Pragmatic rationality and rules

Philosophy and Public Affairs

Volumn 26, Issue 3, 1997, Pages 210-258

  McClennen, Edward F ^a  

^a NONE

Author keywords

[No Author keywords available]

Indexed keywords

EID: 34248103202     PISSN: 00483915     EISSN: 10884963     Source Type: Journal    
DOI: 10.1111/j.1088-4963.1997.tb00054.x     Document Type: Article

References (107)

1. Philosophy & Public Affairs 6, no. 4 (summer 1977): 317-44
2. Philosophy & Public Affairs Ibid, pp. 335-36.
3. I adapt the distinction between "guiding" and "goading" from W. D. Falk's landmark article, "Goading and Guiding," Mind, n.s. 62 (1953): 145-71. The economists' unsuccessful attempt to try to make do with sanctions extends (with a similar lack of success) to their treatment of the notion of being "trustworthy," and the related notion of "(good) reputation."
4. Similar problems seem to me to be lurking in Russell Hardin's otherwise very interesting article, "The Street-level Epistemology of Trust," Politics and Society 21 (1993): 505-29. For a most helpful diagnosis of the conceptual problem here, see C. W. Morris, "Qu'appelle-t-on reputation?" in Les limites de la rationalité. Of course, to insist that being rule-guided involves more than behaving in a certain way, and more than being motivated to behave in that way by the threat of sanctions, is to make only a conceptual point. It remains to show - this being the task I have set myself here - that being rule-guided (in the strict sense of that term) can be defended from the perspective of an instrumentalist theory of rationality.
5. Philosophical Review 64 (1955): 3-32.
6. My sense that one should focus on a somewhat broader class of practices is shaped in part by the insightful analysis to be found in B. J. Diggs, "Rules and Utilitarianism," American Philosophical Quarterly 1 (1964): 32-44.
7. For an illuminating discussion of the advantages of rules see G. Brennan and J. Buchanan, The Reason of Rules (Cambridge: Cambridge University Press, 1985).
8. This somewhat oblique formulation of the "public goods" problem is necessary, in order to emphasize what countless conceptual studies have clearly established, namely, that the problem is not one that arises just for "selfish" individuals. See, for example, D. Parfit, Reasons and Persons (Oxford: Clarendon Press, 1984), Chap. 4;
9. T. C. Schelling, "Game Theory and the Study of Ethical Systems," Journal of Conflict Resolution 12 (1968): 34-44;
10. and H. Sobel, "The Need for Coercion," in J. R. Pennock and J. W. Chapman, eds., Nomos XIV: Coercion (Chicago: Aldine-Atherton, 1972), pp. 148-77.
11. It would seem that it is for this sort of reason that Rawls, for example, in Political Liberalism (New York: Columbia University Press, 1993), pp. 48-54. takes the position that the "reasonable" cannot be derived from the "rational." What I hope to show here is that such a derivation is possible.
12. See, for example, K. Arrow, "Political and Economic Evaluation of Social Effects and Externalities," in M. Intriligator, ed., Frontiers of Quantitative Economics (Amsterdam: North-Holland, 1971), pp. 3-31;
13. and M. Reder, "The Place of Ethics in the Theory of Production," in M. J. Boskin, ed., Economics and Human Welfare: Essays in Honor of Tibor Scitovsky (New York: Academic Press, 1979), pp. 133-46.
14. For a powerfully argued version of this sort of explanation, see R. H. Frank, Passions within Reason: The Strategic Role of the Emotions (New York: W. W. Norton, 1988).
15. The implausibility of this presupposition cannot be established simply by noting that it has the odd implication that, with regard to human behavior, unintentional processes can achieve what rational, deliberative choice cannot. Following Elster's arguments in Ulysses and the Sirens (Cambridge: Cambridge University Press, 1979)
16. and in Sour Grapes (Cambridge: Cambridge University Press, 1983), it must be acknowledged, for example, that the logic of belief is such that we cannot simply will ourselves to believe a proposition to be true; and it must also be acknowledged that there are some psychological and/or physical states, like being asleep, that cannot be brought about by a direct act of deliberative willpower. On the other hand, we are not required to conclude from these observations that a commitment to practice rules cannot be deliberatively achieved. What Elster's argument turns on, I suggest, is the specious assumption that such a deliberative commitment would, in the typical case, be irrational. What I hope to show is that this assumption is incorrect.
