Optimal rules for deciding on instructional routes in intelligent tutoring systems

Intelligent Tutoring Systems (ITSs) can be described as instructional networks with individual routes for the students. The purpose of this paper is to formulate optimal rules for deciding on such routes in ITSs using achievement tests administered to the students and information about their previous history in the system. As an example, three types of elementary test-based decisions (viz. selection, placement and mastery decisions) are combined into an instructional network of an ITS. By simultaneous optimisation of the decisions in this network, rules for guiding the progress of students through ITSs can be designed. The framework for the approach is derived from Bayesian decision theory. Results from an empirical example of instructional decision-making in medicine will be presented to illustrate the differences between a simultaneous and a separate approach.