Performance evaluation of a knowledge-based system

To demonstrate the usefulness of a system, it is important to test it within the boundaries of its limitations. In the case of expert sytems, the premise is that the goal of the technology is to enable a non-expert to complete the steps of solving a problem in a way similar to that of an expert. If computer-based systems are to take the lead in problem solving, they must inevitably be subjected to appropriate and adequate validation. In this paper, the authors present results from a validation study conducted on a knowledge-based system called ICE (Information Center Expert). The ICE system was developed to determine the software requirements of end-users and make appropriate recommendations. It incorporates the expertise of consultants from IBM/Endicott, IBM Tucson and the Center for the Management of Information (CMI) at the University of Arizona. The results reported here are based on the validation exercise conducted for the ICE implementation at the University of Arizona.