Vladimir A. Goodkovsky

ENIKO MIFI, 3 1 Kashirskoe shosse, Moscow 115409, Russia

good @aka.msk.su

Abstract

Intelligent Tutoring Systems (ITSs) are popular among researchers now. Indeed ITSs promise to be highly effective in learning, but the labor consumption for their creation is very high. Known ITSs are unique, very complex, and expensive, and as a result, do not have wide application. To the contrary, "Intelligent Tutor" (Shell, Toolkit, Technology) has general modular architecture based on fuzzy control theory, contains reusable modules, uses dynamic planning of tutoring dialogue, can use ready-made tutoring materials, and supports easy development and maintenance of inexpensive applications in many different disciplines. They may become popular among users and useful for researchers.

Introduction

Intelligent Tutoring Systems (ITSs) are popular among researchers now. Indeed ITSs promise to be highly effective in learning. But the labor consumption for their creation is very high now. Each new research starts from a scratch, without a foundation on which to build. ITSs are designed under assumptions of plentiful computing resources as complex, integrated applications of Artificial Intelligence. As a result, known ITSs are unique, very complex, expensive, and do not have wide application.

We suggest simple and inexpensive type of ITS based on wide-spread IBM compatible PC. "Intelligent Tutor" has general modular architecture founded on fuzzy control theory. It includes Shell, Too/kit, application modules, and manuals in Technology of application modules development. The Shell consists of reusable modules and may be filled in by ready-made tutoring materials to create new application module. The Toolkit is used to make easy an application modules development in accordance with proposed Technology.

Applied ITSs (or application modules) developed on this basis are able to realize dynamic planning of tutoring dialogue and to tutor in adaptive and mixed initiative way. They let students to choose (and change at any step) learning objectives within theme under study and any mode of System functioning.

This System contains all the typical modules of other ITSs, but for the sake of simplicity the following key enhancements add:

• the absence of problem solver (expert system) in domain under study,
• use of pre-stored tutoring materials without their automatic generation;
• an extension of a traditional "overlay" Student Model with Knowledge Delivery Model, Knowledge Genesis Model, and Knowledge Manifestation Model.

These additional models are declarative and easy to build up within a concrete theme by experts   They set experts  free  from  a  traditional   very  difficult  and   error-prone   design   of procedures, or plans, or rules of ITSs functions. It becomes possible because this System can dynamic plan and execute it automatically.

At that time several applied ITSs have been designed in different domains They begin to be used in practice now.

We intend to demonstrate the technological complex at the conference.

1. Applied ITSs

An applied ITS (or an application module) can solve the human tutoring problems in the individualized dialogue with student automatically. They let students actively participate in their own learning by choosing not only the course and the theme for studying, but also current objectives for achievement within the theme, and mode of ITS operation. Within each theme context under study applied ITS offers to a student the didactic full spectrum of:

• External Modes:
  • Pretest (If it has been prescribed by an instructor);
  • Posttest (If it has been prescribed by an instructor);
• Internal Modes:
  • Basic modes:
    • Knowledge delivery;
    • Knowledge checking;
    • Bugs diagnosing;
    • Explanations;
  • Complex modes:
    • DEBUGGING:
      • Bugs diagnosing + Explanations;
    • COACHING:
      • Knowledge checking + DEBUGGING;
    • TUTORING:
      • Knowledge delivery + COACHING.

After the student makes his/her choice the system works automatically, but on any step of learning a student is allowed to change mode and/or current objectives. Thus the applied tutoring/learning strategy is mixed-initiative. All tutoring materials given to the student are pre-stored, not generated by the system. As a feedback a student may give his or her replay in a form of: multiple choice, set ordering, filling in the blank, number, simple text/digit expression, construction (scheme) of predetermined elements.

The learning of any theme begins from pre-test (when system has no information about student's initial knowledge needed for the given theme). Further learning is done in any mode, dynamically chosen by a student in any order. To finish studying a theme and to get the resulting evaluation a student must pass a post-test. But he/she may begin with the post-test, if he/she is confident in his/her own knowledge.

2. Shell

All applied ITSs are based on the Shell. This Shell is the "engine" of the ITS, that is not filled in with concrete tutoring contents of the domain under study.

