Game Theory, Freedom and Indeterminacy
Kevin Quinn (Bowling Green State University, USA)
© Copyright: Kevin Quinn 2006
“A common opinion prevails that the juice has ages ago been squeezed out of the free-will controversy, and that no new champion can do more than warm up stale arguments which everyone has heard. This is a radical mistake. I know of no subject less worn out, or in which inventive genius has a better chance of breaking new ground, - not, perhaps, of forcing a conclusion or of coercing assent, but of deepening our sense of what the issue between the two parties really is, of what the idea of fate and free will imply.”
William James, “The Dilemma of Determinism.”
“All the influences were there waiting for me. I was born, and there they were to form me, which is why I tell you more of them than of myself.”
-Saul Bellow, The Adventures of Augie March.
“Sovereignty, which is always spurious if claimed by an isolated single entity, be it the individual entity of the person or the collective entity of the nation, assumes, in the case of many men mutually bound by promises, a certain limited reality.”
-Hannah Arendt, The Human Condition.
How can we economists reconcile our conviction that we are free with what we spend our lives doing, namely, offering up causal explanations of human behavior? If we’re so free, how come we’re so predictable? If we are rational choosers, then, given our beliefs, desires and opportunity set, our choice is predictable. And if the preferences that shape our choice are in turn predictable, whether by biologists, evolutionary psychologists, anthropologists or sociologists, what sort of freedom is that?
In this paper I contend that classical game theory, and the “inventive genius” (see the James quote above) of the likes of John Von Neumann, Oskar Morgenstern, and Thomas Schelling, have given us a new way of thinking about the old issue of free will versus determinism. I claim that an appreciation of the indeterminacy that obtains in games with multiple Nash equilibria allows us to reconcile the scientific explanation of human agency with human freedom. This can be done if, and only if, the link between scientific explanation and determinism can be broken - and this is precisely what classical game theory allows us to do. In games with multiple equilibria, looking back from an achieved equilibrium, it is manifestly the case that we - though not any one of us taken individually - could have acted differently, in just the sense that champions of free will have always maintained was incompatible with science while being necessary for the ascription of genuine freedom.
The first section of the paper sets the scene with a survey of the traditional free will versus determinism debate, paying special attention to the determinist strategy that argues for the compatibility of freedom, properly understood, and determinism - and indeed for the meaninglessness of an account of freedom that doesn’t in fact presuppose determinism. The second section then fleshes out my candidate account of freedom as one that preserves important components of the traditional free will position - most importantly, the essential link that position makes between freedom and indeterminacy. To make the argument, I defend the indeterminacy of game theoretic explanations from those - many game theorists themselves - who argue that this indeterminacy is a failure of the theory in need of correction. In the next two sections I then examine and criticize the growing tendency among game theorists, fueled in part by embarrassment over indeterminacy, to abandon rationality altogether. This can be done either with the methods of evolutionary game theory (section 3) or with an appeal to conventions, salience or focal points (section 4). The next section makes the case that freedom understood along the lines I have proposed is a species of positive liberty, in Isaiah Berlin’s sense, and indeed allows for a plausible interpretation of that much-maligned notion. I also address the question of “spontaneous order,” in Hayek’s sense and argue that the faith in the emergence of such an order in general can only be grounded on the non-rationality of human beings; that for rational agents, spontaneous order is not an option and politics is therefore inescapable. The last section concludes with a brief comparison of my account of positive freedom with that of the great political theorist Hannah Arendt.
Setting the Scene: Compatibilism and Its Opponents
A host of thinkers have tried to square freedom with predictability, or freedom with determinism - this is the compatibilist project, as it’s called. Mill was a well-known exponent of this view.1 We are free, he says, when we do what we want. The fact that we didn’t choose our wants, that they are in principle predictable, shouldn’t bother us. Kant, on the other hand (in a crude interpretation, to be supplemented below with a somewhat more subtle one), sure that genuine freedom was incompatible with determinism, located our freedom not in the phenomenal realm, where objects in space and time exist subject to causal laws, but in a mysterious “noumenal” realm. On its face, at any rate, he seems to throw out a scientific account of human agency in order to save something like a metaphysical conception of freedom. If explanation in the social sciences is conceived naturalistically - as no different in kind from explanation in the natural sciences - it would seem that the notion of a metaphysically free will can find small purchase, and must be rejected as a holdover from religious views of the world, condemned to obsolescence by the rise of science.
