Lecture 5: Building and Exploring Domain Tools

1
Lecture 5 Building and Exploring Domain Tools
2
Contents

• Building Domain Models
• LOGO experience - model building
• Intermodeller
• Conception
• Simulation Authoring Tools RIDES
• Pedagogical Agents

3
1. Building Domain Models
4
Learning through Model Building

• Assumes learner is active and seeking stimulation
• Making knowledge explicit
• get learner to communicate beliefs
• get learner to model theories and test them
• get learner to reflect on learning
• Learning through confrontation
• student has belief of what happens in environment
• tests belief -gt consequences in environment
• if consequences don't match belief, then (hope)
• cause student to review belief (learn)
• But, are all confrontations beneficial?

5
Quote from Mark Elsom-Cook

• r.e. discovery learning environments in general
• The process of externalisation making explicit
  the learner's model of the educational setting
  makes a concrete representation of the internal
  processes of the learner.
• By seeing the external form of the predictions
  made by these processes, the pupil realises the
  limitations of the external representation,
  repairs it and, in the process, repairs his/her
  internal representation.

6
Some Advantages

• Student sets goals, agenda, and structures
  learning themselves - own v. imposed goals
• Structure provided by environment has to be
  non-intrusive, but supportive
• Allows learner to directly experience behaviour
  inaccessible in real world
• Self-directed learner --gt interaction --gt
  stimulation
• Learner is actively engaged in developing and
  testing theories (echoes child-centred/progressive
  education claims)

7
A distinction

• A. Simulations, Exploratory Environments,
  Micro-worlds
• the investigation of views of a given domain,
  which may differ from the learner's own
• B. Expressive or Modelling Environments
• the modelling of user's own beliefs, and
  reflecting on and exploring own models

8
2. The LOGO Experience
9
Model Building in LOGO

• Allows student to explore their own models by
  building programs written in the programming
  language LOGO
• LOGO - Feurzeig and Papert (1969)
• - based on LISP
• - provides language for describing procedures
• Piaget
• "If you want a child to learn, begin by taking
  something he already knows, and use it as a
  framework within which new learning takes place."

10
Turtles and Button Boxes

• So, start with body awareness and relate body
  movements to those of a small robot.
• Control the robot (turtle) with commands - see
  effect of drawing device attached.
• If doesn't do what expect, try to act out what it
  does do and correct it.
• slides of Turtle and Button Box, figures 1.1 and
  1.2 from O'Shea and Self

11
LOGO uses

• Subjects explored through problem solving
  include
• programming
• maths
• Language
• physics
• Child learns the language and also learns
  problem-solving skills such as problem
  decomposition in the course of writing
  procedures.

12
Claims and Problems

• Claims
• student gets a problem solving methodology
• generalisable to other problems.
• Problems
• Syntax of LOGO - language may be harder than
• domain to be taught
• What is the goal?
• Does it transfer?
• Evaluation
• no clear evidence of generalisable problem
  solving skills through LOGO
• some evidence of improved learning in the domains
  where LOGO used

13
Issues about how used

• PAPERT
• if integrate in existing class this may
  prejudice any success
• take whole new approach to curriculum with it
• only need student and environment
• V. EDINBURGH
• need for some structure to support it - can't
  just say "explore, learn!"
• need additional guidance too

14
However.

• Not all environments equally suited for all
  learning
• a design may facilitate one model
• but be incompatible with another
• (e.g. physics models, Newtonian v. Relativity),
• therefore make different predictions for
  different models
• Should environment suit all users?
• Build so that can hold onto bits of the
  environment and develop better models from them.
• Tools must be meaningful to learner
• turtle geometry assumes can relate movements to
  motor skills.

15
3. Intermodeller
16
Expert System Shells

• Can be used to build knowledge models
• But commercial ones not suitable for schools
• Various rule based shells developed for schools
• But used mostly in technology classes - not as
  modelling tools in the wider curriculum
• Led to research by Conlon on why, and how tools
  might be improved to make them both useful and
  usable

17
INTERMODELLER

• a computer program intended to support children
  in building classification models
• can be in many domains a child who is learning
  about spiders, dinosaurs or planets can build a
  model of his or her developing knowledge of the
  domain
• models once constructed can be run as small-scale
  expert systems that perform interactive
  classification
• models can be demonstrated and discussed, shared
  as files across networks, pasted into word
  processors and graphics programs, printed out to
  create classroom display material.

18
Conlon, 1999

• .. effective support for constructive thinking
  in classification can be provided by a
  model-building environment in which learners
  create software representations (or models) of
  classification structures. The environment
  provides editors for building these models and an
  interpreter which can run any model as an
  interactive classifier.

