Use of ISMAUT (Imprecisely Specified MultiAttribute Utility Theory) in Determining the Most-Preferred Package of Alternatives

摘要

Multiattribute decision analysis serves as a useful model of alternative selection under risk. A specialized form of multiattribute decision analysis assumes that there is only one stage and one consequence for each action and uses a utility function that is additively independent. This approach has served as the basis for several computer based, interactive decision aids. ISMAUT (Imprecisely Specified MultiAttribute Utility Theory) is one such aid that permits lowest level utility scores and trade off weights to be described by linear inequalities. The desirable aspect of this knowledge representation inequalities. The desirable aspect of this knowledge representation is that it permits easy interpretation of many natural language statements of preference, the inclusion of pairwise comparisions of the alternatives, a progressive procedure for knowledge acquisition, and the determination of the most preferred alternative. The intent of this reprint is to present preliminary thoughts as to how an aid designed to select a single most-preferred alternative, ISMAUT, may be integrated with ideas from artificial intelligence to solve this more general problem, which we call the packaging problem. After presenting a brief overview of ISMAUT and its capabilities, we present four possible solution approaches to the packaging problem. Future research and application is required to determine the most effective approach for determining the most-preferred subset of alternatives.