Case memory and the behaviouristic model of concepts and cognition

Abstract

Since the advent of case-based reasoning (CBR) in the early eighties, two schools seem to have emerged: One that is more concerned with the cognitive aspects of CBR, and another, arguably more applied school, which views CBR as an AI ‘workhorse’, as it were, that can provide solutions to real-world problems. The former school is sometimes associated with the US, and the latter with Europe. This work, to some extent, attempts to reconcile the two camps. It proposes the behaviouristic or cognitive (as opposed to metaphysical) concept notion as conceptual framework for CBR. It does so by reviewing the three most intensely researched concept models, and relating them to the CBR paradigm. Two of the more recently evolved concept views—the probabilistic exemplar and prototype view—are then put forward as epistemologically sound foundation to model a memory of cases. Extending this core model by means of a complementary possibilistic dimension, a rich and flexible case-knowledge representation framework is developed that addresses issues like expressiveness, extendibility, context effects, and uncertainty. This scheme is called PERCEPT.