Binmore, Ken. Natural Justice. Oxford: Oxford University Press, 2005. Pp. 207. $29.95 (cloth). Behavioral economics, game theory, and evolution offer powerful models and often provocative explanations of human behavior. The convergence of their methodologies is yielding intriguing new hypotheses, and the developing research program will surely become quite influential. The nature and the origins of the moral sentiments have fascinated thinkers for centuries, and further scientific inquiry into those questions is welcome. Reconciling the wide (but frequently weak) sociality of humans with the selfreplicating gene of evolutionary biology is a major and unsatisfied explanatory challenge. The most promising new work is constrained by the strong empiricism of behavioral economics, and happily is more in the spirit of Kropotkin than Spencer. For example, Bowles, Gintis and coworkers carried out experimental tests of the ultimatum game in 15 diverse cultures, the results of which decisively challenge the realism of the homo economicus assumption (Joseph Henrich et. al, Foundations of Human Sociality. Oxford University Press, 2004).. Ken Binmore is a mathematician who turned to economics and then to political theory. Natural Justice is an abridgement and attempted popularization of his two-volume, 1013- page work, Game Theory and the Social Contract (1994, 1998). The larger work contains some friendly education in game theory, is thoughtfully organized, and makes some detailed arguments. At the same time it contains formal passages challenging even to specialists, frequent digressions, and numerous controversial but undefended assertions. The smaller work at hand diminishes the virtues and exaggerates the vices of its ancestor. As the author explains, the reader who wants proofs, references, and, often, arguments, should turn to the larger work (the smaller work in its margins cites to relevant sections of the larger work). Most indefinitely repeated games of interest yield multiple equilibrium states, for example, the folk theorem of game theory shows that the repeated prisoners dilemma yields equilibria of any mixture of cooperation and defection as good as or better than mutual defection (repetition of the game allows each party to threaten defection in response to defection). Thus, the question becomes, which equilibrium would be selected, which should be selected? Ethics arose from Nature s attempt to solve certain equilibrium-selection problems, according to the author (140). Here is his main line of argument. Two-person bargaining models yield many Paretoefficient distributions, for example, from nothing to everything for one of the parties. If the players have equal bargaining power, then the Nash bargaining solution maximization of the product of the players gains over their disagreement point seems most salient. Discernment of fairness biologically evolves as a calculation of the Nash solution for situations where direct bargaining is too difficult or costly. The mechanisms which produce sympathy towards future selves and reciprocity towards kin generalize to food-sharing in the ancient group, which further generalizes to a natural ability to don as appropriate the Rawlsian veil of ignorance in everyday affairs.
Language is a biological capacity, but the particular language is culturally transmitted. Similarly, fairness culturally evolves such that the weighting of payoffs to each party differs from culture to culture. The same bargaining problem also allows for utilitarian and egalitarian solutions. If there is external enforcement of bargains, then cultural evolution would weight the two parties such that the utilitarian solution would coincide with the Nash solution; if no external enforcement, then cultural evolution weights the two parties such that the egalitarian solution coincides with Nash. Fairness norms prevail in a particular society because they allow its members to coordinate rapidly on efficient solutions to bargaining problems. Societies which coordinate on such norms live, and those which fail to so coordinate die. Memes are to culture what genes are to biology. There is no place for that distinctive human capacity, reason, in evaluative judgments: humans are no more than a mouthpiece for the cultural memes that have taken over their minds (52). The author s science, however, is somehow exempt from such determinism. What is fair differs incomparably from culture to culture, and is defined as the prevailing equilibrium enforced by threat of reciprocal punishments in a repeated game. Nothing else: fairness has no independent content or force. Purely moral motivation is mythological: the author is certain that we are a species that lies and cheats when we think we can get away with it (137, also 9, 40, 68). Daughters feel duty to their aged mothers ultimately only from fear of social censure (87). Other than kinship altruism, and reciprocally enforced interaction, humans are egoistic and amoral. There is no inborn notion of the Right (94). Those not beguiled by theory, however, recognize that most of us are morally motivated, sometimes strongly, often weakly. And humans, within