The Shell contains universal didactic knowledge. It may be said that the Shell itself knows "how to teach" students and needs only to know "whom to teach", "what to teach", and "which materials to use" in order to become the applied ITS.

This Shell contains:

• some empty declarative models (domain model, expert model, student model) of tutor's mental representations;
• procedural models of tutor's activity;
• criteria and constraints of tutoring.

The DOMAIN MODEL (DM) is a list (a set) of domain concepts and relations studied within a theme.

The EXPERT MODEL (EM) is a list of tasks and/or questions and possible final ("correct") results of each task and/or answer to each question together with domain concepts used by the expert during solving, fulfilling, answering processes.

The STUDENT MODEL (SM) has a multiple form and consists of 5 sub-models:

• DKM - Domain Knowledge Model;
• KDM - Knowledge Delivery Model,
• KGM - Knowledge Genesis Model,
• KMM - Knowledge Manifestation Model,
• PTM - Personal Traits Model.

DKM represents domain concepts mastered by a student. Strictly speaking it is a fuzzy subset of the DM expressing the possibility of domain concepts mastering by a student. The DKM is widely known as Overlay Student Model.

KDM is a fuzzy relation between domain concepts and pre-stored teaching materials that tend to form these concepts. This relation shows & possibility (or the probable ability) of each teaching material forming some domain concepts.

KGM is a fuzzy relation among domain concepts, which shows the necessity of all other concepts for learning/teaching each separate concept.

KMM (in its simplest form) is a declarative Expert Model. In this form, KMM is a fuzzy relation between domain concepts and pre-stored known results of the tasks or the answers to the questions received from an ideal (expert) student within a theme. This relation shows the necessity of each concept for getting the given answer or result by ideal (expert) student. In a more complex, form KMM includes not only correct (expert based), but also possible incorrect and mixed student reactions. In this case KMM additionally shows that the cause (bug) of each incorrect reaction (error) may be possibly the non-mastering of some domain concepts by a student.

PTM is a general fuzzy description of some cognitive, psychological characteristics, and preferences of the student.

Using filled-in declarative models, procedures, criteria, constraints, and also some adjustable parameters the Shell can automatically solve:

• problems of:
  • knowledge prediction;
  • knowledge checking;
  • knowledge diagnosing;
• problems of collection of sufficient sets of:
  • teaching materials for knowledge delivery,
  • tasks/questions for knowledge checking;
  • tasks/questions for bugs diagnosing,
• problems of the reasonable choice of the consequent:
  • teaching material during knowledge delivery,
  • task/question during knowledge checking,
  • task/question during knowledge diagnosing.

All algorithms of applied ITS operation are built on the basis of these problems solving (strict mathematical) methods. The most general mode of the ITS operation is TUTORING. All other modes of Intelligent Tutor operation are parts of TUTORING. Its algorithm includes:

• the cycle of active knowledge delivery with passive prediction, checking and diagnosing of student knowledge,
• the cycle of active checking with a passive knowledge delivery (commenting), prediction and diagnosing of student knowledge,
• the debugging cycle (in case of errors detection):
  • the cycle of active diagnosing with a passive knowledge delivery, prediction and checking of student knowledge;
  • the cycle of explanation and checking of its effect.

A concrete style of this operation is defined by adjusting the following main parameters: Level of Teaching Sufficiency, Level of Checking Sufficiency, Level of Diagnosing Sufficiency, Degree of Checking Delay, Degree of Error Tolerance.

3. Toolkit

ITS Toolkit is used by subject matter experts to fill in the ITS Shell with defined tutoring contents in order to set the design of particular applied ITS. The Toolkit provides experts with the hierarchical mam menu. The base level of the mam menu gives experts the possibility to describe domain concepts of the theme (this is the internal theme representation) The top level of menu gives the possibility of forming external presentations of the theme materials for students. The middle level gives the possibility of connecting base and top levels of the theme description by the above described declarative models.

The Toolkit includes:

• ordinary text, fonts, icon, and graphic editors to form external presentation of all theme materials for students,
• a special Editor to fill in the models used and verify them,
• a special Tool to input values of parameters to define the wishing style of tutoring,
• a Debugger of designed ITS for its verification, for diagnosing and correction of its bugs, for adjusting applied ITS.

