APOT : automatic path optimizing for traits

Abstract

This thesis considers the design of automated tutoring systems that customize teaching material to accommodate individual student learning styles. In particular, we consider the following problem: Begin with one or more presentations of a subject, and break them into fragments ("atoms") each expressing a single idea. Given information about an individual student's learning style, how can one select the optimal choice and sequence of atoms ("path of atoms") to create the most effective presentation for that student? We have implemented several algorithms that automatically create such paths, and we investigate the tradeoff between number of constraints imposed by the algorithms and the number of paths they can find. We have tested one of these algorithms ("partition search") in an experiment where student volunteers in computer science studied material about planning and artificial intelligence. The results of the experiment indicate that the algorithms can produce presentations that are effectively tailored to the different learning styles.