certain ranges, often feel trust or moral obligation towards anonymous strangers. Theories of the egoistic ilk, such as the author s, either deny the truth of such observations, or resort to hypothesis-saving mistake theories. Any feeling of trust or obligation towards a stranger is because one mistakes him for kin, says the author (112). Subjects in anonymous prisoner s dilemma experiments cooperate about half the time, but if motivated by the material payoffs alone they would not have cooperated at all. The author retorts that upon repeated experimental play (with no communication) cooperation decays, indicating that players initially were mistaken about how to play the game (66). He neglects to report, however, that repeated games allowing anonymous and noncredible communication produce high and continuing levels of cooperation suggesting that the decay of cooperation is due to noncommunication, not to mistake (e.g., Olivier Bochet, Talbot Page and Louis Putterman, Communication and Punishment in Voluntary Contribution Experiments, Journal of Economic Behavior and Organization, 2006; and an additional option to punish did not significantly increase cooperation). Game theory was conceived at the onset of the Cold War, and was carried to term in the womb of the Rand Corporation (S.M. Amadae, Rationalizing Capitalist Democracy, University of Chicago Press, 2003): the U.S. and the Soviet Union were locked together in an indefinitely repeated two-person prisoners dilemma game. That this is the exemplary game is wholly an accident of history: there is no formal or substantive
reason to assume that typical human interactions are 1) between only two agents, 2) who have consistent preferences, 3) possess complete information, 4), are disciplined by threat of retaliation, 5) and locked in indefinite embrace. There is no reason to expect that the predictions of a two-person model would resemble those of an n-person model, and most social interactions are n-person. States conducting diplomacy are more likely to be consistent than individual humans, who, it has been robustly shown, struggle constantly with time-inconsistent preferences (weakness of will). Game theorists so far have been unable to model complex games of incomplete information, as the author concedes (198). The retaliatory punishments required by the model are rarely observed in real settings. Many social interactions allow for exit, rather than indefinite embrace. It is notorious that the predictions of game-theoretic models can change radically upon small changes in assumptions, and that a model can be jiggered so as to predict any pattern of behavior. Thus it is not fitting that any particular model be promiscuously generalized, and it is not fitting to conjure facts from the assumptions of one s model. For example, instead of the perpetual retaliation of Cold War, what if our game were Out-for-Tat, which allows players to cooperate, defect, or exit the relationship? Then, within certain parameters, it pays to follow the rule: initiate cooperation, if cooperation is reciprocated then stay with relationship, and if partner defects then exit relationship and randomly choose new partner (e.g., Michael Macy and John Skvoretz, The Evolution of Trust and Cooperation Between Strangers, American Sociological Review, 63:638-660, 1998). This is a more descriptively plausible model of trust in strangers, I submit. Formal models blind as much as they illuminate. Enthralled by Cold-War, the author assumes that our hunter-gatherer ancestors enforced cooperation by continuous group supervision (41). Cold-War better describes the settled agricultural village, however. Out-for-Tat better describes the circumstances of wandering and fluidly associating and reassociating hunter-gatherers, and better explains food-sharing and absence of bosses among them: for any individual, if she didn t love it, she could just leave it. The author also offers an anecdote on behalf of Cold-War, which instead supports Out-for-Tat: Sure I trust [the honesty of an antique-store owner]. You know the ones to trust in this business. The ones who betray you, bye-bye (88). Moreover, contrary to the author, there is no reason to expect cultural evolution to resemble biological evolution. The units of analysis, sources of variation, avenues of transmission, and mechanisms of retention are radically different in each sphere. It is difficult to believe that each culture has attained the best of all possible worlds for itself (for evidence, see Robert Edgerton, Sick Societies: Challenging the Myth of Primitive Harmony, Free Press, 1992). The prediction that every inherited cultural practice is Pareto-efficient is painfully falsified by daily experience. An attraction of more modest game theory is that models of coordination, imitation, or punishment can be formed which might account for observations of harmful cultural practices. Finally, although it is legitimate to attempt explanation of the distribution of beliefs and desires in a society, meme theory is a pseudoscientific enterprise, and is additionally mistaken in denying the important place of reason in human affairs.
