Adaptive Work-Centered User Interface Technology

Document Sample
Adaptive Work-Centered User Interface Technology Powered By Docstoc
					Adaptive Work-Centered User Interface Technology (ACUITy)
Andrew W Crapo

GE Global Research Jena User Conference May 11, 2006

Copyright 2006 General Electric Global Research

Context: work-centered support systems Problem: extensible, adaptable decision support Approach: semantic model-based Architecture:
 Jena provides underlying repository and transitive reasoner  Controller embodies Problem/Vantage/Frame concepts, provides special-purpose reasoning (e.g., learned defaults)  UI Engine transforms the Controller’s representation of ontology concepts and data into a user interface representation

Copyright 2006 General Electric Global Research

2/ GE /

Work-Centered Support Systems
Problem: View work as a series of problem-solving events Vantage: Each problem-solving event requires interaction with a set of information, where some information is more critical than other information Frame: The information relevant to a vantage is communicated via a frame (not necessarily in the windows sense); a physical representation of the information set
Problem Classes A
B C Problem Focus


Center Visualization


Work Domain Variable Cluster 1 Work Domain Variable Cluster 2


Work Domain Variable Cluster 3

Work-centered design focuses on the intrinsic nature of work: How do people solve problems, regardless of technology?
Copyright 2006 General Electric Global Research 3/ GE /

Problem: How Do We…
Prepare for knowledge change Knowledge representation and communication must keep pace with the evolution of the autonomic work domain Apply information effectively Decision-makers must assimilate and process more new knowledge, and use it correctly Design for Trust Decision-makers must learn to trust the knowledge base and decision support aids Communicate knowledge, don’t flood with information Decision support systems must avoid information overload Create a shared understanding of how information is used Machines need to understand information use to intelligently communicate knowledge
Knowledge rendering and sharing Knowledge extraction Knowledge representation and reasoning

Improved Decisioning

Communicate the Right Information to the Right People at the Right Time in the Right Way
4/ GE /

Copyright 2006 General Electric Global Research

Why Semantic Models (Ontology) and WCSS
UI design is always based on models of the user and problem domain

Design models: either implicit in the designer’s head, or made explicit as part of a UCD process Implementation models: become part of the system itself, usually hardcoded (e.g. if user is of type x and problem of type y then display z…)

WCSS offers a way to develop models of work that become the basis for design and implementation Ontology offers a way to make these models the computational basis of a learning, extensible system


Creates a shared understanding between users and the system about information meaning and use Represents information in semantic terms that both the user and the system can understand Promotes information reuse

Copyright 2006 General Electric Global Research

5/ GE /

Semantic Model Structure
Domain-specific models nested within upper-level models to facilitate evolution and extensibility

Work Model
Problem Focus

• Work at Hand (problem focus) • Location, Time • Work processes • Other Context • Information locations

• When to present information
• What information to present • How to present information • When to ask for clarification or guidance Frame



User Group Profile: • Preferences for Groups of Similar Users
User Profiles: • User Competence by Task • User Preferences

Interaction Model

User Model

Copyright 2006 General Electric Global Research

6/ GE /

Adaptive Work-Centered User Interface Technology (ACUITy)
Domain 1 Data Sources Domain 2 Data Sources

Requests & Interaction Histories


UI Engine

Data & Metadata: what to present & how

ACUITy Controller

Upper Level Ontology

Public Ontologies

Information requests, Data Capture, Contextual Parameters, Actions

Problem-VantageFrame Ontology
Tailored Information

Modular design enables extensibility through specialized models of the work domain.
Copyright 2006 General Electric Global Research

Domain 1 Ontology Flightline Mx

Domain 2 Ontology Engine Removal Planning


Domain n Ontology

7/ GE /

Key Concepts in the Problem-VantageFrame (PVF) Ontology
PVF Ontology formalizes the principles of WCSS:
Problem-Vantage-Frame: Problems characterize the work being done; Vantages are collections of information presentation objects relevant to a particular problem set; Frames are composed of one or more vantages.

Focus-Periphery Organization: We can specify which vantage within a frame occupies the focal area of the display.
First-Person Perspective: Specific user preferences for information content and presentation are saved with the session; peer group preferences can be inferred.
WorkDomain InformationObject DataSeries Map Acuity Controller



Presentation Object









Copyright 2006 General Electric Global Research

8/ GE /

Value of Upper-Level Ontology
Gives implementable meaning across domains
• Script is the form (independent, abstract) of an occurrence

• Situation is the occurrence of a mediation of physical things. • Purpose is occurrence of an intention—an abstract mediation.

