Representing Arguments as Background Knowledge for the Justification of Case-Based Inferences

Reference: P. Clark. Representing arguments as background knowledge for the justification of case-based inferences. In E. L. Rissland and J. A. King, editors, Proc. AAAI-88 Workshop on Case-Based Reasoning, pages 24-29. August 1988.

Abstract: This paper examines the representation of background knowledge and its use in case-based reasoning. Case-based reasoning can be viewed as a particular form of problem-solving, based on the assessment of similarity of a new case to previously encountered cases, and the subsequent inference that an old solution also applies to the new case. To justify such inferences, we present a representation of background knowledge as arguments for and against a conclusion given known facts, rather than as statements in logic or probabilistic relations. Cases are characterized by the set of arguments for and against a hypothesis of interest, and the resolution of conflicting arguments for a new case is obtained by firstly locating an old case where the same or a similar conflict occurred, then secondly transferring the resolution from the old case to the new case. In this way, we are able both to represent weak, possibly conflicting fragments of background knowledge and also learn about the relative strengths of such arguments from cases where the outcome of conflicting arguments is known. We provide a description of this method in logical form, and analyze the assumptions under which it is valid, its limitations and possible future extensions.