Robustness of Multiple Objective Decision Analysis Preference Functions

This research investigated value and utility functions in multiobjective decision analysis to examine the relationship between them in a military decision making context. New data is presented and data from an earlier study is also analyzed for this relationship. Further, the impact of these differences on the decision model was examined in order to improve implementation efficiency. Specifically, the robustness of the decision model was examined with respect to the preference functions to reduce the time burden imposed on the decision maker. Data for decision making in a military context supports the distinction between value and utility functions. Relationships between value and utility functions and risk attitudes were found to be complex. Elicitation error was significantly smaller than the difference between value and utility functions. Risk attitudes were generally neither constant across the domain of the evaluation measure nor consistent between evaluation measures. An improved measure of differences between preference functions, the weighted root means square, is introduced and a goodness of fit criterion established. An improved measure of risk attitudes employing utility functions is developed. Response Surface Methodology was applied to improve the efficiency of decision analysis utility model applications through establishing the robustness of decision models to the preference functions. Which value-utility differences are significant was determined. An algorithm was developed and employs this information to provide a hybrid value-utility model that offers increased elicitation efficiency for the decision maker.