Transcript E-Learning FORUM 21 Martie 2003 SOCRATES

E-Learning FORUM
21 Martie 2003
Intelligent eLearning
Environments
Paul Dan Cristea
“Politehnica” University of Bucharest
Spl. Independentei 313, 77206 Bucharest, Romania,
Phone: +40 -21- 411 44 37, Fax: +40 -21- 410 44 14
e-mail: [email protected]
E-Learning FORUM
21 Martie 2003
SOCRATES - MINERVA PROJECT
87574-CP-1-2000-1-RO-MINERVA-ODL
Artificial Intelligence and
Neural Network Tools for
Innovative ODL
Coordinator : “Politehnica” University of Bucharest
Partners
• Vrije Universiteit Brussels, BE
Prof. Jan Cornelis, Head of Electronics & Digital Signal Processing Department
Prof. Edgard Nyssen, Prof. Rudi Deklerck
• Universitat Erlangen - Nürnberg, DE
Prof. Manfred Kessler, Director of Institute fur Physiologie und Kardiologie
• University of La Rochelle, FR
Prof. Patrice Bourcier, Assistant Director of Information and Industrial Imaging Lab.
• Universidade Nova de Lisboa, PT
Prof. Adolfo Steiger Garcao, President of UNINOVA
Prof. Jose Manuel Fonseca
• University of Edinburgh, UK
Dr. Judy Hardy, Applications Consultant at EPCC
• Patras University, GR
Prof. Nicolas Pallikarakis, Coordinator of BioMedical Engineering Scool
• Global One Communications Romania, RO
Dr. Pavel Budiu, Strategy Manager
Objectives
Main goal : develop and use a set of innovative ODL tools
for on-line and Internet-based learning, using the methods
and techniques of artificial intelligence and neural networks.
O1. Provide a model of the collaborative learning process
involving human and artificial intelligent agents;
O2. Provide a set of tools based on AI&NN techniques
to develop innovative ODL systems;
O3. Carry out pilot implementations of ODL systems;
O4. Develop a methodology for intelligent ODL production
and performance evaluation;
O5. Evaluate and disseminate the outcomes of the project
for future developments.
Contractual Time Table
Start of eligibility period
1 October 2000
Submission of 1st
Interim Report
1 June 2001
Submission of 2nd
Interim Report
1 June 2002
End of Eligibility Period
1 September 2003
Submission of
Final Report
1 November 2003
Workpackages and Responsabilities
WP0: Project Management, Monitoring and Reporting (PMMR)
PUB + PMG
WP1: Collaborative Learning Model (CLM)
ULR + PUB + UP
WP2: Learner’s Profile Eliciting Tool (LPET) EPCC + PUB + GOC
WP3: Automatic Tutoring Tool (ATT)
UNL + ULR + PUB + VUB
WP4: Learner’s Personal Assistant (LPA)
PUB + UNL + UEN + GOC
WP5: ODL courses on Bio-Medical Data Processing and
Visualisation (BMDPV) using the new AI&NN tools
BMDPV – M1: Medical visualisation
UEN + PUB + VUB
BMDPV – M2: Cortical brain anatomy VUB + PUB + UP + UEN
WP6: Elaboration of Instructions, Guidelines, and Examples of
integrating the AI&NN tools with existent ODL materials (IGE)
UP + UPB + EPCC + all
WP7: Testing, evaluation, assessment and dissemination (TEAD)
of AI&NN tools for innovative ODL
PUB + all
• Professional qualification is no longer a life-long
achievement
• Complex knowledge and skills have to be
transmitted and acquired efficiently
• E- Learning will play a continuously increasing
role.
• Intelligent educational tools can bring the flexibility
and adaptability required to actively support the
learner.
Basic paradigms:
• Intelligent Human-Computer Interaction
• Computer-Supported Cooperative Work (CSCW)
Learning in the system: Cooperative learning by interaction between
student and tutor/expert or inside the group of learners
Organization: Group of learners assisted by artificial agents with
active role in the learning process.
Tutor: Human or artificial agent
Structural features:
• Set of tools to assist the learner at several levels of the knowledge
acquisition process.
