Utility Functions for Multiattributed Consequences

Abstract

One of the important methodological problems concerning the application of decision analysis to complex problems is the restrictiveness of existing techniques for systematically assessing multiattributed utility functions valid for decision making under uncertainty. This is the problem addressed in this paper. Operational assumptions are postulated about the decision maker's preferences for multiattributed consequences, and functional forms of utility functions satisfying these assumptions are derived. The results are a number of representation theorems which simplify the assessment of the utility function provided the requisite assumptions hold. A special case of these results is the well-known additive utility function. A procedure to verify the appropriateness of the necessary assumptions is included.