Dr. habil Erica Melis

1
Educational Technologies WS2006

• Dr. habil Erica Melis
• ActiveMath- Group
• Deutsches Forschungszentrum
• für Künstliche Intelligenz (DFKI)

2
About the Field and the Course

• Intelligent assistent systems for learning
• components of ITSs
• AI-techniques and related ones
• Practical applications
• Interdisciplinary and empirically validated
• Learn actively!!!
• Test described software on Web if available
• Make suggestions yourself
• Hands-on experience and authoring in projects

3
Scheme of the Course

• http//www.activemath.org/teaching/edtechsws0607
• register with Matrikelnummer
• Projects
• Start as soon as possible
• Author interactive script in ActiveMath
• Inform george_at_activemath.org about groups til end
  of week
• Not everything on the slides

4
Approximate Plan of the Course

• 18.10. Introduction and overview
• 25.10. Introduction to ActiveMath
• XML- Knowledge Representation
• 8.11. Student Modelling
• 15.11. Web technologies and security
• 22.11. Tutorial Planning and instructional design
• 29.11. Media Principles
• 6.12. Interactive exercises
• 13.12. Authoring tools, CTAT
• 20.12. Diagnosis model tracing and domain
  reasoning
• 10.1. Diagnosis constraint based
• 17.1. Tutorial dialogues
• 24.1. Action analysis and ML techniques
• 31.1. Cognitive tools
• 7.2. Meta-cognitive support
• 14.2. student projects

5
Why Technology-Enhanced Learning ?

• Independent of time place
• Individual tutoring
• better learning (modalities visualization..)
• (Semi)-automatic assessment
• Information for teachers
• cost effective
• Knowledge resources from the Web

• Distance learning
• Virtual Universities
• Training on the job
• Military training
• Training for disabled

6
Data1From Statistics Bulletin on Economic
and Social Development in
P.R.China 2004
Admission Proportion for High Education
19 4,200,000
81 17,905,000
7
Data 2 From CNNIC (China Network Information
Center) Statistic Report on the Development
Status of Network in China
Increment Rate 18.9
Million Person
199M in America From ComScore
94M
79M
69M
45.8M
26.5M
16.9M
The only purpose of 8.4 (7.89 million) users going online is for education
21.3 users prefer more educational information, 20 million broadband users
8
History First Generation of Tutors, CAI
1960ies Programmed instruction 1970ies CAI..
PLATO, SHIVA

• IF the correct response THEN present new element
  ELSE goto
• Computer-Aided Instruction (CAI) or CAL
• store and retrieve data, exercise bank with
  answers
• pre-defined branches of problem solving
• no understanding of problems, few anticipated
  wrong answers
• Independent of students understanding,
  preferenes, behaviour
• linear (not individualized) progression of
  instruction
• no diagnosis of errors

9
History Second Generation of Tutors, ITS
1970iesScholar Carbonell,Brown
1990iesPACT Anderson

• Internal domain representation knowledge base
• Problem solver, inference engine (XPS)
• -gt cause of errors
• -gt more appr. response
• Exercise bank
• Limited dialogue and QA

• Student modeling
• Domain Expert Module
• Model learners errors
• Tutoring module (intervention modalities)
• Pedcognitive theories developed
• more autonomous student
• Lab- and realistic evaluations
• bandwidth of user interface,
• more variety of responses
• more interaction

10
CAIITS1 Architectures
Knowledge Base
Exercise bank
Expert system
selector
User interface
11
History Third-Generation Learning Systems

• More student modeling
• emotional, motivational, affective, situational
• learning from massive log data
• Natural language tutorial dialogues
• Explorative, interactive, inquiry learning
• Collaborative learning
• Support of meta-cognition
• Web-based systems
• Multimedia and (adaptive) hypermedia based on
  pedagogy
• Semantic knowledge representation (semantic web)
• Retention tests, social skills,
  performance/learning
• AHA, Tectonica, ActiveMath, ELM-ART, Edutella,
  Wayang Outpost, iHelp, Algebra Cognitive Tutor,
  BEETLE, Help Tutor

12
A Generic ITS Architecture
Intelligent Tutorial Component
Curriculum Planner
Problem Selector
Problem Solver
Domain KR
Student
Solution Graph
Curriculum
Action Interpreter
Solution Evaluator
Interaction History
Student Model
Feedback Generator
13
Andes Architecture
Authoring Environment
Student Environment
Workbench
Problem Presentation
Assessor (BN)
Physics Rules
Action Interpreter
Solution Graph
Physics Problem Solver
Student Model
Problem Definition
Help System
tutoring strategy
Procedural help
Conceptual help
Example study help
14
ActiveMath MVC Architecture
15
What is an IST for a Learner

• Intelligent features
• Personalization
• Interactive problem solving
• Error diagnosis and feedback
• Mixed-initiative control
• Tutorial dialogue
• Open learner model
• Interactions
• Tools (search)
• Personalization of
• Content sequencing
• Presentation
• Generated suggestions
• Feedback

