ABSTRACT
This chapter offers a candid review of a seemingly straight-forward expert system (ES) development effort. It focuses on the acquisition of knowledge from experts outside the knowledge engineering/programming group. The chapter provides description of the development of SOYBUG, an ES for making spray recommendations for pests in soybeans. The spray decision process was subdivided into four major problem areas; identification of the insect pests, quantification of insect population, delineation of yield damaging population thresholds, and choice of pesticide. Although projects in which the knowledge engineer is his own expert provide useful on-the-job training with the software and hardware components of ES development, it obscures the knowledge acquisition process. Conceptually, an ES's knowledge base can be derived from a wide range of sources such as textbooks, case studies, simulation models, and anecdotes. The central difficulty that experts have is their inability to define the procedures or rules that they actually use in reaching their conclusions.
