QUALITATIVE MODELING IN EDUCATION

1
QUALITATIVE MODELING IN EDUCATION

• Bert Bredweg and Ken Forbus
• Yesim Imamoglu

2
Computers and Education

• Using educational software
• Supports constructivism in education
• Allows multiple representations and interaction
• Serves to individual differences

3
Qualitative Reasoning and Education

• Qualitative reasoning is valuable for education
  because
• Uses of conceptual knowledge
• What happens, when does it happen, what effects
  it, what does it effect?
• Provides grounding and framework for quantitative
  and traditional mathematical models

4
Recent Applications

• Teachable Agents project (Vanderbilt University,
  Biswas et al. 2001)
• Bettys Brain-qualitative mathematics
• ALI (DSauza et al. 2001)
• Based on qualitative process theory
• Domain independent-can be attached to any
  quantitative simulation

5
Recent Applications

• Application of qualitative process theory in
  chemistry classes (Syedd, Pang and Sharifuddin,
  2002)
• Qualitative model is used alongside classroom
  experiment to determine optimal use of substances

6
Aspects of modeling that existing visual
languages do not address

• The importance of broadly applicable principles
  and processes
• Understanding when a model is relevant
• Qualitative understanding of behavior

7
VModel

• Student friendly visual notation for qualitative
  process theory
• Creates a software environment that helps
  students express their qualitative, conceptual
  models

8
VModel

• Visual notation is based on concept maps
• Nodes represent entities and properties of
  entities
• Each node has a specified type such as Thing,
  Multiple Thing, Substance or Process
• Quantities are used to describe the continuous
  properties of entities
• Links represent relationships, labels are drawn
  from a fixed set of relationships

9
(No Transcript)
10
VModel

• Small, fixed set of structural relationships
• Touches, Contains, Part of...
• Ordinal relationships
• Greater Than, Less Than, Equal To
• Requires relationship links a process to the
  conditions that enable it
• Casual relationships
• Increases, Decreases direct influences (I/I-)
• Influences, InfluencesOpposite qualitative
  proportionalities

11
VModel

• Two coaches
• Qualitative simulation to help the students see
  how well their model make predictions
• Modeling equivalent of spelling and grammar
  checking
• Three sources of feedback
• Visual step-by-step animation of the simulation
• English summary of behavior predicted by the
  model
• Assessment of how well the model supports the
  hypothesis

12
VModel

• Model library contains all the models that the
  students have created
• Students can build their own domain theory

13
HOMER and VISIGARP

• Allow learners to use a qualitative reasoning
  engine for running and inspecting simulations
• They work on top of the domain independent
  qualitative reasoning engine GARP
• Use diagrammatic reprsentations for buliding and
  inspecting qualitative models and simulations

14
Model Building with HOMER

• Organized set of builders and tools
• Builders
• Capture knowledge and use diagrammatic
  representations
• Used for creating building blocks (entitiy
  hierarchy, quantities, quantity spaces...)
• Tools Interactive dialogues for modifying the
  content of builders
• Constructs
• Model fragments and scenarios
• Assembled from building blocks

15
Model Building with HOMER

• Task
• Create a set of model fragments (stored in
  library)
• Specify one or more scenarios (structural
  description of the system)
• Model Building
• Simulator uses the model fragments to predict the
  behavior of the system defined in the selected
  scenario
• For each of the specified scenarios, the
  simulator generates the intended behavior graph

16
(No Transcript)
17
HOMER Prevents learners from making syntatically
incorrect models

• The user interface restricts user actions based
  on
• Content
• Current selections in the builder that the user
  is working on
• Investigates each user action with respect to
  side-effects
• Notifies the user
• Gives user the option either carry on with the
  action or cancel it

18
Problems encountered using HOMER

• Homer was tested on two researchers and two
• master students from a computer science
• department
• Problems caused by poor use of the tool
• Problems caused by subjects not fully
  understanding how to perform a task
• (model building problems)

19
Model building problems

• Model scope
• Determining which features of the real-world
  sytem to include in the model
• Model Structure
• Determining what to put where in the model
• Model Building Concepts
• Understanding the meaning and difference between
  the concepts provided by the tool
• Model Representations
• Knowing the concept, what to represent but not
  knowing how to represent it

20
VISIGARP

• Provides a graphic interface to for running and
  inspecting qualitative simulations
• Simulations might use models constructed by
• teachers
• domain experts
• the learners (for example, by using HOMER)

21
(No Transcript)
22
VISIGARP

• Visual container All model ingredients belonging
  to a particular area are grouped into a single
  box
• Ingredients relating aspects from different
  entities cross the border of the boxes

23
(No Transcript)
24
Usability of VISIGARP

• Experiment was conducted to 30 first year
  university (psychology) students
• Pretests and posttests were conducted on domain
  knowledge and icon language of VISIGARP
• Treatment Prediction exercises using simulations
• Subjects were asked to evaluate VISIGARP after
  treatment

25
Usability of VISIGARP

• Results indicate that
• There was significant difference between pre/post
  tests in domain knowledge
• No significant difference between pre/post tests
  in icon language
• Most icons were easy to learn , lt, , , P,
  P-, I, I-
• A few icons were hard to understand Q, Q, V, V
  (they were not used much during simulation)
• VISIGARP did not always produce insighful graphs
  for comlex models
• Attitude test revealed that subjects evalated the
  usefulness of VISIGARP quite positively

26
Conclusion

• Importance of conceptual knowledge in education
• Reasoning about system behavior
• Two qualitative models VModel and HOMER/VISIGARP
  combination
• Both foster learning as a constructive process

27
Comments

• Experiments are done with subjects at the
  university level
• Further adjustments may be needed if it is going
  to be used in high-middle school level (more user
  friendly)
• Suitable for the new curriculum in Turkey, but
  not easy to put into practice