Review of Morals by Agreement

by David Gauthier
Clarendon Press, Oxford University Press, 1986
367 pages; ISBN: 0-19-824992-6
reviewed by Eric Watt Forste

Originally published in Extropy: The Journal of Transhumanist Thought, Issue 14.

Without a God, is everything permitted? Why not do anything you want? Why accept any limits on our behavior? After all, extropians reject unnecessary limits on our actions. In Paul M. Churchland’s essay “Moral Facts and Moral Knowledge” [1], he discusses moral theories as having a structure in the brain similar to scientific theories (though clearly not based on the same kind of empirical evidence), and discusses paradigms and paradigm-shifts in the context of moral theories. The moral paradigm that human beings have lived under for almost all of history has been the stern parent paradigm; in the particular case of the Judeo-Christian tradition, Dad gives the orders.

But most extropians have rejected this paradigm. Now that we’ve grown up and no longer take orders from Dad, is morality a necessary and reasonable limit, or an unnecessary one which we transhumans can dispense with? Philosophers who have attempted to provide a moral theory that does not reduce to divine commandment have found it frustrating. As Bertrand Russell put it: “I cannot see how to refute the arguments for the subjectivity of ethical values, but I find myself incapable of believing that all that is wrong with wanton cruelty is that I don’t like it.” [2]

There are two traditional sources in which people have sought to find a basis for rational constraint on individual behavior. David Hume and many others since his time sought to find it in our “fellow feeling” for one another, but this certainly doesn’t address the puzzlement expressed by Russell above. Others, such as Kant and Rand, have sought to locate rational constraints in some objective good that exists independently of vagaries in people’s individual preferences. Utilitarianism [3] attempts to reconcile these two approaches, but has a long history full of frustrations, and the problems of utilitarianism are very well documented by now. The new approach of moral philosophers to the search for rational constraints on individual behavior is contractarianism, borrowed from political philosophy. The good news of the contractarian is that we can find such constraints in the rational agreement to certain predispositions, such as a predisposition to keep one’s promises, that will be reached among rational individuals in hypothetical negotiation. The contractarian’s claim, which here is a claim about moral constraints and not political power, does not require an actual negotiation of an actual social contract, but only an explanation of how it is in our interest to adopt, for instance, a mutual predisposition to keep promises.

The quickest way to sum up David Gauthier’s Morals by Agreement in terms that will make its appeal evident to extropians is to say that it develops a theory which explains moral constraints as a spontaneous order arising from rational utility-maximizing behavior. Gauthier operates within the context developed most recently by John Rawls [4] and Robert Nozick [5]. Like Rawls, he develops a rigorous and systematic theory, but like Nozick, he rejects the drastic redistributionism demanded by Rawls. The theory of morals by agreement is an ambitious one that attempts to answer rigorously such questions as “What rational motivation can I have for not taking another’s property, when I can get away with it? What rational motivation can I have for not going back on my word, when it seems advantageous for me to do so?”

Morals by Agreement is unusually pleasant going by the standards of this genre, but the reader should be warned. The going will be much easier if you familiarize yourself with the basic terms and ideas of economics and game theory before you start. Personally, I got a lot more out of Gauthier’s book by pausing to read The Compleat Strategyst [6], a simple introduction to game theory, before trying to finish it. Axelrod’s explanations of game theory [7] would probably do as well. On the other hand, Gauthier does explain these terms before he uses them, but he does so very tersely, as befits background material, so these prefatory remarks are among the most difficult parts of the book.

Gauthier starts by drawing a connection between the suboptimal outcome of the famous Prisoner’s Dilemma of game theory, and the suboptimal outcome of a market afflicted by externalities. What does this mean? Game-theoretical analyses of interactions between people, with outcomes measured according to the utilities received by the people interacting, leads us to one conception of a “good” outcome. Market economics leads us to a different conception of a “good” outcome. The first is equilibrium: the outcome “which maximizes the expected utility of the person choosing it, given the strategies chosen by other persons.” [8] The second is optimality, usually called Pareto-optimality: the outcome in which no one can be made better off without making someone else worse off. Game-theorists look for equilibrium solutions. Economists tend to look for optimal ones. Now in many, many interesting interactions, there seems to be no difference between equilibrium and optimality. In fact, the basis of the satisfactory results derived from free markets is that market interaction exemplifies optimal equilibrium.

