, the complexity of SET led to       large numbers of users, and in those cards the smart-
its being abandoned. In these systems credit card pur-     card feature is currently turned off.)
chases are usually funded with a fee that is charged to        Experts have raised a number of questions about
the merchant. Although rates vary, typical fees are fifty   the security of smartcards. Successful attacks con-
cents plus 2 percent of the purchase amount.               ducted by security testers have been demonstrated
                                                           against most smartcard systems. Experts have raised
                                                           even deeper questions about the privacy protec-
Third-Party Payment Accounts                               tion provided by these systems. For example, in
A merchant must be able to process credit card pay-        Taiwan, where the government has been moving to
ments. This processing is often inconvenient for small     switch from paper records to a smartcard system for
merchants, such as people who sell items in online         processing National Health Insurance payments,
                                                                                           DIGITAL CASH ❚❙❘ 169



considerable public concern has been raised about          Micropayments
potential privacy invasions associated with the use of     One of the most interesting types of electronic pay-
health and insurance records on a smartcard system.        ment is micropayments. In many instances consumers
     A number of devices function like a smartcard         wish to purchase relatively small-value items. For ex-
but have different packaging. For example, some ur-        ample, consider a website that vends recipes. Each
ban areas have adopted the FastPass system, which          recipe might be sold for only a few cents, but sold in
allows drivers to pay bridge and highway tolls us-         volume, their value could be considerable. (Similarly,
ing radio link technology. As a car passes over a sen-     consider a website that offers online digital recordings
sor at a toll booth, value stored in the FastPass device   of songs for ninety-nine cents each.) Currently, mak-
on the car is decremented to pay the toll. The state       ing small payments online using traditional payment
of California recently disclosed that it uses the same     methods is not feasible. For example, as mentioned,
technology to monitor traffic flow even when no toll         credit card companies typically charge merchants a
is charged. The state maintains that it does not gather    processing fee of fifty cents plus 2 percent of the pur-
personal information from FastPass-enabled cars,           chase amount for credit card transactions—clearly
but experts say that it is theoretically possible.         making credit card purchases for items that cost less
                                                           than fifty cents impractical. Most merchants refuse to
                                                           deal with small single-purchase amounts and require
Anonymous Digital Cash                                     that consumers either buy a subscription or purchase
A number of researchers have proposed anonymous            the right to buy large numbers of items. For example,
digital cash payment systems. These would be token-        newspaper websites that offer archived articles typi-
based systems in which tokens would be issued by           cally require that consumers purchase either a sub-
a financial institution. A consumer could “blind”          scription to access the articles or purchase a minimum
such tokens so that they could not be traced to the        number of archived articles—they refuse to sell
consumer. Using a cryptographic protocol, a con-           archived articles individually.
sumer could make payments to merchants without                 To enable small single purchases, a number of
merchants being able to collect information about          researchers have proposed micropayment systems
the consumer. However, if a consumer attempted to          that are either token based or account based. An ex-
copy a cryptographic token and use it multiple times,      ample of an account-based micropayment system is
the cryptographic protocol would probably allow            the NetBill system designed at Carnegie Mellon
the consumer’s identity to be revealed, allowing           University. This system provides strong protection
the consumer to be prosecuted for fraud.                   for both consumers and merchants and acts as an
     Anonymous digital cash payment systems have           aggregator of purchase information. When purchases
remained primarily of theoretical interest, although       across a number of merchants exceed a certain thresh-
some tr ials have been made (notably of the                old amount, that amount is charged in a single credit
Digicash system pioneered by David Chaum).                 card purchase.
Anonymous payment for large purchases is illegal               An example of a token-based micropayment sys-
in the United States, where large purchases must be        tem is the PepperCoin system proposed by Ron Rivest
recorded and reported to the government. Moreover,         and Silvio Micali and currently being commercial-
consumers generally want to record their purchases         ized. Peppercoin uses a unique system of “lottery
(especially large ones) to have maximum consumer           tickets” for purchases. For example, if a consumer
protection. Some researchers have demonstrated that        wishes to make a ten-cent purchase, he might use a
anonymous digital cash payment systems are not com-        lottery ticket that is worth ten dollars with a prob-
patible with atomic purchases (that is, guaranteed ex-     ability of 1 percent. The expected value paid by the
change of goods for payment). The principal demand         consumer would be the same as the items he pur-
for anonymous payment appears to be for transactions       chased; but any single charge would be large enough
designed to evade taxes, transactions of contraband,       to justify being charged using a traditional payment
and transactions of socially undesirable material.         mechanism (such as a credit card).
170 ❘❙❚ BERKSHIRE ENCYCLOPEDIA OF HUMAN-COMPUTER INTERACTION



