Transcript Document

PANEL: Intelligent Learning Tools;
Adaptation to Non-stationary Environments;
Learning & Evolution
An Adaptive Distance Learning
Environment for Language Teaching
Alexandra Cristea
Toshio Okamoto
The University of Electro-Communications, Tokyo, Japan
English teacher system rationale
• Lifting “the language barrier”.
• present language teaching software
• focuses on general items, irrelevant for academics.
• little flexibility
• academics are people with little time
• important to build a user-oriented
environment, with no time restrictions or other
restrictions of physical nature.
Adaptability – up to where?
• Adaptive environments should allow users
to make decisions and correct directly the
user model, or at least its effects
• So, both lazy user, who prefers to be told
things, and dynamic user, bored by standard
ways, are satisfied.
• Main
question:
balance
between
adaptability and user-driven software.
Our adaptation
• Implicit (user tracing) &
explicit (user input)
• Symbolic (labels, pointers) &
sub-symbolic (weights)
• For more information:
• “Student model-based, agent-managed,
adaptive Distance Learning Environment for
Academic English Teaching” (paper at IWALT)
• Answers to questions from chair
(1) How do you define an ILT? What
should be its basic features
& minimum requests?
• An ILT should
• serve for learning, so correspond to some specific or general
learning goal
• Intelligence in ILT should mean user adaptation; therefore,
some minimal user modeling is necessary
• Being a tool and not an environment, it can assure only a part
of the learning goals, or be able to assure learning goal
achievement only together with other tools
• IEEE Learning Technology Standards Committee (LTSC)
(2) Share with us your experience in
using ILTs.
• Prof. De Bra, TUE, Netherlands, examples of
adaptive learning environments:
• http://wwwis.win.tue.nl/~debra/
• MyEnglishTeacher:
• http://www.ai.is.uec.ac.jp/u/alex/MyEnglishTeacher/i
ndex.html
• Etc.
Adaptation in educational systems
• Adaptive presentation of educational material
• providing prerequisite, additional or comparative explanations,
• conditional inclusion of fragments, stretch-text,
• providing explanation variants, reordering information, etc.
• Adaptive navigation support
•
•
•
•
direct guidance,
sorting of links,
links annotation, link hiding, link disabling, link removal,
map adaptation, etc.
What is user adaptation in ITS?
• E.g.: switch among pedagogical strategies
( cooperative strategy contexts – Frasson 1998).
Pedagogical strategies
Explanation
Tutor-tutee
Traditional: computer is teacher, user is student
Learning companion
computer-simulated learner, accompanying the user
Learning by disturbing
Learning with a simulated troublemaker.
Learning by teaching
Human student teaches the simulated companion.
Learning with co-teacher
Both simulated teacher and co-teacher
Within contexts, direct strategies exists, e.g: Learning by examples,
learning by story-telling, learning by doing, learning by games, learning
by analogy, discovery learning, learning by induction/ deduction, etc.
Learner & domain models
• learner model
• to switch between strategies;
• In 1996, Greer pointed at the importance of offering adapted
activities & appropriate feedback, favoring communication
between students & offering assistance.
• But: “student’s values, learning style metacognition and
preferences regarding feedback” have to be correctly inferred
• knowledge domain model
• represents the model of the course contents knowledge
• the student model has to be mapped on it
Layers in student models
• latest student models have layered learner models:
• knowledge & cognitive model level, wrapped by:
• learning profile (curricula), wrapped by:
• believability & emotional layer (which, if correctly
interpreted, is supposed to point to the best learner-tailored
pedagogical strategy - Abou-Jaude, Frasson, 1999)
Acquiring knowledge about learners
• ask the learner (straightforward) single/multiple-choice
questionnaires, where learner inputs preferences &
opinion(s) about his/her knowledge level, learning profile,
emotional profile, etc.
• test the learner, to establish his/her profile (knowledge
tests, IQ tests, even personality tests)
• trace & interpret learner’s steps, choices & results
during learning user’s into a learner model (most difficult)
Pros & Contras of knowledge
acquisition methods
• explicit information gathering :
• info correct (if user knows him-/herself).
• user-model building transparent to learner, who can
directly influence it & correct misinterpretations.
• implicit user tracing:
• lets user concentrate on subject at hand
• doesn’t prompt him/her with numerous questions.
 fine balance of modeling methods necessary.
• Prediction: optimal solutions will imply a
combination + fine tuning of fuzzy goals
set:
• user-friendliness,
• low user overhead &
• learning enhancement.
Advantages of ITS & user
modeling on the Web
• Servers can store large amounts of material &
user models from tiny client machines
• great number of (actual/ potential) users on Internet
makes user modeling, average behavior interpretation,
classifications, etc., more meaningful.
(New:nation&region–oriented
classification&adaptation)
• Internet is loaded with (potentially) useful educational
material using more than just local data & facilities
(3) Can an ILT go over the role of
only assisting a human tutor?
• Yes
Automatic adaptation could
become better than classrooms
• Arguable: an adaptive system can perform
better than a classroom teacher, who is
bound to present a classroom average
material
• very convenient if a system makes correct
assumptions, problematic when not
• e.g., “smart” office software package from
Microsoft
Advantages over classroom teaching
• Classical classroom teaching method is
• limited in time
• learning is synchronous (unlike distance-learning)
• A teacher always addresses average pupil (LE can be
customized)
• Media can enhance human aspect of course contents,
( believability level: smoothing transfer from face-to-face
teaching / learning to learning in front of computer)
• Media presentations can also contain extra clarifications,
part of main contents, etc.
(4) Should ILTs be used only for ODL,
or also in standard class teaching?
What would be their specific tasks in
the two cases?
• Both.
• The minimal task requirements and the definition is
the same, as we are talking about tools and not
systems or environments
• (from the tools point of view, it can be used in
collaboration with other tools, material, etc., generated
automatically or by the teacher, for reaching a learning
goal)
(5) How should an ILT evaluate the
performance of a human learner?
• Tracing and tests for finding the best learning
path
• Note: This is different from grading evaluation,
which should be separated from the learning
evaluation
• Accuracy
• Influence on learning
(6) What are the expectations for the
immediate and medium future?
• Fischer 1999 noted:
• new millennium: marked by changing of mindsets:
• teacher: “sage on the stage”  “guide on the side”
• Student: dependent, passive role  self-directed, discovery-oriented
role
• life-long learning
• We have to prepare for the change. Intelligent, mediaoriented distance learning environments are an answer.
• But: focus should always be on learning enhancement
and educational goals.
A last word …
• It is:
• difficult to break with old customs
• dangerous to throw away old
methodologies, just because they are old
• But:
• education is too important to be merely
fashion oriented!