The current focus of research in this domain is to provide computer-based teaching systems with "intelligence". Above all this means to produce a realistic model of the teacher-student situation in teaching processes.
A crucial point when generating a proper model for the learning situation
is to produce a realistic representation of a human teacher's abilities.
The classical architecture of intelligent tutoring systems (ITS) consists
of four modules:
The teacher's knowledge in the domain is represented in the
expert module.
This knowledge base has to contain all the relevant information about the
domain of knowledge.
But correct information alone doesn't suffice, one needs also knowledge
about typical mistakes and misunderstandings of students to assess the
degree of misconception and to be able to show the student a way to
improve his understanding.
When learning the synchronization with semaphores testing and debugging are
also important aspects.
Some proposals for integrating these into the tutoring system have been
developed, too.
The didactic knowledge of a human teacher is represented within the
tutor module.
This component of an intelligent tutoring system has to contain fundamental
strategies to teach the content of the lesson, e.g. the way of presenting
and choosing proper examples or the appropriate timing of help.
Timing and kind of interaction with the student (tutoring in the small)
in SYPROS during solving
a given task depend on the state of knowledge and the motivational
characteristics of the student.
According to these attributes a proper teaching strategy is chosen.
The further development of a component %structuring the teaching goals
for instructional planning (tutoring in the large) is done in cooperation
with the FU Berlin.
The teacher's knowledge about the student's knowledge state
and his typical method of working, is represented in the student module of
an ITS.
This component stores information about the extent to which the student masters
the domain and about his characteristics of working
(learner-type of the student or his preferences).
This information provides the fundamentals for the decisions the tutor module
has to make during the learning and teaching process.
SYPROS administrates this information in a working model of the learner
during the work on the current task and a long-term model containing
the information of previously handled problems.
Finally an intelligent tutoring system is required to present its knowledge
in a comprehensible way. This is the part of the communication module.
The use of graphical presentation, especially in animations, can provide
illustration, perhaps even better than in conventional instruction.
SYPROS offers windows for editing the processes and simulating runs
of the whole program besides the animation of specific runs.
Additionally the system has an adaptive hypertext component to review
the theory of this knowledge domain.
To improve the coordination of the means of help provided by the system
and to adapt them to the dialogue, it is planned to further
the integration of the four modules in SYPROS.
To achieve this, a representation of knowledge about the correlations
within an ITS and about the dialogue with the student should be provided.
To examine the concepts, developed for SYPROS, in other domains
and to find out, which possibilities different domains may yield,
the intelligent tutoring system POINTRA is developed simultaneously.
POINTRA is an ITS for programming with pointers in Pascal.
Besides the teaching of correct programming techniques, the style
of programming can be a topic of discussion due to the representation
of program variants in the goal-plan-tree.
The intelligent tutoring system SYPROS, which is developed at the moment
at this chair, has the knowledge domain of "SYnchronisation of
parallel PROcesses with Semaphores".
The semaphores, a fundamental concept for synchronization introduced by
Dijkstra, are used in SYPROS for cooperating processes.
The expert module of SYPROS consists of a knowledge base of goals and plans
from which a goal-plan-tree can be generated for every task.
This tree describes a correct way to solve the problem as well as frequently
occurring mistakes in this task.
The student's solution can be diagnosed with the help of the goal-plan-tree,
i.e. it represents the current state of the problem's solution at any given
moment.
Moreover the tree provides information for different mechanisms of help
for the student, for example the correct solution, explanations thereof,
explanations of errors, a testing facility for partial solutions (hypothesis
test) and
a generator of counterexamples disproving the current solution (if the solution
is not yet correct).
On the other hand there is the possibility of a second analysis phase
for tasks and solutions that cannot be handled properly with the knowledge
base.
Mistakes found by this analysis can be shown in an example, but an
explanation thereof is not possible.
To the InTuSys project
To the research group Informatics I
Meike Gonschorek, 1997-08-05 Christian Herzog, 1997-10-24