The debate between compatibilism and incompatibilism commences along with, and in response to, the scientific revolution of the 17th century. The most convincing and influential of the early progenitors of compatiblism was Baruch Spinoza, the dominating figure in Jonathan Israel’s magisterial Radical Enlightenment. Israel writes: “That men suppose themselves to be free, Spinoza ascribes to their consciousness of their desires and appetites while failing to see ‘those causes by which they are disposed to wanting and willing, being ignorant of those causes.’”2 Here Spinoza is debunking the traditional conception of freedom, which is incompatible with determinism. One of Spinoza’s British followers, Matthew Collins, describes the sort of freedom that is compatible with determinism, making the compatibilist case in essentially the same way it would be made by Mill nearly 200 years later:
“Whenever the doing or forbearing any action, according to the determination of my will is in my power I am then always free and at liberty, that is, free from any agents hindering me from acting as I will, but not free from necessity. For when I will, or prefer, going abroad to staying at home, that act of volition as much determines me to act according to that preference, if it is in my power to go abroad, as locks and bars will hinder one from acting according to that preference. The only difference is that in the one case I am necessitated to act as I will, and in the other case contrary to my will.....This seems to me..to contain the whole idea of human liberty.”3
The anti-compatibilists of the time were the Cartesians. Descartes’s radical body/soul dualism, in what looks like an earlier version of Kant’s later move noted above, locates the free will in the soul, not the thoroughly determined body. Human freedom is incompatible with scientific determinism, but real nonetheless, since located in a realm, the soul, where science holds no sway.
Flashing forward to the present, here is Julian Sanchez , reviewing a new book by the compatibilist Owen Flanagan, glossing the latter’s argument, a mainstay of the compatibilist position, that not only is freedom not incompatible with determinism, but that it is, in fact, incoherent without it:
The human capacity for free choice is another cornerstone of liberal thought that seems threatened by a thoroughly naturalized conception of persons. Real choices are supposed to be undetermined by what comes before. When I make a genuinely free choice, no set of antecedent conditions predetermines what I must do . But an exercise of free will is supposed to be something that an agent does, not something that merely happens. It would not count as free will if some non-deterministic quantum fluctuation in my brain caused me to do good rather than evil. These two conditions - indeterminacy and authorship - together define free will as traditionally conceived. But, as Flanagan observes, they are mutually incompatible. To the extent that my actions are undetermined - that I could have turned right just as easily as left -they are not bound to any of my past mental states. To the extent that my own experience and reasoning do explain my actions, those actions are determined and, therefore, not ‘free’ in the radical sense. 4
I will call the compatibilist account of freedom “deflationary” as compared with the traditional conception, which I will refer to, following Sanchez, as “radical,” or sometimes “metaphysical,” freedom. The absence of determinism, the compatibilist argues, would not at all help the case for a non-deflationary, radical freedom.
The attempt to preserve freedom by appealing to indeterminacy and chance in nature itself, excoriated by the compatibilists, nonetheless has a long history. One of the most famous attempts along these lines was by William James, the American philosopher, following a nervous breakdown brought on by struggling to reconcile science and free will. His solution5 was the postulate of a radically indeterministic universe- a non-solution, if the compatibilists are right. James scorned the compatibilists as “soft determinists,” as a determinism which “says that its real name is freedom, for freedom is only necessity understood.” Though flawed, James’ move is interesting because, rather than embracing any sort of supernaturalism, it would make naturalistic, empirical explanations in general more expansive. Like the compatibilists he is a monist6- his thorough-going empiricism has no use for a non-empirical, realm, neither Kant’s noumenal realm nor Descartes’ mysterious soul - but his monism works by inflating nature, if you like, hoping to reconcile naturalism with more inflationary accounts of human freedom, rather than deflating the latter along compatibilist lines.
Dualism, of course, is not touched by the compatibilist critique, although the price of such invulnerability - to the extent that it involves postulating entities such as souls - is not one most of us would be willing to pay. Though Descartes’ version is not widely embraced, dualism retains its appeal in some quarters. Like Descartes, later dualists deny that one can give a scientific account of human agency. If science only allows the choice between freedom as random, unpredictable action and the thin compatibilist notion, so much the worse for science. Think of Sartre’s existentialism: people, unlike things, have no essence; our essence, as he put it, is to exist.7 Escaping the causal nexus, thus, we are capable of radical choice, though constantly attempting to deny this about ourselves. When we think of ourselves as things, determined in our action, we are exhibiting what Sartre calls mauvaise foi, (bad faith).