17. Here I must acknowledge my indebtedness to Elster. This is the approach set forth in Ulysses and the Sirens, and in "Weakness of Will and the Free-Rider Problem," Economics and Philosophy 1 (1985): 231-65.
18. In suggesting that there is a link between the problems faced by the isolated decision-maker and those faced by interacting persons I thereby set foot in a territory many important features of which have been carefully mapped by T. Nagel, in The Possibilities of Altruism (Princeton: Princeton University Press, 1970),
19. and by Parfit, in Reasons and Persons. Space considerations preclude my speaking, except all too briefly, to the question of how their analyses differ from my own. Let me simply say that the argument to be unfolded here is not intended as a brief for the view that a convincing defense of prudence (or self-interest) will also lead one to a defense of some impartial form of consequentialism. Moreover, I have, in general, tried to studiously avoid, rather than remark upon, the great philosophical controversy over various psychological and ethical "isms." And while I do hope that I have managed to offer some sort of brief for the rationality of thinking more holistically about oneself, and more holistically about interactions with others, I have tried to content myself with what can be concluded on the most modest of foundations, namely what is advantageous to a given person from each of the relevant temporal points in a series of choices to be made over time, or mutually advantageous to a set of persons who find themselves faced with a problem of interdependent choice.
20. Ulysses' problem, it should be noted, is subject to two different interpretations. On the first, the story of the Sirens and their Song is a metaphor for a situation in which an agent anticipates that his will-power will be literally overwhelmed by some external power (the Sirens and their Song). Faced with this, Ulysses reasonably takes precaution and has himself tied to the mast. Interpreted in this way, the story connects naturally with the problem posed for a rational agent by a physically addictive drug. On a second interpretation, the story is a metaphor for a situation in which an agent projects that his preferences will change over time (with some change in perspective that a temporal shift can imply). In this case, the agent is presumed to deliberate and decide incrementally rather than globally - to reassess options at each choice point in time in terms of interests (short-range or long-range) as they are perceived from that point forward in time, without regard to whatever interests were previously projected with respect to those same choices, and which originally formed the basis for a choice of a plan. In what is to follow, I shall focus exclusively on intrapersonal struggles of the second kind. That is, my concern is with agents who are liable to being disadvantaged by the "Siren's Song" of incremental reasoning, rather than those who expect their deliberative powers to be overwhelmed by external forces.
21. For a sense of the extraordinary range of stories that can be told here, see, for example, R. H. Strotz, "Myopia and Inconsistency in Dynamic Utility Maximization," Review of Economic Studies 23 (1956): 149-58;
22. P. Hammond, "Changing Tastes and Coherent Dynamic Choice," Review of Economic Studies 43 (1976): 159-73,
23. and "Dynamic Restrictions on Metastatic Choice," Economica 44 (1977): 337-50;
24. M. E. Yaari, "Endogenous Changes in Tastes: A Philosophical Discussion," Erkenntnis 11 (1977): 157-96;
25. Elster, Ulysses and the Sirens;
26. E. F. McClennen, Rationality and Dynamic Choice: Foundational Explorations (Cambridge: Cambridge University Press, 1990);
27. and G. Ainslie, Picoeconomics (Cambridge: Cambridge University Press, 1993).
28. For an illuminating discussion of planning in the context of changing information, see M. Bratman, Intention, Plans and Practical Reason (Cambridge, Mass.: Harvard University Press, 1987).
29. Since the confounding is predicated on a change in your preferences, it might seem that some sort of confounding is unavoidable. Either - as in the case just discussed - the later self confounds the earlier self, or - as in the regular version of the story, where Ulysses has himself tied to the mast - the earlier self confounds the later self: Ulysses tied to the mast, and desiring to follow the Sirens, curses his former self. On this reading, if the preferences of one or the other self must be sacrificed, what is the point of labeling only the myopic self dynamically inconsistent? As we shall shortly see, however, neither pre-commitment, nor certain other ways of overcoming myopia, involve any inconsistency in choice.
30. See F. P. Ramsey, "Truth and Probability," in Foundations of Mathematics and Other Logical Essays, ed. R. B. Braithwaite (London: Routledge & Kegan Paul, 1931), pp. 156-98;
31. D. Davidson, J. McKinsey, and P. Suppes, "Outlines of a Formal Theory of Value, I," Philosophy of Science 22 (1955): 60-80;
32. and F. Schick, "Dutch Bookies and Money Pumps," Journal of Philosophy 83 (1986): 112-19.
