Vector-Adjusted Expected Utility

This paper proposes a representation of (possibly) probabilistically unsophisticated preferences whereby (1) beliefs are jointly represented by a finitely additive probability measure and a vector-valued measure; (2) uncertain prospects are ranked according to the difference between a baseline expected utility evaluation and an adjustment term; and (3) the latter is the norm of the vector-valued expected utility of the prospect under consideration.Vector-valued measures are employed to represent the extent to which ambiguity about different events "cancels out" or "adds up", as revealed by the decision maker's preferences. The proposed representation, vector-adjusted expected utility (VEU), is shown to be consistent with the maxmin-expected utility model (MEU). A necessary and sufficient condition characterizing the class of VEU preferences within the MEU family of preferences is provided.