A more subtle attempt along these lines is that of Kant. He argued that causality and determinism are imposed on the world by our reason; not a property of things-in-themselves but a human construction. Our freedom can never be a matter of knowledge, for knowledge, given the way we are constituted, can only ever be of phenomena, causally determined in space and time. But we are free, Kant argued, when we engage our practical reason, by acting in accordance with the moral law, a law we give to ourselves, in the teeth of inclination.8 Our freedom is bound up with our status as rational agents, and takes the form of acting on obligations to respect both others and ourselves as rational agents. The moral law free agents give themselves carries the well-known injunction never to treat ourselves or others as mere means, always as ends - one formulation of the categorical imperative.
Before trying to salvage something from the apparent wreckage of the metaphysical conception of freedom on the shoals of science, let me mention a parallel discussion of political freedom or liberty. Since Isaiah Berlin’s seminal essay9, it is a commonplace to distinguish between two concepts of liberty, negative and positive - usually, among self-styled Liberals, preparatory to anathematizing the latter. Negative liberty simply means the absence of coercion, the ability to do what one likes.
In Berlin’s essay, positive freedom comes in several varieties. It may be, in the first place, the idea of participating as a citizen, along with other citizens, in democratic self-governance. Second, it may refer to the idea of being enabled to become “who one is.” In this second conception, we are only free when we pare away non-genuine, inauthentic preferences and act on the basis of the remaining, better self. Berlin himself finds the latter conception at the root of utopian totalitarianisms, where the State takes it upon itself to force its benighted subjects to be free. Positive freedom of either variety entails that the uncoerced subject may well be unfree, because unable or unwilling to participate in self-government, on the first reading, or a slave to inauthentic desires, on the second; and the coerced subject, on the contrary - coerced by the will of the people or a dictator with one’s allegedly genuine interests in mind, respectively - free.
There seems to be a natural connection between compatibilism, on the one hand, and an exclusively negative conception of political liberty on the other. On the compatibilist account, after all, what else is there for freedom to be but acting on one’s preferences without interference? The first variety of positive freedom seems to privilege one preference above the rest and give it lexicographic priority. But it is empirically dubious that such a privileging obtains. Many of us are quite happy to trade-off a desire to participate in politics for other stronger desires; and efficiency would seem to argue for benefits of the division of labor here as elsewhere, with some people specializing in governance. The second conception of positive liberty in many formulations smacks of incompatibilism, the criteria for genuineness among preferences sounding often like a requirement of radical authorship, with non-genuine preferences having been caused by external factors, and genuine preferences not so caused. The picture of a self able to reflect on its own preferences - where that is not simply a matter of second-order preferences with no more authority than the preferences they sit in judgement upon - would also seem to defy compatibilist accounts.
The fact that the compatibilist account of freedom and a strictly negative account of political freedom have always met, and continue to meet, strong resistance despite the apparently compelling case that can be made for each, is, I think, significant. With the account of freedom I offer, I hope to capture some of the themes in this alternative literature while maintaining, like compatibilism, a naturalism about human agency ( in this respect, my strategy mirrors James in its monism). In particular, I hope to capture the idea that freedom and indeterminacy are mutually implicating, and to show: that the concept of bad faith has a naturalistic application; that freedom is intimately connected with rationality, and with respect for the rationality of other rational agents; and that positive liberty understood as democratic self-governance has some claim to be seen as freedom par excellence.
I hasten to say that my account will please neither compatibilists nor non-compatibilists. Non-compatibilists will most likely find it almost as deflationary as the compatibilist account, and find my re-interpretations of their views unpersuasive. Compatibilists, on the other hand, will find that it grants too much to the other side. But where positions are so dug in, with neither side conceding defeat over many centuries, a correct account, if such there is, will surely incorporate something from both camps - inevitably offending both. My candidate account, at any rate, satisfies this necessary, though certainly not sufficient, criterion! And I would like to think that William James would be pleased10.