Our problems (and the problems Gauthier is interested in analyzing) begin when equilibrium and optimality lead to two separate outcomes. The two classic examples of this are, in game theory, the Prisoner’s Dilemma, and in economics, the problem of externalities. Essentially, straightforward maximizers of utility will always choose the action which yields them the greatest expected utility, given their expectations about the actions of the people they are interacting with, under the assumption that those people in turn are also straightforward maximizers. The Prisoner’s Dilemma and the problem of market externalities are situations in which interaction of straightforward maximizers leads to suboptimal outcomes. Recall that in a suboptimal outcome, there are actions which could be taken which would increase the utility of some or all of the participants without making any individual participant any worse off, but the “rational” constraints of straightforward maximization prevent people from taking those actions.

Gauthier’s solution to this problem begins when he points out that we can make choices on multiple levels. At the level of each individual choice, it may make sense to be a straightforward maximizer. But we can also choose to adopt certain “dispositions to choose”; we can make choices at a higher level than the level of individual decisions. Although it is rational for us, in any given situation, to choose the course of straightforward maximization, might it not be possible that, at the higher level of choosing a particular disposition-to-choose, it is rational for us to voluntarily choose to constrain our behavior in a way that makes it possible for us to attain optimal outcomes? Optimal outcomes are, on average, of greater utility to us than equilibrium outcomes, since in any given optimal outcome we will never be any worse off than we are at equilibrium, and will sometimes be better off.

Gauthier calls this optimality-enabling disposition to choose “constrained maximization” and contrasts it with straightforward maximization throughout the work. He is careful to contrast his conception of constrained maximization with rational prisoner’s-dilemma strategies such as Robert Axelrod’s TIT-FOR-TAT [9], pointing out that constrained maximization, as a disposition to choose in certain ways, comes into play even when interacting with people in one-shot games, whereas a straightforward maximizer (and Axelrod never discusses any other kind) finds TIT-FOR-TAT rational only in iterated interactions. In this way, constrained maximization seems to find a compromise between simple TIT-FOR-TAT (which is no help in one-shots) and Douglas Hofstadter’s “superrationality” [10] (which contradicts most people’s intuitions about the one-shot prisoner’s dilemma). For, unlike “supperrationality”, the rationality of constrained maximization is not completely insensitive to the identity of the other player(s). It is rational for us to follow a policy of constrained maximization only when we are interacting with people whom we believe to be other constrained maximizers. Given that belief, and good grounds for it, it truly does not matter whether the interaction is iterated or one-shot: when the optimal outcome does not coincide with the equilibrium, it is still the preferable outcome, and constrained maximization is the only course which allows for the optimal outcome instead of the equilibrium.

This raises an immediate practical question for those of us interested in thinking about the problems that will arise among posthumans. The reliability of our current assessments of the morality of people we are dealing with depends on the difficulty of lying convincingly about ones intentions over a long period of face-to-face contact. There are subtle clues of insincerity that most of us are capable of detecting; we have instinctive “hunches” about who is and is not trustworthy, hunches that are accurate more often than not. Without this ability to detect the dispositions of others, our ability to choose the correct disposition and strategy in prisoner’s-dilemma situations is impaired. What happens when we are posthumans dealing with other posthumans, each of us with complete control over every bit of information that trickles out from us? What happens when we (and the people we deal with) have complete override control over the subtle cues of untrustworthiness? The results of Gauthier’s theory indicate that we might find some advantage not only in not masking our intentions, but finding some way to guarantee that we are not masking. We might need a forward-thinking cryptologist to develop a cryptographic protocol for “expression of true intentions” to save us from having to settle for non-cooperative, equilibrium outcomes instead of reaping the benefits of cooperative, optimal ones.

So what does constrained maximization amount to? In large part, it seems to consist of such bedrock ethical principles as not lying, keeping ones promises, and in general, not cheating. Some libertarians such as Jan Narveson [11], have been excited by the political implications of Gauthier’s development of a theory of property and fair distribution from his moral theory. Gauthier cannot find any philosophical ground in his own work for the levelling, equality-at-any-price redistribution that Rawls sought to justify in his A Theory of Justice [12]. In fact, the theory of property he develops is more than superficially similar to Locke’s theory as defended by Nozick in his notorious response to Rawls, Anarchy, State, and Utopia [13]. While Nozick was quick to point out problems in Rawls’ theory, he did not develop a coherent theory of his own. Narveson and other libertarians are excited by the possibilities of Gauthier’s moral theory as providing a pretty good foundation for libertarian political theory.

