A Bayesian Tutoring System for Newtonian Mechanics: Can it Adapt to Different Learners?

Abstract

Newtonian mechanics is a core module in technology courses, but is difficult for many students to learn. Computerized tutoring can assist the teachers to provide individualized instruction. This article presents the application of decision theory to develop a tutoring system, iTutor, to select optimal tutoring actions under uncertainty of students' mastery states. The novelties of this research are: 1) the automation of student diagnosis that is made possible when tutoring alternatives and the utilities for different outcomes are incorporated to the Bayesian network; and 2) the ability of the tutoring system to select test items with difficulties that are appropriate for the students. The results from formative evaluation on iTutor indicate that it is adaptive, working in a well-structured knowledge space, and able to use the information gathered from the student's responses to dynamically modify the presentation in clearly defined ways.