Explanatory mechanisms for intelligent tutoring systems

Abstract

QUE is an exploratory environment for users of rule based intelligent systems. Its original motivation was the question of how to analyze and explain the discrepancies in rule-based intelligent tutoring systems, between “near miss” incorrect responses of a student and the system's knowledge of the “correct” line of reasoning. It is currently under development as a suite of techniques which provide explanation by supporting the exploration of a system's reasoning processes. This paper describes some of the exploratory modes, the underlying mechanisms that support them, and a number of ways in which these modes and mechanisms might be incorporated into intelligent tutoring architectures.