Intelligent decision systems (decision analysis, artificial intelligence, infertility, gynecology, urology)
Samuel Holtzman · 1985
People are often faced with difficult decisions for which they do not have intuitive answers. Although a professional decision analysis could have great benefits, such analysis is usually far beyond the financial reach of most individuals. A promising alternative to individual analyses is to model the task of a decision analyst, within a specific domain, as an expert system. This dissertation develops a methodology for the design and implementation of intelligent decision systems, defined as expert systems for helping users make decisions by developing the intuition necessary for action. The methodology emphasizes the process by which real decisions are formalized, rather than the solution of already formalized problems. To this end, it exploits the generality and flexibility of rule-based architectures as well as the normative power of formal decision methods. An important element of this dissertation is the use of influence diagrams to represent and reason with uncertain knowledge in a manner that is mathematically sound, intuitively appealing, and computationally efficient. Particular emphasis is placed on the goal-directed process by which these diagrams are constructed, and on the use of deterministic functions to construct complex stochastic relations. The methodology described in this dissertation is illustrated by RACHEL, a prototype intelligent decision system for infertile couples seeking medical help.