A Mixture-Set Axiomatization of Conditional Subjective Expected Utility

Abstract

An axiomatization is presented for a Savage-type conditional subjective expected utility model. The axioms consist of extensions of the Herstein-Milnor [11] axioms for measurable utility, a generalization of an averaging condition in Bolker [4], and several structural conditions. The structural conditions are examined in some detail, and examples are given to show what happens to the numerical model when they do not hold. The numerical model expresses the utility of an act (or mixed act), given an event, as a weighted sum of the utilities of the act given events that partition the initial event, the weights being personal probabilities for the partition events conditioned on the initial event. The theory is compared to Savage's theory [18] and to a version of the theory of Luce and Krantz [14] for conditional expected utility. 1. DECISION UNDER UNCERTAINTY THE PREDICAMENT BETWEEN mathematical tractability and situational reality that is characteristic of mathematical models in the behavioral sciences is epitomized in the axiomatizations of subjective expected utility models. These axiomatizations include structural conditions that facilitate the derivation of the desired numerical representations for preference. Unfortunately, actual situations of decision making under uncertainty often fail to exhibit the structural properties that occur in the axiomatizations. Thus there is real concern about the applicability of such models to realistic decision situations. As might be expected, decision theorists have attempted to alleviate this predicament by weakening the structural conditions while maintaining the ability to derive the desired model from the axioms. An early move in this direction was made by Suppes [21] in his alternative to Savage's axiomatization [18]. The more recent axiomatizations of Bolker [3 and 4], based on Jeffrey's decision model [12], and of Pfanzagl [15 and 16] and Luce and Krantz [14], continue this line of research. The present paper is a further effort in this direction. To understand its approach we shall first review briefly some other theories. The formulation of the paper is set in the context of Savage's states-of-the-world approach to decision under uncertainty, and I shall therefore focus the discussion within this context. We suppose that the decision maker is to select an alternative, or act, from a set of acts. The consequence of his decision will depend not only on the selected act but also on which state in a set of exclusive and exhaustive states of the world obtains. The state that obtains is not known beforehand by the decision maker and does not depend on the selected act.3