Reference: Clancey, W. J. Knowledge Engineering Methodology: An Annotated Bibliography of NEOMYCIN Research. Springer-Verlag, 1988.
Abstract: From a broad perspective, knowledge engineering is a methodology for acquiring, representing, and using qualitative models of systems. We distinguish between systems being modeled (physical, cognitive, social, ect.), modeling tasks (such as diagnosis and control), computational methods (such as heuristic classification), and implementation languages (such as rules and frames). The pragmatic value of this perspective is illustrated by uncovering knowledge representation problems in existing expert systems. New languages make explicit the dimensions of task system model, computational method, and implementation. In state-of-the-art expert system shells, the representation of reasoning strategy is emphasized, illustrated here with examples of enhanced explanation, student modeling, and knowledge acquisition. Beyond this, we consider philosophical limitations of the representational approach and implications for future research.