33. The suggestion that one can adapt the Pareto criteria for social choice to this context is to be found in Yaari, "Endogenous Changes in Tastes." Recall here that the problem with which we are concerned arises by virtue of there being an agent who has different preference orderings at different points in time. From the standpoint of the microeconomist or the game theorist, this amounts to supposing that there are two distinct agents who have to interact with each other.
34. Here is where the literature on Dutch books and money pumps becomes relevant. If you are a myopic chooser, your loss can be some potential exploiter's sure gain, which means, in an entrepreneurial world, you can expect to be exploited. Someone, for example, could come forward not only offering to sell you special diet foods, but also (at a later time) offering to buy them back (at a reduced price) when you fail to consume them! All of this makes for a powerful pragmatic argument against being myopic.
35. In this manner, as already noted, Ulysses' present self confounds his future self. But there is no dynamic inconsistency, since, in this case, there is no choice by his future, tied-to-the-mast self. Alternatively, suppose Ulysses resolves the problem by simply choosing another (somewhat longer) route, one that does not bring him within earshot of the Sirens. In this case, he achieves dynamic consistency by simply avoiding the problematic choice point altogether.
36. The term "consequentialism" has a somewhat different meaning in this setting than in that of moral theory. Here it is the view that rationality for a given agent is a matter of that agent maximizing with respect to his or her own preferences for outcomes. The stress on preferences, and on the standpoint of the individual agent, distinguishes this sense of consequentialism from the one to be found in contemporary moral theory. There it is taken to refer to the doctrine that "the right act in any given situation is the one that will produce the best overall outcome, as judged from an impersonal standpoint which gives equal weight to the interests of everyone" (S. Scheffler, ed., Consequentialism and Its Critics [Oxford: Oxford University Piess, 1988], p. 1).
37. Pertinent discussions of consequentialism within the context of rational choice theory are to be found in P. Hammond, "Consequential Foundations for Expected Utility," Theory and Decision 25 (1988): 25-78;
38. McClennen, Rationality and Dynamic Choice, pp. 144-46
39. in particular; I. Levi, "Consequentialism and Sequential Choice," in M. Bacharach and S. Hurley, Foundations of Decision Theory (Oxford: Basil Blackwell, Ltd., 1991), pp. 92-122;
40. and J. Broome, Weighing Goods (Oxford: Basil Blackwell, Ltd., 1991), pp. 1-16, 123-26.
41. The qualifier "still" is important here, since as you move through the tree, certain opportunities are foregone; paths that were in fact not taken lead to outcomes that are, then, no longer possible.
42. Formally speaking, it would seem that concerns of this sort can be captured within a consequential framework by working with a more permissive notion of what counts as an outcome. Someone who now prefers to make choices that are consistent with choices made earlier can be said to view the path by which he or she reaches a given outcome (in the more ordinary sense of that term) as part of the outcome. See the references listed at the end of footnote 22. Once again, however, what is relevant are the preferences entertained by you now.
43. By the same token, of course, nothing on this account mandates that you have such preferences. That is, it is also possible that you have virtually no, or at best only a very imperfect, commitment to past decisions.
44. That is, at each point in the future you will then stand to what is then yourself in the past in just the same relation that you now stand to yourself in the past. But as just observed in the case in question, you will not conceptualize your deliberative problem in such a way that deliberation could issue in a decision to coordinating with the choices you made earlier.
45. Autarky ordinarily implies not just independence but also self-sufficiency. What will emerge shortly is that you do less well (in terms of continuing interests that you have over time) by exercising this sort of independent choice than you would by seeking to coordinate your choices over time. In this respect, such an autarkic approach to choice over time means that you can realize only an imperfect form of self-sufficiency.
46. If I have not misunderstood D. Davidson's argument in "Deception and Division," J. Elster, ed., The Multiple Self (Cambridge: Cambridge University Press, 1986), pp. 79-92, this appears to be the position that he adopts when he argues that there is no reasoning that extends across the boundaries of the divided self, only causal or power relations. Since I propose to challenge this assumption, it seems clear to me that our views on both rationality and, for example, weakness of the will, significantly diverge. I will come back to briefly discuss one contrast in our views, but I must leave to another paper the task of comparing our respective viewpoints.
