Hypothetical Reasoning in Causal Models

Abstract

In this article we propose a logical formalization of the reasoning mechanisms to be adopted for diagnosing multiple faults in a physical (or physiological) system represented by means of causal models. After a brief discussion of the similarities and differences with respect to ontological models, the article presents in detail a formalism to represent causal knowledge and, in particular, an approach to deal with incomplete knowledge. We propose a logical semantic for the formalism and we introduce a precise definition of the concepts of "diagnostic problem" and of "definite and plausible solution to a diagnostic problem." A particular form of hypothetical reasoning is presented in order to deal with incomplete models: the observation (or nonobservation) of findings is used to confirm (reject) the hypothetical assumptions introduced during the diagnostic process. The correspondence between the confirmation criterion we have defined and the circumscription principle is discussed.