16
Some Intelligent Systems

• Cognitive Tutors (Koedinger et al)
• ELM-ART (Weber et al)
• Andes, Atlas-Andes (vanLehn et al)
• Cabri-Geometre (Balacheff et al)
• Wayang Outpost (Wolff,Arroyo,Murray)
• ActiveMath www.activemath.org (Melis et al)
• Belvedere (Suthers)
• I-Help (Greer et al)
• Tectonica, AHA (Murray et al, deBraAroyo)
• AutoTutor, BEETLE (Graesser et al, Moore et al)
• Help-Tutor (Aleven, Koedinger)

17
Interdisciplinary Field
AI
CoLinguistics
TEL
Cognitive Psychology
Web-Technology Multimedia
Pedagogy
Content
18
Contributions of AI

• Knowledge representation
• User modelling
• Intelligent user interfaces
• Presentation planning, intelligent sequencing
• Diagnosis
• Data mining, Machine Learning
• Problem solving systems/automated reasoning
• Agent-based (help) systems
• Adaptive hypermedia

19
AI User modeling

• Bayesian nets

Probability distribution events, causes,
evidences conditional dependences diagnostic/causa
l update
20
AI Knowledge Representation
Frames in Cognitive Tutors
Problem WME
(make-wme composed-cen-insc isa problem
key-quantities (angle-KHP-measure arc-KP-measure
angle-KQP-measure) key-reasons
(angle-KHP-measure ...) questions (question1)
given-relational-quantities (central-angle-KHP
inscribed-angle-KQP) table composed-cen-insc-tab
le )
Relation WME...
inscribed-angle... inputs (arc-KP-measure) output
angle-KQP-measure
Quantity WME ...
angle-KHP-measure...unit..dimension..labels..
21
AI Knowledge Representation

• Semantic networks
• DAML/OIL/OWL decision logics for
  XML-Representation
• Meta data (publ, mathematical, pedagogical)

Ontology
22
Web-Languages and Technologies
Standardization!!!

• IEEE LTSC, LOM
• IMS Global Learning Consortium
• Apple
• Cisco
• IBM
• Microsoft
• Sun
• WebCT
• Universities
• .
• Open e-Book

• Meta data
• Interoperability of services
• Interoperabilty of content (ontologies)
• Architectures
• Presentation of content
• Wiki
• Security

23
Contributions from Pedagogy
difficulty

• Goals
• Content sequence
• Strategies, Methods
• Media, tools
• Competencies

• Didactic
• Socratic
• Inquiry
• Discovery
• LearnNew
• Rehearse
• Collaborate

exercises
Handling errors Frequent mistakes Feedback Multipl
e solutions
MultiMedia
user modeling (competencies)
24
Bloom taxonomy of educational objectives
25
PISA Competencies

• Compute
• Apply
• Model
• Argue
• Solve problem
• Collaborate
• Use tools
• Meta-cognition

26
Contributions Cognitive Psychology

• Behaviourisms vs. constructivisms Piaget,
  Vygotski
• Feedback
• motivation personalized, self-guided, social,
  active DecyRyan...
• zone of proximal development Vygotsky
• gender-specific
• meta-cognition White
• adaptive support Mandl...
• multi-modality
• structured presentation of solutions Catrambone

Effective design vs on-line book with animations
27
Cognitive Psychology Multimedia Learning

• Multimedia Principle
• Integration Principle
• Modality Principle
• Redundancy P.
• Coherence Principle
• Personalization P.
• Learner control

28
Cognitive Psychology some results

• Self-explanation of worked-out examples
  Renkl,Chi,Merrinboer,Siegler
• Why does tutorial dialog help? Chi etal 2001
• even if human tutors dont know tutoring
• no-content prompts
• ask, dont tell ?
• students own communication?
• Learning from errors/impasses only (?)
• Conceptual change (Vosniadou)
• Influence of motivation, self-efficiacy Bandura
• Evaluation of systems

29
Conclusion

• Pursue learning
• Learn actively and believe in yourself
• Ask questions if you dont understand
• Discover the world of research

30
Student Projects

• 1.Visualization of the pedagogical knowledge
  domain
• Analyze and visualize the structure of
  pedagogical tasks
• 2. SLOPERT exercise generator
• Explore the problem space and create a ActiveMath
  exercises.
• 3. Learner Model for iCMap
• Catch and analyze events generated by iCMap
• 4. Domain Viewer
• Render an ActiveMath domain (concepts, relations)
• 5. Exercise generation with extended randomizer
• to support intervals and (adaptive) randomizing
  over a set of elementary functions and their
  compositions

31
Student Projects

• 6. Mathematical Rendering Tester
• Support authors by rendering mathematical
  formulae on the fly
• 7. Analyzing Online Collaborative Data
• Generate Machine Learning classifiers from log
  data
• 8. E-Portfolio Viewer
• Implement an interactive viewer for the IMS eP
  Spec