47. This is the case upon which Ainslie focuses in Picoeconomics. Again, space considerations preclude my exploring the relation between my own account of dynamic intrapersonal choice and that which is to be found in this most interesting and insightful work.
48. The model that emerges in all these cases, then, can be described as the intrapersonal analogue to the model of "coordination" or "cooperation" that is central to virtually all work in game theory on interpersonal interaction, namely, the model in which the choices of the distinct individuals are in equilibrium. Within the framework of that model, individuals do not negotiate or bargain and thus compromise their respective interests or concerns with a view to reaching an arrangement that, while mutually beneficial, calls upon each to constrain his or her choices. Instead, each maximizes with respect to his or her concerns, given an expectation as to how each other person will choose. Here again, the appropriate model is the one already encountered, in which each (in this case distinct) self proceeds to choose in a thoroughly autarkic manner. I shall return to this point in Section VI.
49. The condition formulated here constitutes a generalization of the one formulated in Rationality and Dynamic Choice. Within the present context, separability cannot be defined in the way that I defined it there. In that work I was concerned only with a very constrained set of decision problems, in which other, entrepreneurial agents are motivated, and are in a position, to exploit an agent who has nonstandard preferences with respect to certain gambles. It is plausible to assume, in that case, that the content of the preferences that such an agent has regarding various gambles allows one to "snip off" earlier stages in a given decision tree, and imagine that the agent is faced simply with a choice between whatever courses of action remain in that "truncated" tree. Separability, then, can be characterized as requiring a match between what an agent would do at any given point in a decision tree, and what he would do were he to face the truncated tree as a de novo decision problem. But if we allow now that the agent can have concern for what has come before, what sense is to be made of the de novo version of a truncated decision problem? To snip off the earlier part of the tree is to disconnect the agent from concerns which I am now prepared to suppose that the agent could have - since in the de novo version of that part of the tree, there is no past self whose concerns (in the case now under consideration) may well be the object of concern of the present self. But the discussion regarding the autarkic manner in which the time-defined self proceeds to deliberate turns out, as I suggest in the text, to provide a suitable way to reformulate and generalize this condition.
50. For a much fuller treatment of the more technical details of this, see Rationality and Dynamic Choice, Chaps. 6 to 8.
51. 1 on remains.
52. Tyranny is exactly what the sophisticated self typically achieves, by the device of precommitment. Ulysses before he hears the Sirens does not respect the preferences of Ulysses after he hears the Sirens, and once he precommits, his later self has no choice but to accept the constraints imposed by his earlier self. His resolve, then, does not unravel; but this is simply because he has tied his hands in advance.
53. One can contrast this case with one in which, say, a person acts on a sudden strong altruistic impulse and gives away his entire fortune, only to rue this too-charitable act. Here it would be appropriate to say that the individual chose in a manner that amounted to tyrannically imposing his preferences of the moment on his subsequent self. See Strotz, "Myopia and Inconsistency," for other examples of this sort.
54. For a more detailed discussion of the relation between the argument pursued here and Elster's analysis, see Rationality and Dynamic Choice, Section 13.7.
55. Here the argument makes contact, of course, with Chapter 7 of Rawls, A Theory of Justice (Cambridge: Harvard University Press, 1971), "Goodness as Rationality," and especially the discussions in Sections 63 and 64 of plans of life and deliberative rationality. Once again space considerations preclude my exploring this connection, but it seems clear to me that his remarks about planning involving scheduling for the sake of effectively and harmoniously satisfying sets of desires (pp. 410-11), and his appeal to a principle of inclusiveness (pp. 412-15), implicitly direct us to what I have here characterized as a "non-separable" approach to rational deliberation.
56. See T. Schelling, The Strategy of Conflict (Cambridge, Mass.: Harvard University Press, 1969).
57. Correspondingly, I think that it is possible to develop a theory of coordination to apply to successive generations of people. What poses a problem in the intergenerational case is not that the agents exist in temporally disjointed periods; rather, it is, once again, the limited conception of coordination that the standard theory of rationality offers us, and specifically the autarkic nature of the deliberation to which the separability principle commits us.
58. For economists who embrace the theory of revealed preference, according to which what you choose defines what you prefer, this view in unavoidable. See Hammond, "Consequential Foundations for Expected Utility," pp. 25-78. When pushed to the limit, as it is in Hammond's account, this doctrine has the decidedly odd implication that no sequence of choices you make could ever reveal a problem of dynamic inconsistency: as myopic, no less than as a sophisticated chooser, you simply execute the plan you prefer (since what you end up choosing reveals what you prefer, and any path through a decision tree defines a set of choices according to some plan)!
