Building expert systems : a phenomenological approach to accessing expert knowledge

The purpose of this study was to demonstrate a knowledge acquisition technique for developing expert systems. Employee performance problem cases were prepared which represented a wide range of organizations and problem situations that managers could be expected to encounter. An expert manager/trainer was presented these cases and asked to select an appropriate intervention to correct employee problems. Data were collected by an audio-taped interview using the Action-ReasonThematic-Technique (ARTT). ARTT involves a phenomenological interview and analysis procedure designed to identify and describe a person's actions and supporting reasoning structure. An ARTT analysis of the expert's reasoning structure revealed 52 rules used by the expert as he selected intervention strategies. These rules were encoded into an expert system shell to create an operational expert system. These results demonstrated that (1) the ARTT successfully described the expert's actions and reasoning structure; (2) the analysis procedure identified and described rules that could explain the expert's problem solving process; and (3) a phenomonological analysis of an expert's account of a problem solving process can successfully produce a knowledge base for an expert system. Implications of the study include: (1) ARTT should be used in additional studies that would demonstrate its utility as a knowledge acquisition technique; (2) the effectiveness of the expert system developed with the use of ARTT should be tested in experimental conditions to determine the system's effectiveness as a decision aid; and (3) expert systems have potential use as training aids where expert knowledge is scarce or other access to expert knowledge is not cost efficient.