The author argues, in the spirit of Hume, that morality is merely conventional, that the good and the right is defined as coordination on one equilibrium in one culture and coordination on another equilibrium in another culture. I shall argue that his cultural relativism is based on several confusions. Game theorists explicate conventions with a simple coordination game, illustrated by the story of which side of the road to drive on. It doesn t matter whether we drive on the left or on the right just so long as everyone drives on the same side. Define the following states of the world, and assign the corresponding utilities: Self drives left, other drives left (1); self drives left, other drives right (0); self drives right, other drives left (0); self drives right, other drives right (1); there are two stable coordination equilibria: all drive left (1,1) or all drive right (1,1). Similarly, it doesn t matter whether we call the perplexing furry creature a cat or un gatto, just as long we all use the same word. We are indifferent between certain conventions, which misleads some to assert that conventions are arbitrary. Underlying the supposedly arbitrary convention, however, are nonarbitrary value-judgments: that automobile collision is worse than its avoidance, that communication failure is worse than communication success, that miscoordination is worse than coordination (in other circumstances, miscoordination would be better than coordination, for example, to search for firewood it is better that one of us goes north and that the other goes south). The underlying judgments seem almost universal, almost objective. Further, sometimes one convention is better than alternative conventions. Suppose that our country drives on the left and surrounding countries drive on the right, and increasing international traffic is causing fatal accidents. In these circumstances, rankings change as follows: Self drives left, other drives left (1); self drives left, other drives right (0); self drives right, other drives left (0); self drives right, other drives right (2); there are two stable coordination equilibria: the inferior convention of all drive left (1,1) and the superior convention of all drive right (2,2). Similarly, a 25-syllable word is cross-culturally inferior to a one- or two-syllable word for our furry friend, here we are no longer indifferent between naming conventions. To define the good and the right as an equilibrium in a coordination game is woefully insufficient, and thus fails. It neglects that the social equilibrium is based on prior individual valuations that one state is better than another; neglects that one equilibrium can be judged better than an alternative equilibrium, either within or between cultures; and it neglects to explain why any individual does favor or should favor one particular state over another. The author often asks us to receive charitably his speculations (ix) and inconsistent (53) models, but he is not ready to reciprocate such charity to the thinkers he claims to supersede. Other than Binmore s improvements on Hume and snippets of Hobbes, Nietzsche, Harsanyi, and Rawls, the entirety of moral and political philosophy over thousands of years is bunk. They are all just expressing their personal opinions, and this is no less true of the pundits who claim privileged access to the wisdom of the ages than of a drunk sounding off in a neighborhood bar (50). Rousseau and Kant are neurotic oddities (52). Kant, Marx, and Hitler are denounced in a single breath, as creatures of superstition (56). Philosophers and political scientists are no better than the witch doctors of darkest Africa : Instead of Mumbo-Jumbo being personified as a god, he is reified as some abstract Platonic form (44). The folk theorem of game theory, however, emerges as Binmore s Platonic form. In commenting on the weak fit to reality of one
nice mathematical model, he says that, The real world is imperfect in many ways (72). For his descriptive purposes, however, it is the models that are imperfect, not the realities. He claims to be an empiricist, but there are as few actual facts in Natural Justice as there are in Plato s Republic, and far fewer arguments. Argument is scarce, perhaps because of the author s view that, I have no source of moral authority at all.i know perfectly well that my aspirations for what seems a better society are just accidents of my personal history and that of the culture in which I grew up (19). I was raised differently. Gerry Mackie, University of California, San Diego