Calibration of expert systems

Expert systems provide automated vehicles for providing advice, and enacting decisions which are commonly associated with human `experts´. This research investigates calibration of expert systems to bring them in line with end user expectations. Calibration is typically involved in initial learning and setup of the knowledge base. Little research exists to provide insight into how much calibration to perform, or effective approaches to calibration. The research develops a short run comparative static analysis of expert systems development which focuses on expertise. The potential user of an expert system will demand a certain level of expertise and the potential provider of an expert system will be able to provide a certain level of expertise, depending on the sophistication of the expert system provided. The level of expertise provided depends, in part, on how well it is calibrated to the needs of the end user. Since calibration may require significant additional expenditures, as well as require a change in the life cycle approach used to develop expert systems, the decision to calibrate may incur significant additional costs in expert systems implementation. The research finds that only in certain situations is calibration justified