The connection between indeterminacy and freedom
James was right to link the freedom with indeterminacy, and to reject the idea that scientific explanation is incompatible with indeterminacy. The indeterminacy I want to link with freedom is the indeterminacy that obtains in non-cooperative games with multiple Nash equilibria played by rational agents with common knowledge of one another’s rationality. Our freedom is then the freedom to coordinate on any of the multiple equilibria: the rational choice account of human agency, arguably the best scientific account we have, cannot say anything about which one that will be. If this right, then note that this is properly our freedom: isolated agents cannot be said to be free in this sense11. This I think helps to explain how standard compatibilist accounts can be persuasive: there is nothing that corresponds with radical freedom that can be said to characterize individuals qua individuals. At the same time it explains why radical accounts cannot be defeated, since there is something highly congruous with radical or metaphysical accounts that pertains to the interaction of rational agents with common knowledge of rationality: how we successfully coordinate is radically unpredictable, and it is at the same time our achievement, for which we are responsible. James thought that a universe full of chance might solve the problem of reconciling science and free will. But this just makes our actions thoroughly contingent: we are not responsible for them - they are not our actions, so they cannot be the actions, however unpredictable, of free agents. In my account, individuals who succeed in coordinating on one of many equilibria, by, e.g., talking together, have determined themselves: freedom is the freedom we have to so determine ourselves12.
This account has advantages over the standard compatibilist account, in virtue of its greater ability to capture some of what traditional accounts encompass. It is less deflationist than the standard compatibilist account, while remaining, I claim - and this is its advantage over the radical account - compatible, not with determinism, but with science. Obviously, this needs to be argued for.
Freedom as I understand it is compatible with scientific explanation, with naturalism about human agency, provided that game theoretic accounts that stipulate rationality and common knowledge of rationality are so compatible. This, for many, may appear to beg the question. For many, the very fact that game theory cannot tell us which of the multiple Nash equilibria encountered in so many games will be selected, is a signal failing of the theory, serious enough, indeed, to make its explanatory force suspect.13 The fact that game theoretic explanations14 imply indeterminacy (and thus freedom, in my sense) is for many, if not most, a reason, not to accept indeterminacy, but to reject game theory.
Quine taught us15 that there are no beliefs in the web of beliefs we take to be our fallible knowledge at any time that may not, under certain circumstances, in the face of new evidence, be jettisoned, including, famously, what prior to Quine we were pleased to call conceptual or analytic truths. Admittedly, the belief that a scientific explanation, to be such, must make a unique prediction is a highly embedded belief - right up there with the belief that bachelors are unmarried!
Nevertheless, given the progress of our discipline, so, too, is the belief in rationality and the sophisticated development given to that concept by game theoreticians, including the idea of common knowledge. Certainly, in the wake of the seminal work of Thomas Kuhn16 and the whole revolution in philosophy of science he inaugurated, whatever prior temptation one may have had to believe that there are any analytic truths about scientific explanation, to believe that the criteria for scientific explanation have not themselves evolved along with the growth of knowledge, ought to have been sufficiently allayed.
So it is arguable - I will so argue - that the belief that agents are rational has as much claim to being in the hard core of our web of beliefs as does the belief that scientific explanations must make unique predictions. Where they conflict - and they do conflict in games with multiple equilibria - I am inclined to jettison the latter, not least because doing so allows us, as I have argued, to salvage something of the equally hard-core beliefs about human freedom that the standard compatibilist account has never succeeded in undermining.
It is finally only the prejudice against the idea of unpredictability in an explanation that keeps us from saying, instead of “game theory fails insofar as it cannot allow us to predict a unique outcome,” that “game theory succeeds at showing us why many of the interactions of rational agents with common knowledge of each other’s rationality are in principle unpredictable. This unpredictability, I claim - our freedom - is a fact about the world, a fact that game theory can persuasively explain, not a failure of the theory. Game theorists, instead of being embarrassed about indeterminacy, or promising future refinements that will get rid of it, or, even worse, abandoning rationality altogether17, need to hold their heads high and tell the critics, “We have been able to give, without any appeal to metaphysics or the supernatural or any sort of religious thinking, an account of the deep sense we have that rational animals, unlike all others, are free. And we needn’t use anything more than plain old instrumental rationality here - we needn’t appeal, like Kant and many others, to a scientifically suspect non-instrumental reason18. Pascal was right that we are “ni ange, ni brute,” neither angels or brutes - not brutes, because we are rational; but not angels either, because it is plain old vanilla instrumental rationality that we have, nothing fancier, nothing of the angelic species that Kant imagined.
