Modelling expertise for educational purposes

Abstract

This paper describes the knowledge acquisition phase for an Intelligent Computer Coach for physiotherapeutic diagnosis, FysioDisc. Physiotherapeutic diagnostic expertise, like many forms of human expertise, is associative and heuristic in nature, but therefore hard to teach. Students may learn the “rules” and “shortcuts”, but this seems inefficient and tedious, and moreover will leave them helpless when confronted with new or strange cases. This is why we tried to find a more systematic approach, and together with domain experts performed a rational reconstruction of the observed expert behaviour to come up with a prescriptive model of physiotherapeutic diagnosis. Such a reconstruction is time consuming and takes a lot of effort, but the resulting model is more explicit, easier to maintain, control and explain. On the other hand, such a model is more error prone, but the domain experts appear to be good “debuggers” of reasoning models they do not (longer) use.