Knowledge Acquisition for Expert Systems: Experience in Leukaemia Diagnosis

Summary EMYCIN was used to develop an expert system for the interpretation of immunological data obtained in the cell surface phenotyping of leukaemia. Access to a recognised expert, and a large quantity of data on the leukaemias, has facilitated a systematic study of knowledge acquisition and knowledge base refinement based on tape recorded commentaries made by the expert. System performance was analysed at six stages in its development, and ways in which it differed from that of the human diagnostician were identified. Among the most suggestive observations were differences in the way that “undiagnosable” patients were treated and a failure of the elicitation technique to reveal structural aspects of the task. The tools and techniques of knowledge engineering are a significant advance, but a better methodology for developing high quality knowledge bases is needed.