Operationalizing approximate multiattribute utility functions for use in practice

Decision analysis provides defensible methods for rank ordering alternatives based on the calculation of expected utility. This technique can be quite time consuming and conceptually difficult for the decision maker as many input parameters are needed to be assessed. The recognition of such difficulties prompted researchers to call for simplifying approximations which retain the integrity of the original problem but relieve the assessment burden on the decision maker. This call has been answered with the recent development by Kirkwood of Taylor series approximations to expected utility, which have never been applied in practice. This paper further addresses the need to simplify decision analysis through the development of a simulation-based tehnique which allows the Kirkwood approximations to be used in practice. The technique generates confidence intervals for the true but unknown certain equivalents which are used to rank order alternatives.