59. See Sen, "Rational Fools," pp. 317-44.
60. must confess, however, that in my book, Rationality and Dynamic Choice, pp. 213-15, I defended such a view.
61. See here, in particular, I. Levi, Hard Choices (Cambridge: Cambridge University Press, 1986);
62. "Consequentialism and Sequential Choice";
63. and S. Shapiro, "Rules and Practical Reasoning," Ph.D. dissertation, Columbia University, 1996.
64. See, in particular, J. Raz, Practical Reason and Norms (Princeton: Princeton University Press, 1990), Chapter 1, on exclusionary reasons;
65. H. G. Frankfurt, "Freedom of the Will and the Concept of a Person," Journal of Philosophy 68 (1971): 5-20;
66. and Sen, "Rational Fools."
67. Of course, as discussed earlier, you might be the sort of person who values choosing in a manner that is consistent with earlier choices made. Given preferences of the type in question, however, you need not be resolute in the sense with which I am concerned: ordinary motivations carry you through.
68. Hammond, "Consequentialist Foundations for Expected Utility," p. 28.
69. This is true, not only of Hammond, "Consequentialist Foundations," but also, for example, of K. Arrow. See his "Exposition of the Theory of Choice Under Uncertainty," in C. B. McGuire and Roy Radner, eds., Decision and Organization (Amsterdam: North-Holland Publishing Company, 1972), pp. 23-24.
70. It is also true of Bellman, who, as far as I have been able to determine, is the first to invoke such a principle. See his "The Theory of Dynamic Programming," Bulletin of the American Mathematical Society 60 (1954): 503-15.
71. In such cases, it surely makes more sense to invoke a principle of tolerance and let each theorist nurse his or her own intuitions. On this reading, however, separability has only limited, intersubjective standing, that is, standing only within the circle of the committed.
72. For a classic survey of the problem here, see A. Tversky and D. Kahneman, "The Framing of Decisions and the Psychology of Choice," Science 211 (1981): 453-58.
73. The literature of the last two decades bears testimony to the great number who have, with varying degrees of reluctance, had to embrace this thoroughly odd conclusion. For a sampling, see Yaari, "Endogenous Changes in Tastes";
74. J. Elster, Ulysses and the Sirens;
75. and Parfit, Reasons and Persons.
76. For a fuller discussion of various pragmatic arguments, see my Rationality and Dynamic Choice, especially Chapter 10. For an even more recent discussion of some unresolved issues, see W. Rabinowicz, "To Have One's Cake and Eat It Too: Sequential Choice and Expected-Utility Violations," Journal of Philosophy 92 (1995): 586-620.
77. J.J.C. Smart, "Extreme and Restricted Utilitarianism," in P. Foot, ed., Theories of Ethics (Oxford: Oxford University Press, 1967), pp. 171-83.
78. Bratman, I must note, would disagree. He levels this charge against resolute choice in "Planning and the Stability of Intention," Minds and Machines 1 (1992): 1-16.
79. See, however, the rejoinder by L. De Helian and E. F. McClennen, "Planning and the Stability of Intention: A Comment," Minds and Machines 2 (1993): 319-33.
80. Some will perhaps grant that one may be better off with a disposition to take earlier decisions as providing reasons for current action, but go on to pose the following questions. "Suppose I make my plans subject to the expectation of resoluteness, and then am presented with a pill that would eliminate the disposition. Wouldn't instrumental reason speak in favor of taking the pill? And if it would, doesn't that show that there is a strong pull to separability?" I would argue that instrumental reason does not speak in favor of taking the pill. Agents who have access to such pills and who take them will have no choice, if they are to effectively and advantageously coordinate their choices over time, but to employ second-best methods of precommitment. Those who discipline themselves to not only choose in a resolute fashion, but to refuse the pills, will do better. There is no strong pull to separability unless one has already begged the question of the validity of the separability in its favor. The whole point of the foregoing discussion is to try to find a neutral standpoint from which the rival claims for separable and non-separable approaches can be comparatively assessed. I claim that the more holistic perspective on consequences provides just such a standpoint.
81. With this turn in the argument, I hope to make contact with a point of view that finds powerful expression in Frankfurt's "Freedom of the Will and the Concept of a Person."