Kant was absolutely right, though, in two respects: first, in tying freedom closely to the idea of rational agency; and in his conviction, second, that freedom is bound up with respect for the rationality of others - treating them never as mere means. I would say that freedom emerges when we stop forming beliefs about others based on statistical probabilities, treating one another parametrically, on a par with the weather, and start treating one another as rational agents; it emerges, that is, with rationality. This is a far cry from Kant’s moral law, I know: these are just instrumentally rational agents who are trying to advance their interests. And although freedom isn’t associated with any categorical imperatives (do x unconditionally, whatever you happen to want), it is associated with the failure of any straightforward hypothetical imperative to be action guiding in the context of strategic interdependence and indeterminacy - “if you want y, do......what, exactly?” 19
Abandoning Reason I: Evolutionary Game Theory
The following coordination game is justly famous; it is Rousseau’s Stag Hunt:
Payoffs: Player 1, Player 2
On opposite sides of a clearing, two hunters sit in the brush. A stag will be along soon, as will several hares. The stag can only be caught if both act together to trap it in the clearing. If caught the large game gives payoffs of 6 to each player. On the other hand, each can easily catch, without help, one of the hares who frequently appear. This choice would leave the other hunter, who had chosen stag, with neither stag nor hare and a payoff of zero; the smaller hare has a payoff of 2 for its captor. There are two equilibria in pure strategies, Stag/Stag and Hare/Hare and an equilibrium in mixed strategies which has each playing Stag (Hare) with probability 1/3 (2/3) and expected payoffs of 2. The game is incidentally a marvelous metaphor for the emergence, or failure to emerge, of a complex market-coordinated division of labor, where choosing Stag is the analogue of choosing to specialize, intending to trade the bulk of what one produces for a variety of consumption goods, which will have a very low payoff indeed of others haven’t chosen to specialize as well; while the hare strategy is the analogue of autarky.
One way to achieve determinacy here is to abandon rationality altogether. Let people among a large population randomly pair up and play the game. Let p be the proportion of the population choosing Stag at any time, and let people be rational only in the sense that they learn from experience and gravitate toward the most successful strategy over time. p will then increase over time whenever the Return to Stag (6*p) is greater than the return to Hare (2) and decrease when the inequality is reversed. Then if p > 1/3 it will increase, and if p< 1/3 it will decrease. The dynamics of p are then perfectly determinate; if p is initially above 1/3, it rises to 1 over time; if < 1/3, it falls to zero. Knowing the initial p, we know exactly which of the pure strategy equilibria will be reached.
But what an enormous sacrifice this is to make for the sake of determinacy! The agents in this scenario are not rational agents, but brutes, no less brutish than their prey. In fact, as the reader doubtless recognizes, this is simply an application of the methods of evolutionary game theory, which are making more and more inroads among economists, alarmed enough about indeterminacy, or so it would appear, that they are prepared to sell their birthright - the proud tradition of rationality - for a mess of evolutionary pottage.20 The only change we would need to make to make the application exact would be to “hard-wire” a pure strategy into the agent and then postulate greater reproductive success for those hard-wired with the higher return strategy.
Achieving a determinate outcome would be assured, too, if we formed beliefs about each other just as we form beliefs about the weather, if we started with a guess and then adjusted our subjective p towards the last period’s actual p adaptively. Starting from an initial distribution of subjective priors, the outcome would be entirely predictable. But this would be to ignore the fact that we are playing rational agents just like ourselves, to treat one another not as people but as things. The echo of Kant and of Sartre here is intended: we would be acting in bad faith, denying our freedom if we were to act in this fashion. Using game theoretic techniques in this way is using a mean and paltry version of the theory, one shorn of its very heart and soul: rationality. It is Ulysses without Leopold Bloom.
Abandoning Reason II : Conventions, Focal Points, Salience
Coordination games have been thought either to throw light upon, or alternatively to be enlightened by, the notion of a convention. In either case, this is a mistake. Margaret Gilbert’s work21 explains why, pace David Lewis22, a convention is not reducible to one of the multiple equilibria in a coordination game which has been achieved. I want to argue against the reverse implicature , the idea that a convention can solve the “equilibrium selection” problem in a coordination game - relying on her arguments and those of Martin Hollis.
Suppose, then, that there is a convention among us that one plays Stag in the Stag-Hunt Game. How does this solve the problem?23 Does it do so by creating the belief among us that the other will follow the convention? But that is inconsistent with how beliefs are formed in games among rational players whose rationality is common knowledge. I should believe that you will play conventionally, just in case I believe that you believe I will play conventionally; that is, if I believe that you believe that I believe that you believe I will play conventionally - and so on ad infinitum. But equally, I should believe that you will play unconventionally if I believe that you believe I will play unconventionally...and so on. This is obviously the same coordination game, with strategies of playing Stag or Hare, respectively, replaced by strategies of playing conventionally or unconventionally - and has two equally good equilibria. Rational people are not bound by conventions, or, as I would put it, rational people are free. Whether they will follow the convention is in principle unpredictable. When I act based on the brute belief that you will follow the convention, I am not respecting your rationality, which requires that you play a best response to what you believe I will do.