19
Interchangeable Representations

• Classification trees organise classes or
  categories hierarchically, with arcs representing
  subclass relationships
• Decision trees flowchart-like representations
  which use a branching structure of questions and
  answers to distinguish between categories
• Decision (factor) tables category named in the
  right column is defined by a row specifying
  features as values of attributes named on the
  header row
• Rules statements of if/then relationships
• Graphical tools for Knowledge Base development

20
Screenshot of Intermodeller
21
Model building methods

• The methodology comprises the following seven
  steps
• Decide on the purpose of the model
• Identify decision factors (factors used to
  distinguish different categories)
• Select a form of representation
• Review the design
• Start the model
• Develop the model
• Reflect and evaluate

22
Functionality

• Learner can switch between representations
• a knowledge base can be switched between
  representational forms automatically.
• In-built machine-learning
• models can be automatically slimmed to improve
  efficiency (uses ACLS induction).
• Full expert-system style runtime features
• how and why explanation
• certainty handling
• consultation review

23
Evaluation of modelling course

• 82 children, 15 years old, c. 8 hours of class
  time.
• 632 models - found that rule models not as high
  in quality (by indices of correctness,
  efficiency, and conciseness) as those built using
  the alternative factor table, decision tree and
  classification tree representations.
• Questionnaire responses
• children least enjoyed working with rules.
• Children's ability to construct representations
  of classification improved significantly as a
  result of the modelling course.
• (Say more in relation to methodology later)

24
4. Conception
25
Using Multiple Representations

• Users with a repertoire of representational
  skills can describe, reason with and build models
  of information.
• Constructivist learning theory stresses
  relationship between building internal mental
  models and building external information models
  (Cox Brna 1995).
• Cox Brna showed that when people reach an
  impasse or become 'stuck' while solving a
  problem, the ability to reformulate the problem
  using a different type of representation can be
  an effective way of making progress.
• By making thinking visible, the process of
  constructing an information map can help to
  understand, refine and communicate ideas.

26
Information Maps as Classroom Tools

• Used by teachers to communicate information
• Used by pupils to learn subject matter.
• Building information maps improves generic
  representational techniques and transferable
  thinking skills.
• eg.1 creating an argument map about transport
  policy - learn that argument can be understood as
  a hierarchical structure of claims,
  justifications and objections.
• eg.2 creating a decision map about the budget -
• learn that decision-making process can be
  understood in terms of options, factors and
  evaluations.

27
(No Transcript)
28
(No Transcript)
29
Argument mapping

• Informal discussion this involves learners
  working in twos or threes around a computer,
  debating a main claim while simultaneously
  constructing an argument map.
• Preparation for formal classroom debates
• Planning an essay
• Reporting on research

30
(No Transcript)
31
Map types and applications
32
Concept mapping by learners

• Concept mapping activities by learners can be
  placed into three categories
• Tabula rasa ('blank slate') mapping involves the
  creation of a map from scratch
• Scaffolded mapping tasks elements of the map are
  provided by the teacher, leaving the learner to
  supply the rest.
• Buggy map correction tasks present learners with
  concept maps containing deliberately introduced
  bugs (errors). Learners task is to locate and
  correct them.

33
(No Transcript)
34
(No Transcript)
35
(No Transcript)
36
5. Simulation Authoring Tools RIDES
37
IMTS, RAPIDS and RIDES

• Munro et al, 1997 University of Southern
  California
• IMTS (Towne and Munro, 1988, 1992)
• tools for authoring interactive graphical
  simulations of electrical and hydraulic systems
• supported troubleshooting assistance but could
  not deliver other kinds of instruction
• RAPIDS (Towne and Munro, 1991) and RAPIDS II
• (Coller, Pizzini, Wogulis, Munro, and Towne,
  1991)
• direct manipulation authoring of instructional
  content in the context of graphical simulations
• no low level control over instructional
  presentations
• Both only available on specialized AI
  workstations
• RIDES more robust, less constrained, simulation
  authoring tools and instructional editing
  facilities

38
RIDES

• Integrated software environment for computer
  based tutorial instruction and practice
• Uses interactive graphical simulations of devices
  or other complex systems
• Students explore graphical simulations.
• Build graphical models by fetching library
  objects and pasting them into scenes.
• Draw objects directly onto the scenes
• Behavior of objects specified by authoring rules
  that control the value of object attributes.

39
RIDES Course

• A course a set of learning objectives that must
  be realized by a student
• each associated with a lesson for teaching it
• student model based on course objectives
• what to present next controlled by relationships
  between course objectives and student model
• Instruction author can 'record' a procedure that
  students must learn, by carrying out the
  procedure.
• During training, the student is prompted to carry
  out the correct sequence of actions.

40
Claimed advantages

• For the student
• More robust, more realistic simulations
• Support for free play immersive learning
• For the developer and the manager
• Potential for the reuse of objects
• Ease in developing more robust, flexible, and
  realistic simulations

41

• Graphical tutor, simulates behaviour of human
  circulatory system, by Carol Horwitz at
  Armstrong Laboratory, USAF.

42
Control panels for ground support unit for
maintenance of the B-2 bomber, by Stephanie
Perdomo, Northrop Grumman Corporation
43
Pulse oximeter, for monitoring patient's pulse
and blood oxygen, Carol Horwitz, Armstrong
Laboratory, USAF.
44
Diesel engine, Michael Crumm, Armstrong Lab,
USAF.
45
Pedagogical Agents

• Steve agent cohabits a 3D simulated mockup of a
  student's (complex) real work environment (US
  Navy ship).
• The agent can demonstrate physical tasks, such as
  operation and repair of equipment.
• Students can learn and practice skills in a
  virtual world
• (Rickel and Johnson, 99)

46
Steve and another agent interacting
47
(No Transcript)
48
(No Transcript)
49
Trends in Domain Modelling

• Multiple representations of Domain Knowledge
• Simulations
• Simulation Authoring Tools
• Further Use of Qualitative Reasoning Methods
• Empirically Informed Design