The Economics of Decision Making

Abstract : The design of a decision analysis is itself a complex decision problem. In theory, each aspect of analysis, encoding the probability density functions of state variables, encoding the von Neuman-Morgenstern utility function, and computing profit lotteries is an experiment. The results of the experiments, the data, are used to update the probabilities in the primary decision problem. The economic value of the experiment is the well known value of imperfect information. The drawback to the theoretical approach is that the data are functions. Practical methods for encoding prior distributions over functions do not exist. Therefore, the traditional approach is to parameterize the data. The authors' approach is unique because we show that for an interesting class of decision problems, arbitrary parameterization is not necessary.