A Programming by Demonstration Authoring Tool for Model-Tracing Tutors

Abstract

Model-tracing tutors have consistently been among the most effective class of intelligent learning environments. Across a number of empirical studies, these tutors have shown students can learn the tutored domain better or in a shorter amount of time than traditionally taught students (Anderson, Boyle, Corbett, & Lewis, 1990). Unfortunately, the creation of these tutors, particularly the production system component, is a time-intensive task, requiring knowledge that lies outside the tutored domain. This outside knowledge-knowledge of programming and cognitive science-prohibits domain experts from being able to construct effective, model-tracing tutors for their domain of expertise. This paper reports on a system, referred to as Demonstr8 (and pronounced "demonstrate"), which attempts to reduce the outside knowledge required to construct a model-tracing tutor, within the domain of arithmetic. By utilizing programming by demonstration techniques (Cypher, 1993; Myers, McDaniel, & Kosbie, 1993) coupled with a mechanism for abstracting the underlying productions (the procedures to be used by the tutor and learned by the student), the author can interact with the interface the student will use, and the productions will be inferred by the system. In such a way, a domain expert can create in a short time a model-tracing tutor with the full capabilities implied by such a tutor-a production system that monitors the student's progress at each step in solving the problem and gives feedback when requested or necessary, in either an immediate or delayed manner.