Adaptive Systems in Educational Context

FCE Seminar, May 6, 1996
Presented by Nick Sawhney

Why Build Adaptive Systems?

An overview of key issues (paper by Brown, Norman & Riches)

Reasons for Providing Adaption:

  • Heterogeneous Users
  • Perceptual & Psychomotor Skills
  • Capability (user's model of system & familiarity)
  • Understanding of Tasks / Domain
  • Expectations & Motives
  • Cognitive Abilities/Styles
  • Personality & Social/Organizational Skills
  • Preferences for Interation styles and Modalities

    • Key Issues for Adaptive Systems:
  • What is the purpose/ goal of building such a system?
  • What is it adaptive to?
  • Evaluation Criteria (metrics) for measuring success?
    - eg. Do the adaptions produce an improvement in performance?
  • Need specific knowledge of Problem Domain
  • Knowledge of differences between users
  • System must have a means of meeting its intended purpose

    • How do we design Adaptive Systems?

    - Taxanomy & Methodology (paper by Totterdell & Rautenbauch)

    Question: If you design good systems to begin with, should they require adaption?
    Consider following Issues:

  • Degrees of Freedom between Users, Groups, Tasks and Applications
  • Differences may be persistant but also Dynamic
  • Designing for Different Environments of Use
  • Exlicit Design Assumptions
  • Can Designer Anticipate all situations or build a range of soultions (dynamically selected)

    • Purposes:
  • Extend System Lifespan
  • Widen User Base and hence Product Market
  • Enable user goals and increase satisfaction/accaptance
  • Improve operational accuracy and hence overall performance
  • Reduce Operational Learning & enhance User Understanding

    • Aspects of an Adaptive System:
  • User & Environment
  • Stimulus
  • Available Resources
  • Responses
  • Monitors/Triggers

    • Taxanomy of Adaptive Systems
    Level of SystemVariationSelectionTesting
    DesignedDDD
    AdaptableDUD
    AdaptiveDSD
    Self-RegulatingDSS
    Self-MediatingDSS
    Self-ModifyingSSS

    - Based on Amount of Control System has on Negotiating Change
    - Who is considered the Agent of Change - Designer, User, System?

    Case Studies of Adaptive Systems:

    Case Study 1: AIDA - Adpative CAD system

    Short-term Adaptions

  • User may not understand reason for Change
  • Useful for simple situations (error correction) or repeated tasks

    • Long-term Adaptions
  • Changes in System Behavior / Application Functionality (new requirements)
  • Based on Cognitive changes in user (hard to measure)
  • Changes must occur gradually

    • Consequences of using AUIs

  • User better understands System Complexity/ functions
  • Better user performance (5-20 times)
  • Gain & Maintain user Attention
  • Avoid overloading or underloading the user

    • Three components of Adaptive System:

    User Model

    Application Model

    Monitor

    Case Study 2: Interactive Note-taking System

    (Software Agents paper by Schlimmer & Hermens)
    Goal: Speed Information entry and reduce user errors

    Adaptive Features:

  • Contextual Prompting: Predicts what user will write and presents completion selections
  • Constructs "Button-box" UI automatically

    Model / Process of Representation & Learning:

    Notes --> Tokenzation --> FSM --> Prefix Tree --> Interface Options

  • To characterize syntax of user's notes, the system learns FSM.
  • To generate predictions, the system learns decision tree classifier as states within FSM.
  • Constructs GUI by converting FSM to a set of buttons.

    • Definition of Problem Domain: Education

    - Paper on Study/Review Strategies for Learning from Lectures (Alison King)

    Students viewed Lecture --> took regular notes --> after lecture, used 3 study strategies:

    Study StrategyShort-termLong-term
    Self-QuestioningGoodBetter
    SummarizingBetterGood
    Reviewing NotesWorstWorst

    Design of an Adaptive Note-taking and Review System?