Should rule-based reasoning be enhanced by case-based reasoning for conceptual database design? A theory and an experiment

Traditional knowledge-based systems, such as expert systems, are rule-based. Inference rules are used for representing procedural knowledge. Some recent studies have suggested, and some have even provided evidence, that when knowledge-based systems are endowed with combined powers of multiple reasoning paradigms, the resulting systems will be more effective than when a single paradigm is deployed for the same domain of problems. We set out to test this claim within the conceptual database design domain by comparing rule-based paradigm to one that combines rule- and case-based reasoning. Case-based reasoning is an inference paradigm in which a problem is solved via the retrieval of a closely related past case and the adaptation of this past solution to become the solution for the present problem. A depository of past cases represents a rich body of knowledge. We theorized that the combination of both will be more effective than relying on rules alone. To test this theory, we created two systems. The first was based on pure rule-based reasoning (RBR). The second was created by adding case-based reasoning (CBR) to the first. We compared the effectiveness of the two systems. The results were mixed. The two did not differ from each other in their degree of effectiveness. However, subjects' single-choice preference was for the CBR system.