Inferring an Expert's Reasoning by Watching
David C. Wilkins, Bruce G. Buchanan, William J. Clancey · 1984
This aper introduces a system to infer automatic J y the model of an expert medical diagnostician by watching the expert diagnose a patient. Our ap res ondence YI g roach relies heavily on a close coretween the system and a human wit respect to knowledge organization, inference methods and discourse language. The described system is a major component of a fearning by wutching system being created to facilitate automatic ac uisition of new problem solving knowledge an x to refine cognitive models of expert reasoning. In this paper we describe the closed world learning environment that allows an expert to solve a problem and explain his or her reasoning to the system. Each time the expert requests data, a hypothesis generator reduces a set of alternative interpretations of t Y/l e behavior. These are pruned by a hypothesis evaluator currently under construction.