• Personalised model of the trainee
Combine the traditional style of teaching
with the problem-centered style:
• learning by being told,
• problem solving demonstration,
• problem solution analysis,
• problem solving,
• creative learning
Knowledge transfer
Learning
by being
t old
Problem
solving
demo
Skill development
Solut ion
analysis
Problem
solving
Level of learner’s active participation
Creat ive
learning
Tut or A gent
Module for accessing learning
resources and managing
interactions:
- Problem Solving KB
- Selection of relevant
knowledge
- Coordination activities
Communication
Module
Control Module
Module to select
learning modalities
and to adapt to
learner profile
Problem
Solving
Knowledge
Base (PSKB)
Module to respond
to learner requests
and needs
Tutor Agent KB:
 Knowledge to access
PSKB
 Methodological
 Knowledge on how to
adjust to learner
profile
Other
agents
Tut or A ssi st ant A gent
Module for accessing learning
resources and managing
interactions:
- Problem Solving KB
- Coordination activities
Communication
Module
Control Module
Module to extract:
- tutoring knowldege
- tutoring strategies
- creative learning
experiences
Problem
Solving
Knowledge
Base (PSKB)
Module responsible
for monitoring tutor
actions and guiding
Tutor KB:
 Knowledge to
retreive elements
from PSKB
 Training history
 Elicited tutoring
knowledge
Other
agents
Lear ner Per sonal A gent
Module for accessing learning
resources and managing
interactions:
- Problem Solving KB
- Coordination activities
Communication
Module
Control Module
Module to develop
the learner profile
Problem
Solving KB
Module responsible
for monitoring
learner actions and
requests
Learner KB: learner
history and learner
profile
Other
agents
Learner’s Profile Eliciting Tool
Control Module
Student
input
Registration
form
Learning
Objectives
Communication Module
Learning
Modalities
Student
Tracking
Tool
Knowledge
Watch
Questionnaires
Content
Management
Tutor
input
On-line
students
monitoring
Validation of
students
proposals
• Curricular study
for a diploma
• Complementary
study
• Executive
up-dating
• Specialist
up-dating
• Problem centered
• Test oriented
Preferredly /
Predominantly:
• Descriptive
• Demo
• Analytical details
• Practical aspects
• Examples
• Multimedia / Text
Material
to study
1 First Chapter xxxxxxxxxxxxxxxxxxxxxxxxxxxxxxx
1.1 Section 1.1 xxxxxxxxxxxxxxxxxxxxxxxxxxx
1.1.1. Paragraph xxxxxxxxxxxxxxxxxxxxxX
1.1.2. Paragraph xxxxxxxxxxxxxxxxxxxxxX
1.1.3. Paragraph xxxxxxxxxxxxxxxxxxxxxX
1.2 Section 1.2 xxxxxxxxxxxxxxxxxxxxxxxxxx
1.2.1. Paragraph xxxxxxxxxxxxxxxxxxxxxx
1.2.2. Paragraph xxxxxxxxxxxxxxxxxxxxxX
1.2.3. Paragraph xxxxxxxxxxxxxxxxxxxxxX
1.3 Section 1.3 xxxxxxxxxxxxxxxxxxxxxxxxxxx
1.3.1. Paragraph xxxxxxxxxxxxxxxxxxxxxx
1.3.2. Paragraph xxxxxxxxxxxxxxxxxxxxxx
1.3.3. Paragraph xxxxxxxxxxxxxxxxxxxxxx
2 Second Chapter xxxxxxxxxxxxxxxxxxxxxxxxxxxx
2.1 Section 2.1 xxxxxxxxxxxxxxxxxxxxxxxxxxx
2.1.1. Paragraph xxxxxxxxxxxxxxxxxxxxxx
2.1.2. Paragraph xxxxxxxxxxxxxxxxxxxxxX
2.1.3. Paragraph xxxxxxxxxxxxxxxxxxxxxx
…………………………………
Standard
Path
?
Studied
material
1 First Chapter xxxxxxxxxxxxxxxxxxxxxxxxxxxxxx
1.1 Section 1.1 xxxxxxxxxxxxxxxxxxxxxxxxxx
1.1.1. Paragraph xxxxxxxxxxxxxxxxxxxxxxxxxxxxx
1.1.2. Paragraph xxxxxxxxxxxxxxxxxxxxx
1.1.3. Paragraph xxxxxxxxxxxxxxxxxxxx
1.2 Section 1.2 xxxxxxxxxxxxxxxxxxxxxxxxxx
1.2.1. Paragraph xxxxxxxxxxxxxxxxxxxx
1.2.2. Paragraph xxxxxxxxxxxxxxxxxxxxx
1.2.3. Paragraph xxxxxxxxxxxxxxxxxxxxx
1.3 Section 1.3 xxxxxxxxxxxxxxxxxxxxxxxxxx
1.3.1. Paragraph xxxxxxxxxxxxxxxxxxxx
1.3.2. Paragraph xxxxxxxxxxxxxxxxxxxxx
1.3.3. Paragraph xxxxxxxxxxxxxxxxxxxx
2 Second Chapter xxxxxxxxxxxxxxxxxxxxxxxxxxx
2.1 Section 2.1 xxxxxxxxxxxxxxxxxxxxxxxxxx
2.1.1. Paragraph xxxxxxxxxxxxxxxxxxxxx
2.1.2. Paragraph xxxxxxxxxxxxxxxxxxxxx
2.1.3. Paragraph xxxxxxxxxxxxxxxxxxxxx
…………………………………
Recommended
Path
Self
Testing
Mandatory
Testing
Contribution to
Collaborative
Learning
No purely empirical approach to modelling.