    Despite the promise of micropayment systems,
they remain largely unused. Most merchants prefer                                       DIGITAL DIVIDE
to support small-value items by using Web-based
advertising or subscriptions. Nonetheless, advocates                          There is both optimism and pessimism about the ul-
of micropayment systems maintain that such sys-                               timate impact of the digital revolution on individ-
tems enable new classes of electronic commerce.                               ual, societal, and global well-being. On the optimistic
                                                                              side are hopes that access to information and com-
                                                                              munication technologies, particularly the Internet,
Challenges for Digital Cash                                                   will facilitate a more equitable distribution of social,
Although digital cash is being increasingly used, a                           economic, and political goods and services. On the
number of challenges remain. The principal chal-                              pessimistic side are beliefs that lack of access to these
lenge is associating payment with delivery of goods                           technologies will exacerbate existing inequalities,
(this challenge is often known as the “atomic swap”                           both globally and among groups within societies.
or “fair exchange” problem.) Merchants also need to                           The phrase digital divide was coined to refer to this
be protected from using stolen payment information,                           gap between the technology haves and have-nots—
and consumers need to be protected from merchants                             between those who have access to information and
who inadequately protect payment information (or,                             communications technologies, most notably the
even worse, engage in fraud.) Finally, effective pay-                         Internet, and those who do not. The overriding
ment methods need to be developed and accepted to                             concern is that a world divided by geographic, re-
support both large and small purchases. A balance                             ligious, political, and other barriers will become
must be reached between consumers who want                                    further divided by differing degrees of access to dig-
anonymous purchases and government authorities                                ital technologies.
who want to tax or record purchases. These challenges
make digital cash a rapidly developing research area.
                                                                              Evidence of a Digital Divide
                                                           J. D. Tygar        Evidence supporting the existence of a global digi-
                                                                              tal divide is overwhelming. Of the estimated 430 mil-
See also E-business                                                           lion people online in 2001, 41 percent resided in the
                                                                              United States and Canada. The remaining Internet
                                                                              users were distributed as follows: 25 percent in
FURTHER READING                                                               Europe, 20 percent in the Asian Pacific (33 percent
                                                                              of this group in Asia, 8 percent in Australia and New
Chaum, D., Fiat, A., & Naor, M. (1990). Untraceable electronic cash.          Zealand), 4 percent in South America, and 2 percent
   In G. Blakley & D. Chaum (Eds.), Advances in cryptolog y
   (pp. 319–327). Heidelberg, Germany: Springer-Verlag.
                                                                              in the Middle East and Africa. Even among highly
Electronic Privacy Information Center. (2003). Privacy and human              developed nations there are vast differences in
   rights 2003. Washington, DC: Author.                                       Internet access. For example, in Sweden 61 percent
Evans, D., & Schmalensee, R. (2000). Paying with plastic: The digi-           of homes have Internet access compared to 20 per-
   tal revolution in buying and borrowing. Cambridge, MA: MIT
   Press.
                                                                              cent of homes in Spain.
Kocher, P., Jaffe, J., & Jun, B. (1999). Differential power analysis. In M.        In light of the global digital divide evidence
   Weiner (Ed.), Advances in cryptology (pp. 388–397). Heidelberg,            that Internet use is rapidly increasing takes on ad-
   Germany: Springer-Verlag.                                                  ditional significance. According to data compiled
Mann, R., & Winn, J. (2002). Electronic commerce. Gaithersburg, MD:
   Aspen Publishers.                                                          by a variety of sources, the rise in Internet use
O’Mahony, D., Peirce, M., & Tewari, H. (2001). Electronic payment sys-        extends throughout both the developed and de-
   tems for e-commerce (2nd ed.). Norwood, MA: Artech House.                  veloping world. The rapidly increasing global reach
Tygar, J. D. (1998). Atomicity in electronic commerce. Networker, 2(2),
   23–43.
                                                                              of the Internet intensifies concerns about its po-
Wayner, P. (1997). Digital cash: Commerce on the Net (2nd ed.). San           tential to exacerbate existing global economic and
   Francisco: Morgan-Kaufmann.                                                social disparities.
                                                                                                                             DIGITAL DIVIDE ❚❙❘ 171




