This paper describes a model of efficiency in diagnostic problem solving. It is based on the protocols of three management consultants, who were asked to diagnose a performance problem given to them in the form of a tab-item. They were specifically required to test the maximum number of hypotheses with the minimum number of questions. A bounded rationality model of their judgment was developed, incorporating features of earlier knowledge-based and rule-based approaches. The nature of the diagnostic process is represented as an alternation between knowledge-driven operations on a mental model of the situation, and sensory-driven operations on the problem givens. Insights gained in the former type of episode act as an executive program to direct the search carried out in the latter type of episode. Data generated in the latter type of episode are assimilated into the problem representation constructed in the former type of episode and, by activating existing knowledge structures, generate fresh insights. These in turn direct further search, and so on. Efficiency is achieved both through intuitive leaps which occur in the knowledge-driven episodes, and by a fixed sequence of logical operations in the sensory-driven episodes. The implications of this model for the education of diagnosticians are discussed.
Read full abstract