4. Technology

The full-scale technology of the applied ITS design consists of the following steps:

• An analysis of predicted professional activity of students and a resultant definition of professional tasks, problems and corresponding needed knowledge. (This stage may be omitted when contents of tutoring has been pre-determined).
• A structuring of learning contents (needed knowledge), a determination of courses and themes within each course. A definition of domain concepts within each theme of the course Preparing of an explanation for each concept under study. Data input, KGM build up, an automatic verification and debugging of KGM.
• A collection of ready-made teaching materials and/or design of new teaching materials for each theme Data input, KDM build up, an automatic verification and debugging of KDM. (This stage may be omitted if expert decides that one level delivery of knowledge in the form of early prepared explanations will be enough).
• A collection of ready-made tasks and questions and/or design of new tasks and questions within a theme (on the base of the predetermined set of professional tasks and problems). Data input, KMM build up, an automatic verification and debugging of KMM.
• An assembling, debugging and adjusting of produced applied ITS.

The work of experts in suggested technology is deeply structured. It is very simple and well understood in each step by the experts, supported with special guides and automatic tools. It does not take much intellectual efforts, therefore it does not cause subjective errors and gives good results in practical applications.

5. Practical results

On the basis of the technology described a number of applied ITSs have been designed. The themes have been taken from the following domains: mathematics, physics, chemistry, geometry, book-keeping, strength of materials, automated mechanisms, chemistry practice of NPP, electrical practice of NPP and others. These last, more advanced commercial applied ITSs have been designed for the rigors of training Nuclear Power Plant personnel.

Our experience shows that experts easily master this technology and use it successfully. All experts appreciated the quality of designed ITS functioning with student. As for students, they were glad to have simultaneous high degree of learning freedom and a wide spectrum of useful tutoring services.

Conclusion

The suggested Shell, Toolkit and Design Technology enables experts to reduce an enormously sophisticated intellectual process of applied ITS design to a more simple one, actually through filling in model forms with ready-rnade tutoring materials on the domain under study. This considerably reduces the labor consumption and makes applied ITS accurate and inexpensive.

As for students, they receive an "Intelligent Tutor" for one-on-one work. The Intelligent Tutor "understands" how the students learn, automatically finds and corrects the causes of errors made by student, adapts to individuals, and allows the student to be the driver. It increases the motivation to learn and learning effectiveness.

References

Goodkovsky,  V.A.   (1988).   An automated diagnostics of student knowledge:  theory and application. Thesis. Moscow, Badman Moscow State Technical University. Russia, 295p.

Goodkovsky, V.A. & Kazennov, A.Y. (1992). Intelligent Tutoring Systems: Theory, Technology and   Practice.   In P.Brusilovsky, V.Stefanuk   (Eds.).   Proceedings   of the   EAST-WEST conference on emerging computer technologies in education (pp. 123-127). Moscow: ICSTI.

Goodkovsky,  V.A. (1993). Intelligent Tutoring Systems. From theory to practice.  In P. Brusilovsky, V. Stefanuk (Eds.) Proceedings of the East-West conference on Artificial Intelligence (pp.

305-309). Moscow, Russia. Goodkovsky,   V.A. (1993). Practical Knowledge Diagnostics. Theoretical Systems Approach. In V.Petrushm, A.Dovgyallo (Eds.). Proceedings of the International Conference on Computer Technologies in Education, (pp. 141-143). Kiev, Ukraine.

Goodkovsky, V.A. (1994). Intelligent Tutoring System: Theoretical System Approach. In Haruki Ueno, Vadim L.Stefanuk (Eds.) Proceedings of Japan - CIS Symposium on Knowledge Based Software Engineering.(pp. 106-109). Pereslavl-Zalesskiy, Russia.

Goodkovsky, V.A., Kirjutm, E.V. & Bulekov A.A. (1994). Shell, Tools and Technology for pop class   ITSs  production.   In P.Brusilovsky, S.Dikareva, J.Greer &V.Petrushin (Eds.). Proceedings of East-West International Conference on Computer Technology in Education. Part 1. pp.87-92.