                                   HomeNetToo Tries to Bridge Digital Divide

       B e g u n i n t h e f a l l o f 2 0 0 0 , Hom e Ne t To o w a s a n                 “If I’m stressed out or depressed or the day
       eighteen-month field study of home Internet use in low-                              is not going right, I just get on the computer
                                                                                           and just start messing around and I come
       income families. Funded by an Information Technology
                                                                                           up with all sorts of things like ‘okay, wow.’ ”
       Research grant from the National Science Foundation, the
       project recruited ninety families who received in-home                              “You get a lot of respect because you have
                                                                                           a computer in your house. I think people
       instruction on using the Internet, and agreed to have their
                                                                                           view you a little differently.”
       Internet use recorded and to complete surveys on their ex-
       periences. In exchange, each family received a new home                             “A lot of times I’m real busy, and it was hard
                                                                                           for me to get a turn on the computer too.
       computer, Internet access, and in-home technical support.
                                                                                           My best chance of getting time on the com-
            The comments of the HomeNetToo participants about
                                                                                           puter is I get up at 6 AM and the rest of
       their computer use provide a broad range of views about                             the family gets up at seven. So if I finish my
       the pleasures and problems of computer interactions:                                bath and get ready quickly I can get on
                                                                                           before anyone else is up. And I can have an
                 “When somebody’s on the computer what-
                                                                                           hour space to do whatever I want while
                 ever it is they’re doing on that computer at
                                                                                           they’re sleeping and getting up and dressed
                 that time, that’s the world they’re in…it’s
                                                                                           themselves.”
                 another world.”
                                                                                           “I feel like I don’t have time ...who has time
                 “With the computer I can do things…well,
                                                                                           to watch or play with these machines.
                 I tell the computer to do things nobody else
                                                                                           There’s so much more in life to do.”
                 will ever know about, you know what I am
                 saying? I have a little journal that I keep that                          “Instead of clicking, I would like to talk to
                 actually nobody else will know about un-                                  it and then say ‘Can I go back please?’ ”
                 less I pull it up.”
                                                                                           “They talk in computer technical terms. If
                 “I escape on the computer all the time...I                                they could talk more in layman’s terms, you
                 like feeling ‘connected to the world’ and I                               know, we could understand more and solve
                 can dream.”                                                               our own problems.”

                                                                                  Source: Jackson, L. A., Barbatsis, G., von Eye, A., Biocca, F. A., Zhao, Y.,
                                                                                  & Fitzgerald, H. E. (2003c). Implications for the digital divide of
                                                                                  Internet use in low-income families. IT & Society., 1(5), 219–244.