Even the definition of attributes/features &
the selection of the relevant ones in a given context
are actually theory driven, explicitly or not.
Prototype model of the learner
• Encodes general theoretical knowledge in the field of learning.
• Can not be used directly in practice - rigid and biased:
• Large variability in human personality and in human behaviour,
• The essential traits are context-dependent.
Customised model by using empirical data - sets of examples
collected for the given user, while interacting with the system.
New refined theory
If tuning parameters can not adapt the model to user's profile,
new features are extracted from data and added to the model.
No systematic way to empirically identify the domains of the
feature space that are not properly represented in a set of examples.
• The available collection of examples is never large enough to cover
all the possible classes in an unbiased manner, to avoid spurious
correlation when elaborating a model.
• Small sets of exceptions may be poorly represented or even ignored.
The underlying theory
• helps eliminate irrelevant features,
• guides the selection of relevant examples to scan of the input space,
• gives confidence in the solutions produced.
A purely theoretical approach may be brittle, i.e.,
• can yield dramatically incorrect results for exceptions,
• scores of instances that fall in the limits of validity domain are
treated correctly (abrupt degradation).
Exhaustive theories may become intractable
• The domain of validity must be restricted.
• Compromise scope - accuracy.
Combined use of theoretical knowledge and experimental results allows:
• Incomplete and/or incorrect theoretic knowledge,
keeps the model in the range of an acceptable approximation.
• Incomplete or noisy experimental data
inherent ability to recover from errors.
The user model being developed uses a hybrid approach:
• Artificial Intelligence (AI) -- symbolic representation of theory,
• Neural network (NN) -- sub-symbolic representation of data.
NN has the ability to represent "empirical knowledge",
but behaves almost like a black box:
• Information expressed in sub-symbolic form,
not directly readable for the human user
• No explanation to justify the decisions in various instances,
forbids the direct usage of NNs in learning/teaching and
safety critical areas
• Difficult to verify and debug software that includes NNs.
Extraction of the knowledge contained in an NN allows the portability
• to other systems in symbolic (AI) and sub-symbolic (NN) forms,
• towards human users.
AI and NN approaches are complementary in many aspects
• can mutually offset weaknesses and alleviate inherent problems,
• able to exploit both theoretical and empirical data - hybrid aproach,
• efficient to build a fault tolerant and adaptive model,
• help discover salient features in the input data.
First phase. The system operates using statistics about:
• which buttons were selected by the lerner when using the system,
• in which order,
• which error messages have been generated.
The system is trained to use this input to offer advice in the form of
• access to some additional data and information,
• additional reading,
• recommend or trigger an interaction with the human tutor.
Subsequent phase. The system uses:
• error databases,
• special interest databases,
• preference databases,
including the input from a human tutor.
The output helps identifying some profile of the user,
defined roughly by the set of classes the user belongs to.
This influences the future interaction of the system with the user,
e.g., changing the type and level of the exercises presented to the user.
Next step. The system includes some voluntary feedback learners,
offered to all the other learners, to help conveying original ideas
and generate groups of interest.
Increase of tutor "productivity“. The system is a useful assistant,
not a replacement of the human tutor.
The work done traditionally by two or three tutors could be
accomplished in this approach by only one assisted tutor.
The basic contribution of this research is twofold:
• Identification of several Learning Modalities that combine
traditional teaching with “problem-centred” learning
to better motivate the student and to increase the efficiency
of the learning process,
• Conception of a Collaborative Distance Learning System in which
human and artificial agents collaborate to achieve a learning task.
The Tutor Agent tries to replace partially the human teacher, in
assisting the learners at any time of their convenience.
The development of the learning system is a collaborative effort
to develop a novel intelligent virtual environment for ODL at
“Politehnica” University of Bucharest.
The system is currently under development; several components
written in Java are already functional.
To test the system, we are concurrently developing learning materials on:
• Sorting Algorithms,
• Resolution Theorem Proving,
• Neural Networks,
• Advanced Digital Signal Processing.
The distributed solution has the advantage of creating an ODL environment
that can be joined by any interested learner.
The system is an effective response to the
• the increased demand for cooperation and learning in today's open
environments, academic and economic,
• the necessity of developing effective learning tools that can be smoothly
integrated in the professional development process and with company work.
Care is taken to prevent such an approach to generate an "elitist" system.
The system is designed to enhance the specific features of each user,
without increasing the differences between users in what concerns the level
of understanding or the ability to creatively use the acquired knowledge.