Validation of expert systems : two perspectives
journal contributionposted on 01.06.2001, 00:00 by Eswaran Subrahmanian, Joseph G. Davis, Carnegie Mellon University.Engineering Design Research Center.
Abstract: "Expert systems have generated a lot of interest in both academia and industry. The development of tools, techniques, and methodologies drawn from from artificial intelligence (AI) for building expert systems have not been matched by proposals for evaluating and assesing them. This paper examines the assessment problem in the context of two distinct perspectives of expert systems in the spectrum of possibilities a) expert systems as psychological theories of human problem solving, and b) expert systems as decision aids. We review the assessment methods/approaches for each of the two categories, with the objective of ensuring that they are consistent with the basic assumptions of each category. Based on the law of requisite variety, for evaluation to be effective, the variety of methods employed have to match the variety to be found in the phenomena to be evaluated."