82. Historically, of course, this restricted set of games turned out to be pivotal for the whole of the (then) newly emerging theory of games.
83. This is a fixed point of the standard theory of games: if you are rational your choice must be a preference-maximizing response to what, at the moment of choice, you expect the other player to do. There is a huge literature on refinements in, and modifications of, this way of thinking about rational interpersonal choice. What is basic is the concept of an equilibrium of choices, as developed originally by J. F. Nash, in "Non-Cooperative Games," Annals of Mathematics 54 (1951): 286-95.
84. A most useful elementary exposition is to be found in R. D. Luce and H. Raiffa, Games and Decisions (New York: John Wiley & Sons, 1957), Chap. 4.
85. For a sense of the wide range of variations on, and modifications in, this way of thinking, see in particular J. B. Kadane and P. D. Larkey, "Subjective Probability and the Theory of Games," Management Science 28 (1982): 113-20;
86. B. D. Bernheim, "Axiomatic Characterizations of Rational Choice in Strategic Environments," Scandinavian Journal of Economics 88 (1986): 473-88;
87. and W. Harper, "Ratifiability and Refinements (in Two-person Noncooperative Games)," in M. Bacharach and S. Hurley, eds., Foundations of Decision Theory (Oxford: Basil Blackwell, 1991), pp. 263-93.
88. The concept of choices that are in equilibrium is central not only to game theory in general, but also to that part of the theory which focuses upon games of pure coordination and cooperation, starting with the work of Schelling, The Strategy of Conflict,
89. and D. Lewis, Convention: A Philosophical Study (Cambridge, Mass.: Harvard University Press, 1969), and continuing on with virtually all those who have subsequently commented upon the nature of coordination games.
90. have detailed what I find bothersome about this tradition in "The Theory of Rationality for Ideal Games," Philosophical Studies 65 (1992): 193-215,
91. and more recently, and with special reference to the theory of cooperative games, in "An Alternative Model of Rational Cooperation," in M. Salles and J. Weymark, eds., Justice, Political Liberalism, and Utilitarianism: Proceedings of the Caen Conference in Honour of John Harsanyi and John Rawls (forthcoming).
92. Note, however, that it is not a matter of strategic indifference whether you play first rather than simultaneously. The player who goes first, just like one's earlier self, is faced with the task of determining what will maximize her present preferences for outcomes, given that the player who goes second will maximize in an autarkic manner.
93. Once again I have modified the formulation of the relevant separability principle, specifically the one that I employ in "The Theory of Rationality for Ideal Games," so as to leave open the possibility that an agent might just happen (for some nondeliberative reason) to have a preference for coordinating her choice with the other participating agent.
94. It might be objected, of course, that in a game such as a simultaneous-choice prisoners' dilemma you will have a quite distinct reason for choosing the noncooperative strategy, namely, so as to minimize the loss that the other person could impose on you. But this argument cannot be sustained within the context of ideal games played under conditions of common knowledge. Under such conditions, once the separability assumption is replaced by the assumption that rational players will resolutely act so as to secure gains that coordination can make possible, each will expect the other to cooperate, and thus the risk factor is eliminated. It might also be objected that in such games there is a distinct argument for taking mutual noncooperation as the rational outcome, an argument that is based on an appeal to a simple principle of dominance with respect to outcomes. But dominance considerations, no less than equilibrium considerations, carry little weight in contexts where a case can be made for a coordinated approach to choice. I have already had occasion to express my doubts about certain principles of dominance, including Savage's "sure-thing" principle, in Rationality and Dynamic Choice, Chap. 4, but there my concern was with a very strong version of the principle, framed with respect to arbitrary partitions of events.
95. Just how problematic even simple dominance reasoning can be is revealed in N. Howard, The Paradoxes of Rationality (Cambridge, Mass.: MIT Press, 1971). Howard is forced to admit that his reliance on the dominance principle in his own theory of "meta-games" generates a serious paradox of rationality. Following Howard's lead, my own objections reach even to the most simple form of this principle, as a principle appropriate to interdependent choice. The point is very simple, and implicit in what I have already said: those who reason in accordance with dominance considerations in interactive contexts are doomed to end up, as in the classic prisoners' dilemma problem, with second-best outcomes. The consequentialist perspective I have sought to develop here, then, undercuts dominance no less than standard equilibrium considerations.
