Bridging Gaps in Models of Expertise

This paper presents an approach to the explicit integration of uncertain reasoning mechanisms into conventional KADS-based models of expertise. We assume that uncertainty is present in the model of expertise in the sense that terminological knowledge is available but some assertions cannot be determined due to uncertainty. These gaps can be bridged in a three-step inference scheme using a general model of uncertainty in which knowledge is represented as sets of hypotheses. Our work is based on Shenoy's valuation-based systems and on Pearl's structural equation model of causality. Combining these two approaches we obtain a model of uncertainty including a scheme for uncertain inferences. The resulting structure can be used to integrate the specification of uncertainty into conventional models of expertise. The inference scheme presented in this paper can be seen as a first step towards the implementation of uncertainty reasoning specified in a model of expertise using gr..