Game Theory and "Social Value" States

Abstract

IN HIS article, Reflections on Game Theory and Nature of Value (Ethics, Vol. LXXII [April, 1962]), Robert Paul Wolff takes position that 'game theory is capable of providing a formal analysis only of those situations in which egoistic are to be maximized. He further believes that the concept of strategy, with its postulation of value-maximizing monads, cannot helpfully be applied to problems of realizing values. It is my contention that Wolff's position involves a misunderstanding about use of strategy -s a criterion of value. As a result, his taper shows that application of game theory to 'social values cannot be denied .or reasons that he gives. Wolff takes social values to refer to those valued states and experience where value in question depends upon existence of some form of interaction. He gives as examples of freedom of speech, love, and economic co-operation. These, he says, belong to that class of quite distinct from egoistic values to which he thinks application of game theory is limited. Game theory, as Wolff explains it in a footnote, is a way of understanding that, when parties are engaged in a problem, each has before him branching sets of alternatives. The branches arise from fact that other party's action in a situation will affect one's choices of action at various stages in relationship. As Wolff goes on to point out, aim of each party is to maximize value from his point of view. What stands as a good choice in such situations, as game theory makes clear, is choice which contributes strategically to aim of party choosing. Wolff believes then that gametheoretic analysis of rationality in choice situations has serious limitations because it cannot be used to judge choices where a party is not aiming at some advantage for himself, or, in Wolff's terms, where no egoistic aims are involved in choices made. Wolff begins his subject with example of prisoners' dilemma. Two prisoners in separate cells are each told following: you confess, and your partner does not, you will go free and he will get life imprisonment. If you both confess, you will each get fifteen years. If neither of you confess we have enough evidence to send you both to prison for three years on a lesser charge. The prisoners know that they both have been given same ultimatum. The problem, of course, is what constitutes rational choice. Wolff observes that various alternatives at hand, of confessing and of not confessing under conditions of other prisoner doing one or other, might be assigned a numerical evaluation to indicate which choice will maximize value. The game-theoretic criterion of rationality would hold that choice which maximizes value at any stage would be good, or rational choice. He therefore argues that in this example concept of rationality breaks down. It is worth examining evaluation Wolff makes of choices open to prisoners, in order to see how he separates rational choice from maximizing value here. To confess leads to alternatives of