Gauthier himself doesn’t seem to be any kind of fervent libertarian, though, and libertarians who eagerly delve into his work may find a few puzzling dissonances. While most libertarians see the action of the market as a case of cooperation between buyer and seller, Gauthier is more interested in the contrast between cooperation (the theory of which he is trying to develop) and market competition. His emphasis on the contrast leads him to characterize the market as a “morally free zone” throughout the book. What is curious about this point of view is that, while indeed competition in a market without externalities should not be constrained by any concerns other than consumer demand, it is cooperation (even in Gauthier’s technical sense) between buyer and seller that enables market transactions in all but the simplest case. The universality of credit or futures-delivery in wholesale markets exemplifies this.

In discussing the market, Gauthier seems to be so impressed with how the system functions despite the intense competition between seller and seller (and between buyer and buyer) that he overlooks the cooperation that is both routine and necessary for the market’s existence. Fortunately, although his characterization of the market as a “morally free zone” is strong and reiterated, and used to good rhetorical effect throughout the work, Gauthier does not actually rely on it in developing the foundations of his theory of morals by agreement. He simply relies on the existence of market failures to provide room and relevance for his moral theory, and his failure to recognize that even a market free of externalities requires a moral theory does no harm to his foundational argument.

It’s easy to come up with a plethora of questions about a theory as ambitious as this, and Gauthier, a good self-critical thinker, spends much of the book trying to formulate and answer as many of them as possible. He elaborates about what exactly his theory does and does not do. He deals with the question “Once I’ve adopted and displayed a predisposition to be a constrained maximizer, what it is that makes is rational for me a actually choose according to that “predisposition” when I find myself in the heat of the crunch?”. Needless to say, a lot of this gets bogged down in technical details. In a “Note to the Reader” at the beginning of the book, Gauthier helpfully points out the more abstruse sections so we can skip them on the first reading, outlining the remaining two-thirds of the book which deal with the theory rather than with details of its elaboration. Given the interest extropians have in self-determination and freedom from limits, eventually each of us must come face to face with each of the moral principles we’ve learned over the course of our lives, and make conscious choices in each case to adopt (or reject) that principle on the basis of our own autonomous decisions. The idea of morals by agreement gives us a background of thought that can help us try to apply the principles of rationality we use in our everyday lives to the profound ethical choices that serve, in some ways, as the anchors of our identity.


[1] Paul M. Churchland, A Neurocomputational Perspective, Cambridge: MIT Press, 1989.

[2] “Notes on Philosophy, January 1960”, Philosophy, April 1960, as quoted in The Retreat to Commitment, 2nd ed., p. xxv (Preface)

[3] Utilitarianism in its most familiar formulation is the idea that we should act so as to produce the greatest happiness in the greatest number of people. Among the more obvious problems of utilitarianism is the definition of happiness (in which the schism between the hedonist and non-hedonist varieties of utilitarianism finds its origin), the incommensurability of the happiness of different individuals (how do we choose between making two people somewhat happy or making one person extremely happy?), and the schism between rule-based and outcome-based utilitarianism. To put it mildly, this catalog of problems is not exhaustive.

[4] John Rawls, A Theory of Justice, Cambridge, Mass.: The Belknap Press of Harvard University Press, 1971.

[5] Robert Nozick, Anarchy, State, and Utopia, New York: Basic Books, 1974.

[6] J. D. Williams, The Compleat Strategyst, New York: McGraw-Hill, 1954.

[7] Robert Axelrod, The Evolution of Cooperation, New York: Basic Books, 1984.

[8] David Gauthier, Morals by Agreement, Oxford: Clarendon Press, 1986, p. 65

[9] Axelrod, op. cit.

[10] Douglas Hofstadter, “Dilemmas for Superrational Thinkers, Leading Up to a Luring Lottery”, in Metamagical Themas, New York: Basic Books, 1985.

[11] Jan Narveson, The Libertarian Idea, Philadelphia: Temple University Press, 1988.

[12] Rawls, op. cit.

[13] Nozick, op. cit.