Current Projects: Student Modeling in Intelligent Tutoring Systems.

INSTRUCT: An Approach to Student Modeling

SINT - a Symbolic Integration Tutor

Abstract: We present an intelligent tutoring system in the area of symbolic integration. The system is capable of solving problems step-by-step along with the student. SINT monitors the student while solving problems, informs the student of errors and provides individualized help and advice when appropriate. The main focus of the research was on student modeling. The technique developed, referred to as INSTRUCT, builds on two well-known paradigms, reconstructive modeling and model tracing, at the same time avoiding their major pitfalls. The approach is not only incremental but truly interactive, since it involves the student in explicit dialogues about his/her goals. The student model is used to guide the generation of instructional actions, like generation of explanations and new problems.

Proceedings of ITS'96 conference, Montreal, June 1996, Lecture Notes in Computer Science, C. Frasson, G. Gauthier, A. Lesgold (eds), Springer, pp.587-595.

INSTRUCT: Modeling Students by Asking Questions

Abstract: The paper reports an approach to inducing models of procedural skills from observed student performance. The approach, referred to as INSTRUCT, builds on two well-known techniques, reconstructive modeling and model tracing, at the same time avoiding their major pitfalls. INSTRUCT does not require prior empirical knowledge of student errors and is also neutral with respect to pedagogy and reasoning strategies applied by the student. Pedagogical actions and the student model are generated on-line, which allows for dynamic adaptation of instruction, problem generation and immediate feedback on student's errors. Furthermore, the approach is not only incremental but truly interactive, since it involves students in explicit dialogues about their goals and problem-solving decisions. Student behaviour is used as a source of information for user modeling and to compensate for the unreliability of the student model. INSTRUCT uses both implicit information about the steps the student performed or the explanations he or she asked for, and explicit information gained from the student's answers to direct question about his or her goals and operations being performed. Domain knowledge and the user model are used to focus the search on the portion of the problem space the student is likely to traverse while solving the problem at hand. The approach presented is examined in the context of SINT, an ITS for the domain of symbolic integration.

User Modeling and User-Adapted Interaction, Vol. 6, No. 4, pp. 273-302, 1996.

Interactive Reconstructive Student Modeling: a Machine Learning Approach

Reconstructive bug modeling is a well--known approach to student modeling in intelligent tutoring systems, suitable for modeling procedural tasks. Domain knowledge is decomposed into the set of primitive operators and the set of conditions of their applicability. Reconstructive modeling is capable of describing errors that come from irregular application of correct operators. The main obstacle to successfulness of this approach is such decomposition of domain knowledge to primitive operators with very low level of abstraction, so that bugs could never occur within them. The other drawback of this modeling scheme is its efficiency, since it is usually done off--line, due to vast search spaces involved.

This paper reports a novel approach to reconstructive modeling based on machine learning techniques for inducing procedures from traces. The approach overcomes the problems of reconstructive modeling by its interactive nature. It allows on--line model generation by using domain knowledge and knowledge about the student to focus the search on the portion of the problem space the student is likely to traverse while solving the problem. Furthermore, the approach is not only incremental but truly interactive, since it involves the student in explicit dialogs about his goals. In such a way, it is possible to determine whether the student knows the operator he is trying to apply. Pedagogical actions and the student model are generated interchangeably, thus allowing for dynamic adaptation of instruction, problem generation and immediate feedback on student's errors. The approach presented is examined in the context of the SINT system, an ITS for the domain of symbolic integration.

Int. J. Human-Computer Interaction, Vol 7(4), 385-401, 1995.