| OCR Text |
Show Chapter III. The Utility-Maximizing Model of Rationality: Informal Interpretations 100 that this conception is weak enough to satisfy universality yet strong enough to avoid vacuity. However, a theory of measurement adequate to this conception would first have to solve the problems raised by interpersonal comparisons that are discussed in Chapter IV, Sections 1.1 - 1.4 below. Second, weakening transitivity to acyclicity does not alter the behavioral options actually available to the choosing agent, because the two are logically equivalent; though a demonstration of this must await the apparatus I develop in Volume II, Chapter III.6.2.1. Third, the argument of Chapter IV, Section 1.6 below implies that it is in any case not possible to exclude cyclical preferences by imposing any further familiar normative requirements - neither transitivity, nor irreflexivity, nor independence, nor substitutability, nor continuity - unless these are subordinated to strictly logical constraints on preference orderings for which the canonical notation of decision theory affords no resources. I offer some in Volume II, Chapter III. Finally, a closer look at the money pump in Chapter IV, Section 2.3 shows that utility-maximization does not require excluding cyclical preferences in the first place. So the criticisms of (U) I make in the following pages apply to this more complex variant on it as well. (U) is not the only possible formulation of the principle. Some utility theorists would insist, on Bayesian grounds, that (U) should address the maximization of expected utility. Whether (U) is formulated so as to address the maximization of utility or of expected utility does not, for the most part, affect the substance of my arguments. Unless explicitly stated otherwise, my arguments do not depend on whether a rational agent is assumed to act under conditions of certainty, risk, or uncertainty. Rather, they focus on the concept of a preference ordering that must be presupposed in the assignment of objective as well as subjective probabilities to options. But I shall assume for the sake of argument that the agent has full information and that any probability assignments to outcomes are based on multiple trial repetitions, again unless specifically indicated otherwise. Others would substitute "human beings" for "a rational being," emphasizing the descriptive over the normative and reducing the explanatory scope of (U) in that way. Still others would object to the suggestion of intentionality in (U), on the grounds that action implies intentionality and agents maximize utility whether or not they intend to. I think it is a mistake for a Humean to raise this objection, for reasons explicated in Section 4 below. Yet others would complain that (U) conceals an essential normative dimension, in that people should maximize utility but often do not in fact. I address expected utility theory's conception of the relation between what human beings actually do and what fully rational beings are conceived to do in Chapter IV, Section 4, below; and the issue of intentionality in this chapter's Section 2, below. © Adrian Piper Research Archive Foundation Berlin |