What we need to do, rather, we rational agents, is to agree on a way to play. Gilbert argues that what we do when we agree on a way to play is to form what she calls “a plural subject” We agree. This plural subject, she claims, then gives each of reasons that are not reducible to individual reasons: from “We have agreed on Stag,”it would , she argues, follow directly, without any need to specify individual goals, that each of has a reason to play Stag. I am not very comfortable with this way of putting things, but I cannot decide whether I have a substantive or simply a semantic disagreement. Here is the way I would prefer to put it: coming to an agreement here is an exercise, quite literally, in self-determination. In the next section, I make this the defining characteristic of the political realm, the realm of positive freedom. What is irreducible, then, for me, is not, as I think Gilbert says, “the Social,” but the Political24.
Similar objections pertain to another candidate for equilibrium selection, Schelling’s notion of focal points or salience. If I want to meet you in New York City, the fact that Grand Central Station is salient does not give me a reason to go there under the terms of classical game theory. However salient it is, I have just as much reason to go to a non-salient spot if I think you will. And I do think you will - you have good reason to - if you think I will. Believing you are rational, I believe you will play a best response to what you think I will do. To believe that you will go to a spot because it is salient, whether or not it’s a best response to what you think I will do, is to believe that you are not rational: “Because it is salient” is not a reason. Of course, it may be a compulsion - but here again, even more obviously, we’ve left the realm of reason.25
Positive Freedom and The Inescapability of Politics
Let us call the realm of the political that realm, wherever it happens to arise in life, where instrumentally rational agents reach agreement on how to coordinate their behavior in games with multiple Nash equilibria, and thereby achieve self-determination. Then the realm of the political is the realm of freedom par excellence, as I have been using the term. If negative freedom is the ability to do what we want, then positive freedom is simply the ability to reach an agreement with others when it is not clear, due to strategic interdependence among rational agents who know themselves to be rational, what it is we should do to get what we want - when, that is, the game has multiple equilibria. Politics becomes the art of the possible, not in contrast to the impossible, as the phrase is usually intended, but in contrast to the uniquely determined: it is the realm where we have options. Nor is this a small realm, especially when it is appreciated that every iterated prisoner’s dilemma with no terminal date is a coordination game.
Hayekians have always argued against political coordination in favor of the virtues of spontaneous order. This argument is generally persuasive only when coupled with their equally perennial denial that agents are rational in the sense of classical game theory. As we have shown above, and as Sugden’s26 interesting work has made clear, a collection of non-rational agents will indeed achieve a spontaneous order, a determinate equilibrium in games where they are multiple. It may not be an efficient equilibrium, but it will be an equilibrium nonetheless. Genuinely rational agents, on the other hand, have no idea what to do in such a situation: to achieve any equilibrium at all will require politics.
Let us take a look at one of Sugden’s and John Maynard Smith’s favorite games for an illustration.
Here there are multiple Nash equilibria, two asymmetrical (H/D and D/H), and one symmetrical the latter an equilibrium in mixed strategies with Hawk and Dove played with probablity 1/2 each. The mixed strategy equilibrium gives lower average payoffs per person, 1.5 versus 2, because it involves the resource being destroyed by mutual hawkishness with probability 1/4. Sugden ingeniously shows how non-rational agents (trial and error rational27, that is) playing anonymous random opponents can stumble spontaneously and predictably into an equilibrium where the conditional strategy “Play Hawk, if you are in possession; Play Dove, otherwise” is adopted by each. (The conditioning factor could be anything: play Hawk if are the taller player, or if you are not in possession. e.g.; Dove otherwise. Sugden argues that some of these asymmetric roles, depending on the game, will effectively be more salient.)
But truly rational agents can’t stumble into anything, and, for them, the indeterminacy here makes politics inescapable. This stands the Hayekian message on its head. Hayek would say that spontaneous order here is desirable compared with the inevitable distributional struggle that an opening to politics would entail. Again, with sub-rational agents this is persuasive. But with rational agents, it is only a political settlement that can end what would otherwise be a potentially chaotic struggle for resources. Once there was an Eden of spontaneous order, where innocent a-rational agents had no need for politics. But having tasted of the fruit of the Tree of Rationality, despite having been warned, they were evicted from the Garden and forced to explicitly order what had once been spontaneously done.
