Abstract
Experts acquire their specialized knowledge not only from explicit information of the sort found in textbooks, but also from their own personal experience. The greater their expertise, the more important their personally constructed knowledge and, consequently, the more their thinking deviates from that of typical practitioners, rendering it difficult for experts to communicate all of their knowledge to students, colleagues, and knowledge engineers. It follows that the development of knowledge-based computer systems should be essentially a constructive modeling process and not simply a matter of “expertise transfer.” Model-based representations of a specific domain (e.g., radiology) facilitate communication among developers, experts, and users of an expert system. Ideally, knowledge elicitation tools should support the entire modeling life cycle, from initial conceptualization to eventual implementation of any knowledge-based system.