Decision making procedure based on Multiattribute Utility theory Using Bayesian inference

In multiattribute utility analysis, it is required for the decision maker to repeatedly answer questions for indifferent points and subjective probabilities to evaluate the tradeoff between attributes. It is noted that it is difficult for a decision maker to accurately answer such questions. To avoid such difficulties, by asking the decision maker to choose one from a pair of certain outcomes that differ only in two attribute values, we develop a method for identifying the multiattribute utility function. Assuming that scaling constants are probability distributions, the probability corresponding to a parameter profile of the stochastic scaling constants is updated according to the idea of Bayesian inference. In our method, we can obtain the best alternative for the decision maker with less preference information, and by adding more preference information, it can increase the credibility that the alternative suggested by the proposed method is truly best for the decision maker. We demonstrate the procedure for finding the best alternative according to the proposed method with a simple numerical example. Furthermore, as an application example of multiattribute utility analysis, we examine an agricultural decision problem.