Kees de Koning and Bert Bredeweg
In intelligent educational systems, assessment of what the learner is doing is a prerequisite for proper, knowledgeable guidance of the educational process. We propose to use existing techniques from the field of model-based reasoning for this purpose. This paper describes how a modified version of GDE can be exploited in diagnosing a learner’s problem solving behaviour. The problem solving task for the learner is structured prediction of behaviour. We present models of this problem solving knowledge that adhere to the representational requirements of model-based reasoning, and show how GDE-like diagnostic techniques can be employed to determine those reasoning steps that the learner cannot have applied correctly given the observations. Our approach of diagnosing the learner’s problem solving behaviour, rather than his or her misconceptions, induces an educational strategy that focusses on learning from errors and stimulates the learner’s "self-repair" capabilities.