A Bayesian Approach to Diagnosing Problems with Prolog Control Flow

The use of uncertain reasoning techniques have recently been advocated as a means of diagnosing errors in terms of both performance and conceptual understanding (Petrushin & Sinitsa 1993; Martin & VanLehn 1993; Tokuda & Fukuda 1993; Sime 1993). In this paper, we describe an investigation of the utility of a simple uncertain reasoning (Bayesian) technique for the diagnosis of novice Prolog programmer’s misconceptions aboutProlog control flow. We find that, while the use of more sophisticated techniques may mitigate some of the difficulties, there are certain fundamental problems connected with the nature of misconceptions that need to be overcome before such techniques can be used with confidence.