Envoi : Arendt and Politics
Without departing (if my earlier arguments are convincing) from a scientific account of human agency, this account of the political captures, I think, some of what our greatest philosopher of positive freedom, Hannah Arendt, had in mind in defining the political as the realm of freedom. Arendt made the political the realm of unpredictable action, as opposed to predictable behavior . And the freedom she saw in the political realm was indeed “our” freedom - tied constitutively to the plural nature of the public realm. In the quote that appears at the head of the paper, she claims that no individual man can be said to be sovereign, only many men mutually bound by promises. Arendt would not, I’m sure, be pleased to have her ideas deflated in this way - she clearly rejected the economists’ idea that we are simply instrumentally rational agents. In her polis, we are free in virtue of acting and speaking together, but as soon as we begin speaking about how to obtain our individual, pre-political ends, we have lost our freedom, falling into the realm of the social or what she sneeringly calls “national house-keeping.” Jon Elster28 wondered along with many just what it is we do talk about in that case. My account of positive freedom maintains the qualitative distinction between the political and the non-political that Arendt wanted, but gives us something to talk about in the forum, without appealing to any other kind of agency than the instrumental.
1. Martin Hollis, in his wonderful Introduction to the Philosophy of Social Science (1994) makes Mill the spokesman for compatibilism.
2. Israel, Radical Enlightenment (2001), p. 232
3. Collins cited in Israel, p.616
4. Sanchez, “Self Delusions: Does Morality Require a Soul, in Reason, January 2004.
5. In “The Dilemma of Determinism” (1968)
6. A monist in this respect - in rejecting any appeal beyond experience - he nonetheless calls himself a pluralist, because he makes experience itself plural, meaning that the parts of the universe “have a certain amount of loose play on one another, so that the laying down of one of them does not necessarily determine what the others shall be.”
7. Sartre surely intended us, with this phrasing, to hear the echo of the medieval conception of God as self-caused: the only entity whose essence entails existence. Sartre’s philosophy makes each one of us authors of ourselves in just the way theologians saw God to be.
8. Amartya Sen’s notion of commitment as ‘counter-preferential’ choice has a strong Kantian flavor. See his “Rational Fools” (1991).
9. “Two concepts of liberty.” (1969)
10. The James who wrote: “Of the two alternative futures we conceive, both may now be really possible; and the one becomes impossible only when the other excludes it by becoming real itself. Indeterminism thus denies the world to be one unbending unit of fact. It says there is a certain ultimate pluralism to it; and so saying, it corroborates our ordinary, unsophisticated view of things. To that view, actualities seem to float in a wider sea of possibilities from out of which they are chosen; and somewhere, indeterminism says, such possibilities exist, and form a part of truth. (“The Dilemma of Determinism,” op. cit., p.591.
11. The liberty of one requires, if not the liberty of all, the liberty of at least one other!
12. Consider a simple coordination game, such as driving on the left or right side of the road. In either equilibrium, each is driving on the side in question because others are. Thus, at the aggregate level, we drive on the left (right) side of the road because we drive on the left (right) side of the road - the connection with the notion of self-authorship is here obvious. Our actions in any particular equilibrium are their own causes.
13. In his witty principles primer, Hidden Order, David Friedman titles the few pages in the book that use game theory ( the section on oligopoly, p.165 )“TOO MANY ANSWERS.” Hundreds of remarks by others along these lines might be cited - I don’t single Friedman out - by people for whom more than one is “too many.”
14. That is, game theoretic explanations with rational agents and common knowledge of rationality.
15. In “Two Dogmas of Empiricism.” (1980)
16. in The Structure of Scientific Revolutions (1962)
17. The sad topic of the next section.
18. See Jean Hampton, The Authority of Reason (1998) on the “queerness” of non-instrumental reasons. I should say she rejects this position and argues that to the extent that science finds the idea of the authority of a norm queer, it undermines itself.
