Danielle McNamara University of Memphis by DeptEdu


									    Developing a Reading Strategy ITS:
    Competing Constraints from Theory,
  Technology, Pedagogy, and Experiments

  Danielle S. McNamara, Irwin Levinstein,
Chutima Boonthum, and Srinivasa Pillarisetti

         University of Memphis
Psychology / Institute for Intelligent Systems

  Funded by the IES Reading Program and the NSF IERI Program
           iSTART Investigators
Co-PIs and Senior Researchers: Irwin Levinstein
(ODU), Keith Millis (NIU), Joe Magliano (NIU),
Grant Sinclair, Katja Wiemer-Hastings (NIU), Max
Louwerse, Art Graesser
Postdocs & Staff: Cedrick Bellissens, Rachel
Best, Chutima Boonthum, Zhiqiang Cai, David
Dufty, Joyce Kim, Chris Kurby, Phil McCarthy,
Tenaha O’Reilly, Yasuhiro Ozuru, Margie
Petrowski, Srinivasa Pillarisetti, Roger Taylor
Many students at Memphis, ODU, and NIU
Interactive Strategy Training for Active Reading and Thinking

  • Currently provides self-explanation reading strategy
    training that
       • combines training to self-explain text and training
         to use active reading strategies
       • is adaptive
       • engages the trainee in an interactive learning
         environment using animated agents

  • Goal is to provide via the internet
     • a variety of empirically supported interventions to
       improve strategies for reading and thinking
                                  McNamara, Levinstein, & Boonthum, 2004
• Introduction Module
  – Teacher-Agent and 2 Student Agents discuss
    reading strategies
• Demonstration Module
  – Genie self-explains a text
  – Merlin provides feedback
  – Trainee is asked to identify strategies
• Practice Module
  – Trainee types self-explanations to science text
  – Merlin guides the trainee and provides feedback
            Based on SERT

        Self-Explanation Reading Training
Training to self-explain text using reading strategies
     (e.g., paraphrasing, bridging, elaborating)
Self Explanation: say aloud or type an explanation

                                           McNamara, 2004
• 1996-2002: Funded by McDonnell and ODU
    – Develop and test SERT
• 2000-2006: Funded by NSF IERI
    – Test SERT in high-school classrooms
    – Develop iSTART
• 2004-2008: Funded by IES Reading
    – Increase adaptivity - add texts, modules, student model
    – Develop teacher interface

•   2000-2002: Developed iSTART v1.0
•   2002-2004: Developed iSTART v2.0
•   2004-2006: Developed iSTART v3.0
•   Currently developing Teacher Interface
          Overarching Goals and
• Follow Original SERT Script as closely as possible
• But, take advantage of computer environment
  –   Facilitates individualized interaction
  –   Enables more fine-tuned feedback
  –   Increases time for practice
  –   Escapes (some) social dynamics of classroom

• Anticipated older computers in high schools
• Anticipated recursive development and frequent
       Initial Decision Making (2000-2001)

•   Architecture (e.g., software on server)
•   Programs – nonproprietary: Java, MySQL
•   Agents vs. Text
•   Pedagogical Agents vs. Real People
•   Synthesized Voices vs. Human Voices
•   Full bodied vs. Talking Heads
•   Cartoon-like vs. Human-like
•   Develop Agents vs. Use Microsoft Agents
Initial Version
        Version 2.0 Changes
• Presentation order of the five strategies
• Mini-demonstration
• Dialogue scripts (e.g., more examples,
  short dialogues)
• Multiple-choice quizzes revised
• Synthesized voices improved
• Revised interface
background colors
to reduce
(probably not one
of our best moves)
Added box labels
Changed boxes
from yellow on blue,
to black on white
Boxes are
uniform size –
they don’t grow
Developed pause
and repeat buttons.

Scroll Bars
replaced buttons.
Boxes, characters,
text, and speech
bubbles no longer
      Motivation for Changes
• Human Factors
• Observations during testing
• Data and theory
     Theory vs. Pedagogy vs. Data
• Come up with more ideas than we can test
• But, have to avoid the kitchen sink
    – Can’t make every modification you think of
•   Progress is made by relating ideas to theory
•   But, testing them remains complicated
•   And, not all ideas turn out to be good ones
•   Testing the revisions
    – We ain’t in Experimental Psychology Land anymore
    – So, hard to know if each revision ‘works’
• Time constraints
           Two Examples
• Data indicated that there was a problem
• Theory pointed to solutions
• Testing told the tale
       Demonstration Section
• Students asked to do a wide range of tasks
  – Identify and locate strategies
  – Locate text that is the source of strategies
  – Point, click, highlight
• College students did fine
  – High School students – not so much
• Revised to increase scaffolding and reduce WM
• Data indicates that changes were effective
  – Students increase in levels
         Increase Paraphrasing
• Hypothesized that less skilled students needed more
  practice at basic skills
   – Developed Paraphrasing Practice Module
• Conducted Experiment
   – Students received dedicated practice in paraphrasing (without
     self explanation)
• Predicted that less skilled students would benefit from
  more paraphrasing practice

• Au Contraire – the more skilled students benefited from
  the extra practice and the less skilled students benefited
  more from the version without it
            Can you test it?
• The New Demonstration included a host of
  – scaffolding, reduce number of choices,
    visually chunk self-explanation, etc.
• The Paraphrase Module consisted of a
  single modification (per se)
• The fun factor
   – must be inherently interesting or challenging, but not too much
• The boredom factor
   – Can’t be too repetitious or too long
• The embarrassment factor
   – e.g., 'Genie to the rescue' failed because other
     students saw rescued students singled out
• Avoiding distractions
   – developed a means for students to adjust volume pitch and
     speed of voice, but never used it because they would play
• Theory vs. Intuition (stay the path)
• For more information and papers

To top