This research is about case.based reasoning ("CBR") and the complementary use of CBR with other reasoning paradigms, such as those using rules or models. The specific topic is how and when to interleave CBR and traditional AI rule.based reasoning (e.g., as used in expert systems) and model.based reasoning, in particular in the situation when the concepts used in the rules or model are not fully defined or admit judgmental decisions. In such problems of interpretation a reasoner seeks to determine the meaning of certain terms and rules as they apply to a given fact situation. This is done by consideration of its applications in past cases. One reasons about the similarities and differences with past cases (precedents) in order to determine whether the new case is "in" or "out" of the category of the rule or predicte in question. This investigation will focus on the architecture and control of a mixed paradigm system that will enable CBR and other reasoning processes to complement each other. The research has three high level goals: (1) describe mixed paradigm reasoning on interpretation tasks, such as exemplified by problems of interpretation in statutory law; (2) use the results of (1) to suggest and build architectures and control regimes; (3) perform experiments to explore and evaluate the approaches of (2). Because certain areas of statutory law offer excellent examples for study of such issues, we propose to begin our investigations in an area such as tax law, where we already have experience with CBR techniques. However, our results will have wider application, even to those domains which to date have been the province of purely rule.based approaches (e.g., diagnostic expert systems).