19. Joseph Heath in his excellent Communicative Action and Rational Choice (2001) argues from indeterminacy to the need to postulate non-instrumental reasons for choice. In his account we have “normative” preferences which rank actions directly, along with standard instrumental ranking of actions based on the their varying perceived efficacy of achieving outcomes, and a weighting scheme which assigns weights to the two different sorts of reasons. I am uncomfortable with this sort of move, because it restores determinism. I argue from indeterminacy, not to a new and improved rationality, but to freedom. An excellent discussion of the implications of indeterminacy for game theory is Hargreaves Heap and Varoufakis, Game Theory (1995). They argue - or at least one of them argues; they disagree - that indeterminacy may sound the death knell for methodological individualism, that irreducibly social phenomena -such as a convention, on some understandings- may be required to “solve” the equilibrium selection problem in games with multiple equilibria. Again, for me this is not a problem that needs to be solved!
20. Don’t get me wrong: evolutionary methods have their place in modeling animals. But once evolution has thrown up - not at all mysteriously - rational animals, matters need to left to the economists. The problem is not that the evolutionary psychologists don’t understand our immortal souls - the problem is that they don’t understand instrumental rationality.
21. See her “Rationality, Coordination and Convention,” “Rationality and Salience,” and “Notes on the Concept of a Social Convention,” all reprinted in her Living Together (1996)
22. Convention (1969)
23. Here I follow Hollis closely, op cit., p. 137 et seq.
24. This is somewhat misleading. Gilbert wants to analyze conventions themselves as fundamentally agreements. Her category of the social, that is, is already through and through political. I remain uncomfortable, though, with the “plural subject” locution: it seems to me to reify our agreement.
25. In Natural Reasons (1989) Susan Hurley points out that Schelling himself never thought that salience as a brute fact could solve a coordination game. She quotes this, adding her own emphasis: “ ‘ In the mutually recognized response of players to salient characteristics, the fundamental psychic and intellectual process is that of participating in the creation of traditions. The players must jointly discover and mutually acquiesce in a mode of play that makes the outcome determinate. They must together find rules of the game or suffer the consequences.’”(p.155)
26. in The Economics of Rights, Cooperation and Welfare (1986)
27. “I shall assume that individuals tend to adopt those strategies that proved successful over a long sequence of games.” And, “The theory of games is often defined as the theory of how games would be played by completely rational individuals...it is here that my approach to game theory diverges form the traditional one. Indeed, on a strict interpretation of these definitions, this book is not about game theory at all.” op. cit. p.16
28. In his “The market and the forum” (Elster ,1986)
Arendt, Hannah, 1958, The Human Condition. Chicago: University of Chicago Press.
Berlin, Isaiah, 1969 , “Two Concepts of Liberty,” in Four Essays on Liberty. Oxford: Oxford University Press.
Elster, John, 1986, “The market and the forum: Three varieties of political theory,” in Elster and
Aanund Hylland, eds., Foundations of Social Choice Theory. Cambridge: Cambridge
Friedman, David, 1997, Hidden Order. New York: HarperCollins.
Gilbert, Margaret, 1996, Living Together. Lanham, Md. : Rowman and Littlefield.
Hampton, Jean, 1998, The Authority of Reason. New York: Cambridge University Press.
Hargreaves Heap, Sean amd Yanis Varoufakis, 1995, Game Theory: A Critical Introduction. (London: Routledge.)
Heath, Joseph, 2001. Communicative Action and Rational Choice. Cambridge: MIT Press.
Hollis, Martin, 1994, The Philosophy of Social Science. New York: Cambridge University Press.
Hurley, Susan, 1989, Natural Reasons. New York: Oxford University Press.
Israel, Jonathan, 2001, Radical Enlightenment. New York: Oxford University Press.
James, William, 1968, “The Dilemma of Determinism,” in The Writings of William James. New York: The Modern Library.
Kuhn, Thomas, 1962 , The Structure of Scientific Revolutions. Chicago: University of Chicago Press.
Lewis, David, 1969 , Convention: A Philosophical Study. Cambridge: Harvard University Press.
Perelman, Sidney, 1973 , Crazy Like a Fox. New York: Vintage Press.
Quine, W.V.O., 1980, “Two Dogmas of Empiricism,” in From A Logical Point of View. Cambridge: Harvard University Press.
Sanchez, Julian, 2004. “Does Reason Require A Soul,” Reason, January, 2004.
Sen, Amartya, 1990 , “Rational Fools,” reprinted in Jane Mansbridge, Ed., Beyond Self-Interest. Chicago: University of Chicago Press.
Sugden, Robert, 1986, The Economics of Rights, Cooperation and Welfare. (Oxford: Basil Blackwell.)