Hierarchy of Top-Level Categories (after Sowa 2000)
Copyright 2006 General Electric Global Research 9/ GE /

Example Use of Scripts by AcuityController



WorkDomainInformation Object

Acuity Controller






encodes, hasEffect
contains isTranformedBy contains

Parameter VantageSelection


Data Repositories, External Resources

Copyright 2006 General Electric Global Research

10 / GE /

Acuity Controller Functionality
Interfaces with Jena for persistence, transitive reasoning Tightly coupled to Problem/Vantage/Frame ontology
 “Understands” containment, e.g., a Frame or Vantage contains Presentation Objects, a Graph contains Series  Keeps an updatable map of a Presentation Object’s parameters  Supports explicit and implicit cloning of Presentation Objects, Parameters

Implements special behavior
 Auto-instantiation of “missing properties”, e.g., MyFrame has someValuesFrom MyWorkVantage  Ask user for missing properties, e.g., AcuityController has someValuesFrom Frame, cardinality 1  Transparent pass-through of legacy data as if it were in the ontology
Copyright 2006 General Electric Global Research

11 / GE /

Acuity Controller: Interface to Legacy Databases

OWL Reasoner
Ontology - tBox - aBox


Legacy DB

Pseudo-instance of class X that is also of type DBInstancesDescriptor or DBStatementsDescriptor and enables transparent retrieval of instances of X or statements about X from legacy database
Copyright 2006 General Electric Global Research 12 / GE /

Acuity Controller Functionality (cont.)
Integrates procedural knowledge (scripts)
 Modeled through OWL classes and properties  Extensible through user-defined Java packages  Controller functionality implemented as scripts • Modifiable, extensible • Example: replacable algorithm for learning defaults • Example: XRDQL with UPDATE, INSERT, DELETE, CREATE  Result of one script execution may be input to another

Selectively persists instance data
 Allows user to return to previous sessions  This “memory” provides data for simple learning of preferences

Limited implementation of work domain model
 Problem types, e.g., state-mismatch problem
Copyright 2006 General Electric Global Research 13 / GE /

UI Engine
UI Engine

Transforms the ACUITy Controller’s representation of ontology concepts and data into a user interface representation (masking between XML and something actionable)


ACUITy Controller Interaction Logic

ACUITy Controller


User Interface Model Objects

Interfaces between the ACUITy Controller and client application • Understands how to retrieve and process the contents of a frame
• Updates model properties & sends them back to the ontology

e.g. HTML
e.g. Eclipse

Invokes client application renderers to control look and feel Allows for multiple types of client interfaces produced by the same UI engine; e.g. web (current application), extensible to fat-client
Copyright 2006 General Electric Global Research 14 / GE /

Simple Example
Create: 1. An instance of a Script—MyDataScript (data source) 2. A subclass of SimpleDataTable—MyDataTable
• MyDataTable encodes (hasValue) MyDataScript

3. A subclass of SimpleDataSeries—MyDataSeries

MyDataSeries encodes (hasValue) MyDataScript
MyGraph contains someValuesFrom MyDataSeries

4. A subclass of Graph—MyGraph
5. A subclass of Vantage—MyTableVantage

MyTableVantage contains someValuesFrom MyDataTable
MyGraphVantage contains someValuesFrom MyGraph

6. A subclass of MyGraphVantage
7. A subclass of Frame--MyFrame
• MyFrame contains someValuesFrom each Vantage subclass
Copyright 2006 General Electric Global Research 15 / GE /


Client Interface for Simple Graph and Table

Copyright 2006 General Electric Global Research

16 / GE /

Benefits of ACUITy
Enables automated reasoning based on computational models of users and their work
• Intelligent information display • Platform for interaction with remote services, such as agents, infospace services, etc.

Simplifies software implementation and maintenance
• Users finish the design: reduces the need to predict exactly what the user needs to see and how (particularly important in designing for future domains like JSF)

• New applications are specified through ontology extensions rather than by creating each new user interface element manually • Developing a visual model editor that allows non-programmers to define new applications, extend existing applications
Copyright 2006 General Electric Global Research 17 / GE /

AJAX: finer-grained client/server interaction ACUITyEditor: easier application development, user extensions Open Source release: community interest, feedback? Enhanced learning
 User -> peer group, user-defined grouping  Pattern recognition -> tbox model extensions  Best practice recognition, propagation?

More Problem modeling: greater work-centeredness Semantic web services integration
 Dynamic content discovery  Semantic search for relevant content, by user and/or by application
Copyright 2006 General Electric Global Research 18 / GE /

Shared By: