1 Gauthier, David, Morals By Agreement (Oxford University Press, 1986). Hereinafter referred to as MA.Google Scholar

2 MA, p.9.

3 MA, p.158.

4 MA, pp.166–167.

5 What counts as a fair cooperative scheme is the subject of Chapter V of MA.

6 See MA p.170.

7 Thus, for example, while both an SM and a CM can be expected to employ tit-for-tat in iterated PD games, a CM will, while an SM will not, cooperate on the last round of a finite nteraction, if he expects that the other player is a CM.

8 See MA, p.183. Such a conclusion, it should be noted, is consistent with his remark that the “constrained maximizer does not reason more effectively about how to maximize her utility” (MA, p. 170). A CM is not simply an SM in his most effective guise. Nonetheless, the disposition expressed by a CM is to be defended on the grounds that, within the class of cases to which it is to apply, it is a more effective way to maximize expected utility. That is, from the ex ante vantage point, where the agent deliberates whether to be a CM or an SM, the option of becoming a CM has associated with it a greater expected utility. The essential point is presumably that one's disposition to choose affects the likelihood of being in situations in which mutual advantages can be secured. A straightforward maximizer must expect to be excluded from cooperative arrangements which he would find advantageous. A constrained maximizer may expect to be included in such arrangements. He benefits from his disposition, not in the choices he makes but in his oppportunities to choose.

9 Notice here that no mention is made of the outcome that would result were the individual in question to act on the cooperative strategy while the other player acts on an individual strategy. That is, the analysis abstracts from the possibility that the deliberating agent mistakenly thinks the other player will be cooperative. I shall return to this point shortly.

10 MA, pp.171–172.

11 MA, p.172.

12 MA, p.172.

13 Implicitly, then, this is also a dominance argument.

14 MA, pp.173–174.

15 MA, pp.174–177.

16 Gauthier explores this problem at great length in the book. See in particular, MA, pp.174–187.

17 See his “Deception and Reasons To Be Moral”, American Philosophical Quarterly (forthcoming).

18 To be sure, there will clearly be situations in which it will pay to be disposed to be cooperative. If being cooperative towards others engenders a cooperative response from others, it may well be that, in straightforward utility-maximizing terms, the agent's best strategy is to be cooperative. Imagine, for example, a version of the standard PD game in which each agent is permitted to revise his strategy choice in the light of information about how the other player has chosen – with reconsideration terminating only when each player has “stood pat” with his last announced choice, and where these announced choices are then executed by a referee. Under those conditions, the rational solution will be for each to cooperate – since the rules in question preclude unilateral defection. But in such cases, it can be argued, a CM is simply an SM in a more effective guise. That is, what a CM policy calls upon the agent to choose is exactly what an SM policy calls upon him to choose.

19 To be sure, it is also true that ex ante he would most prefer that the other player trust him while he defects from the cooperative arrangement. But the assumption here is that this possibility is not open to him: under conditions of transparency, he cannot hope that the other player will act cooperatively when he himself is known to be an SM.

20 What supports the assumption that u¹¹ is greater than u¹, that the agent prefers the outcome of unilateral defection over the outcome of mutual cooperation? On the traditional way of thinking about this sort of problem, the notion is that it is just such an order of preferences, and hence utility values, that generates the problem to be analyzed. The most familiar version of this type of problem is, of course, the standard PD game. The agent is presumed to prefer fewer years in jail to more. Those who prefer more years to fewer, or who prefer that both get only a few years to one getting many years in jail while the other goes free, do not face a standard Prisoners' Dilemma.

21 The approach developed here is treated in much greater detail in my book, Rationality and Dynamic Choice: Foundational Explorations (Cambridge University Press, forthcoming).Google Scholar

22 For an earlier statement of the implications of resolute choosing for such situations, see my paper, “Prisoner's Dilemma and Resolute Choice,” Campbell, F. and Sowden, L., eds., Paradoxes of Rationality and Cooperation: Prisoner's Dilemma and Newcomb's Problem (Vancouver: University of British Columbia Press, 1985).Google Scholar

23 Notice also that the criterion of rationality to which appeal is made is a crtierion of individual rationality, not group rationality. The argument is that rational persons who know those with whom they interact to be rational owe it to themselves to choose cooperatively, for by so doing they advance their own interests.

24 Since on the account offered here a rational agent does not ever act contrary to the preferences he has for actions at the time of choice, and, correspondingly, does not choose other than a utility-maximizing action, it is perhaps misleading to describe him as a constrained maximizer. He maximizes in an unconstrained sense his preferences for available actions, although not his preferences over (separably considered) outcomes.

25 Sen, Amartya K., “Behavior and the Concept of Preference,” Economica, vol. 40 (1973), pp.241–259.CrossRefGoogle Scholar

26 The qualifier “rational” is revealingly necessary. The phenomenon of contextually sensitive preferences that appear to qualify as irrational is well-documented. See, for example, Tversky, A. and Kahneman, D., “The Framing of Decisions and the Psychology of Choice,” Science, vol. 211 (1981), pp.453–458.CrossRefGoogle Scholar My suggestion, of course, which runs counter to what most have been inclined to argue, is that in some situations having preferences that do in fact vary with the context turns out to be useful and hence rational.

27 The distinctions developed in this Section are explored at much greater length in my book, Rationality and Dynamic Choice: Foundational Explorations.

28 MA, pp.183–184.