96. I am suggesting, in effect, that we must march in a direction opposite to the one taken by virtually the entire discipline in recent years, where the stress has increasingly been on noncooperative game theory. But, then, radical measures are needed if game theory is to be rescued from the absurdities generated by the standard theory. What the standard theory offers is a marvelous elaboration of the behavior of "rational fools" (if I may be allowed to borrow that phrase from Sen). In this regard, I have found another of Howard's articles, "A Piagetian Approach to Decision and Game Theory," in C. A. Hooker, J. J. Leach, and E. F. McClennen, eds., Foundations and Applications of Decision Theory (Dordrecht: D. Reidel, 1978), pp. 205-25, most useful.
97. This is, of course, central to the argument in Gauthier's Morals by Agreement (Oxford: Clarendon Press, 1986).
98. have tried to say something along parallel lines in "Justice and the Problem of Stability," Philosophy & Public Affairs 18 (1989): 3-30,
99. and in "Foundational Explorations for a Normative Theory of Political Economy," Constitutional Political Economy 1 (1990): 67-99, although the particular account of fairness that I defend differs significantly from the one that Gauthier offers.
100. For a relatively recent survey of this work, see D. Fudenberg and J. Tirole, Game Theory (Cambridge, Mass.: MIT Press, 1992), Chap. 5, and the many citations therein.
101. For an excellent, and not overly technical survey of some of the more recent work in this area, see R. L. Calvert, "Rational Actors, Equilibrium, and Social Institutions," in J. Knight and I. Sened, Explaining Social Institutions (Ann Arbor: University of Michigan Press, 1995). It should be noted that there is a recurring problem that arises in regard to all these mechanisms, formal or informal, for stabilizing cooperative forms of interaction. Such an equilibrium account requires not only that would-be defectors face the threat of sanctions, but that those who are charged with identifying defectors and carrying out such sanctions be sufficiently motivated to do so. The usual move is to require that there be a second level of sanctions, to be applied to those who fail to carry out first-level sanctions, sufficient to motivate the first-level punishers to do their job. Logically, this requires, in turn, a third level of sanctions, etc., etc. It is unclear, however, just what are the implications of this line of reasoning. If the system of enforcement is "open-ended" in this way, i.e., if there is no "highest" level of enforcement which is itself motivated in some other way, how is conformity at any lower level to be secured? And if it is required that there be such a "highest" level, then the standard game-theoretic account of the stabilization of cooperation must be supplemented in some fashion or other.
102. The question posed by this consideration, to which I alluded at the outset, is whether such a disposition to cooperate voluntarily is conditioned merely on an expectation that others will act in accordance with the rules, or on an expectation that others are similarly rule-guided in their choices. My sense, as I have already indicated, is that the latter is the correct analysis. I am inclined to that conclusion by the consideration that what I have been exploring here is one species of the sort of "joint intentional" activity that Bratman analyzes in "Shared Cooperative Activity," Philosophical Review 102 (1993): 327-41, and also a species of what R. Tuomela has analyzed in a number of papers.
103. For the latter, see, for example, G. Sandu and R. Tuomela, "Joint Action and Group Action Made Precise," Synthese 105 (1996): 319-45. Space considerations preclude my exploring the precise relationship between their views and mine, or offering a fully worked out argument for the position I have taken, but it does seem to me that one should be wary about how the argument presented here could be extended beyond a situation in which there is genuine coordination of actions, that is, a mutual sense of an intentional alignment of choices.
104. From the perspective of the account of rationality I have sketched here, it is tempting to infer from the widespread use of surveillance and enforcement systems that human beings are only imperfectly rational. Certainly that has to be part of the story. But, as I indicated above, it is plausible to suppose that a commitment to be guided by rules is conditioned on a sense of the fairness of the arrangement denned by those rules. It seems to me an open question as to how much of the need to resort to "second-best" surveillance and enforcement systems is due to imperfect rationality, and how much is due to a need on the part of those who are unfairly advantaged by this or that system of rules to enforce them against those others who are unfairly disadvantaged.
105. It would seem that this conceptual account holds even if resoluteness is conceived as merely the imposition, by the earlier self, of a regimen which the later self accepts, or, in the case of interpersonal choice, of a pure convention, among a group of people, regarding how each is to constrain choice in certain situations.
106. I take this phrasing from Rawls, "Two Concepts of Rules," p. 24.
107. The phrase, "a new form of activity," is again taken from Rawls, "Two Concepts of Rules," p. 24.