    Evidence for a digital divide within the United                          to the following factors have been observed in
States is a bit more controversial, and has shifted                          all surveys to date:
from irrefutable in 1995 to disputable in 2002.
In its first Internet repor t in 1995, the U.S.                               ■ Income: Income is the best predictor of Internet
Department of Commerce noted large disparities                                  access. For example, only 25 percent of house-
in Internet access attributable to income, educa-                               holds with incomes of less than $15,000 had
tion, age, race or ethnicity, geographic location,                              Internet access in 2001, compared to 80 per-
and gender. In its fifth Internet report in 2002, all                           cent of households with incomes of more than
disparities had shrunk substantially. However, only                             $75,000.
a few disappeared entirely. Although 143 million                              ■ Education: Higher educational attainment is as-
U.S citizens now have access to the Internet                                    sociated with higher rates of Internet use. For ex-
(54 percent of the population), gaps attributable                               ample, among those with bachelor’s degrees or
172 ❘❙❚ BERKSHIRE ENCYCLOPEDIA OF HUMAN-COMPUTER INTERACTION



    better, over 80 percent use the Internet, compared     ■ African-Americans are less likely than whites to
    to 40 percent of those with only a high school           say the Internet helps them to stay connected
    diploma.                                                 to family and friends.
  ■ Age: Internet use rates are highest between the        ■ Women and parents are driving the growth of
    ages of twelve and fifty; they drop precipitously         the African-American Internet population.
    after age fifty-five.                                    ■ Mirroring the pattern of gender differences in the
  ■ Race or ethnicity: Asian/Pacific Islanders and           general population, African-American women
    whites are more likely to use the Internet (71 per-      are much more likely than African-American men
    cent and 70 percent, respectively) than are              to search for health, job, and religious informa-
    Hispanics and African-Americans (32 percent              tion online. African American men are much
    and 40 percent, respectively). However, growth           more likely than African-American women to
    in Internet use has been greater among the lat-          search for sports and financial information and
    ter than the former groups.                              to purchase products online.
                                                           ■ Compared with older African-Americans, those
    The gender gap so evident in the 1995 U.S.
                                                             under age thirty are more likely to participate in
Department of Commerce survey disappeared by the
                                                             chat rooms, play games, and use multimedia
2002 survey. However, gender-related differences re-
                                                             sources. Older African-Americans are more likely
main. Among those over sixty years old, men had
                                                             to search for religious information than are
higher Internet use rates than did women. Among
                                                             younger African-Americans.
the twenty-to-fifty-year-old group, women had higher        ■ The gap in Internet access between African-
Internet use rates than did men. Also diminishing if
                                                             Americans and whites is closing, but African-
not disappearing entirely are gaps related to geo-
                                                             Americans still have a long way to go. Moreover,
graphic location. Internet use rates in rural areas
                                                             those with access to the Internet do not go on-
climbed to 53 percent in 2002, almost as high as the
                                                             line as often on a typical day as do whites, and
national average, but use rates for central-city resi-
                                                             online African-Americans do not participate on
dents was only 49 percent, compared to 57 percent
                                                             a daily basis in most Web activities at the same
for urban residents outside the central city.
                                                             level as do online whites.
    In addition to the five Internet reports by the U.S.
Department of Commerce, a number of other or-
                                                              A number of researchers have also been inter-
ganizations have been tracking Internet use and is-
                                                          ested in race differences in U.S. Internet access.
sues related to the digital divide. The Pew Internet
                                                          Thomas Hoffman and Donna Novak, professors of
and American Life Project devoted one of its several
                                                          management at Vanderbilt University, examined the
reports to African-Americans and the Internet,
                                                          reasons for race differences in Internet access and
focusing on how African-Americans’ Internet use
                                                          concluded that income and education cannot fully
differ from whites’ use. These differences are im-
                                                          explain them. Even at comparable levels of income
portant to understanding the racial digital divide in
                                                          and education, African-Americans were less likely to
the United States and are potentially important to
                                                          have home PCs and Internet access than were whites.
understanding global digital-divide issues that
                                                          The psychologist Linda Jackson and her colleagues
may emerge as access to the Internet becomes less