| Summary: | In this dissertation we have addressed the problem of modeling expertise in domains characterized by unquantifiable, often subjective, information, and using that model of expertise as the foundation for building computer-based decision support systems. The key feature of the expert model is to make explicit the essential characteristics of the knowledge experts use to process objective, quantitative information, for making decisions in environments rich in qualitative data. This model is then used as the basis for an "intelligent" interactive assistant that presents information appropriate for the context to operators who may not have developed the necessary expertise.
The core of the assistant is a heuristic algorithm that reflects what an expert decision maker would actually do. The algorithm incorporates a set of production rules, i.e., if-then-else rules, to define relevance conditions of quantitative data. These rules employ a dominance principle, i.e., a heuristic association of the relevance of quantitative data with the attributes of qualitative data, characterized as a set of ordered values. The heuristic algorithm is embedded in the assistant and is used to assist non-expert operators in locating information useful for making decisions.
The modeling methodology and the heuristic algorithm are applicable for modeling expertise in a class of decision problems characterized by large amounts of qualitative and quantitative data. The process of structuring the expert's knowledge requires empirical evidence from actual decision problems; this evidence feeds the algorithm with heuristic associations between qualitative and quantitative data. The algorithm uses the dominance principle to decide what information to present for a particular set of conditions.
|
|---|
