I. Interpersonal Justification: Second- and Third-Personal

I shall understand second- and third-personal views of justification more narrowly than some thinkers do, as views of how claims about distributive justice are to be justified.^{Footnote 5} More specifically, I take them to be views about where considerations that count in favor of claims of distributive justice get their authority or force as practical reasons, as reasons for acting.^{Footnote 6} In her explanation of third-personal justification, Anderson uses the example of a “policy conclusion” bearing on distributive justice. “In a third person justification,” she says, “someone presents a body of normative and factual premises as grounds for a policy conclusion. If the argument is valid and the premises true, then the conclusion is justified.”^{Footnote 7} Thus, according to the third-person conception of justification as Anderson explains it, justification is sound argument. An example of such an argument would be one that relied upon a true premise about the efficacy of a policy in promoting some end which another true premise asserted to be unconditionally desirable, and that moved impeccably from those premises to the conclusion that the policy itself is desirable. The argument would then provide a reason to favor that policy the force of which is derived entirely from the argument’s soundness.

According to second-personal justification, by contrast, considerations that count in favor of claims of justice get their force, or some of their force, from authority relationships in which parties to the justification stand. To see this, it is helpful to turn from claims of distributive justice to claims that someone in a familiar authoritative role can make on a subject’s will. A military officer, for example, can issue a command to an enlisted person that provides the enlisted person conclusive reason to act but that would be void of any force at all if issued to a civilian. The officer’s command does not get its reason-giving force, its normative authority, from its being the conclusion of a sound argument that compliance with the order would promote a desirable end. It gets it from the authority relation that holds between the officer and the enlisted person. Indeed, the ability to issue commands with such force is partially constitutive of that relation. The source of the command’s force is appropriately described as “second-personal” because the command has reason-giving force only when and because it is addressed by the person in authority to a person subject to her. Justification of the officer’s claim to obedience would appeal to the relationship between officer and enlistee, and to the authority that relationship confers on the former to address commands to the latter.

According to Anderson, relational egalitarians think the justification of claims of justice also depends upon reasons the force of which is grounded in an authority relationship — in this case, in the relationship of equal reciprocal authority that holds among persons who owe one another justice.^{Footnote 8} The idea is as follows. To be owed justice is to be owed it by an agent, either natural or corporate. To be owed something by an agent is to be authorized to demand it and to be authorized to hold the agent accountable for doing or providing it. And to be authorized to hold an agent accountable entails the authority or standing to demand that the agent provide a justification of conduct. Agents who owe one another distributive justice have equal authority or standing of that sort with respect to each other. To press a claim of justice, on this view, just is to demand justification and, if appropriate, redress. The claims parties who owe another justice make on one another count as reasons to provide justification and redress because of the standing those parties have, because they stand in relationships of mutual authority. It is because those relationships authorize parties to address claims to one another that their relationships are ones of second-personal authority and that the reason-giving force of their claims is second-personal.

The justifications offered in response to claims or demands will appeal to considerations that purport to tell in favor of distributive outcomes. The most fundamental of those considerations are the first principles of justice. Are these principles themselves justified second-personally? To answer this question, we need to see how the relational egalitarian identifies those principles.

Principles of distributive justice say how persons with mutual authority are — as such — to relate to one another, both when they distribute and when they justify distributions. That is why the principles are principles of relational egalitarianism. A sufficient condition of persons’ relating to one another as equally and mutually authoritative is, it is said,^{Footnote 9} that their treatment of one another be justifiable by principles which all of them could accept — or, in T. M. Scanlon’s view, what principles they could not reasonably reject^{Footnote 10} — to play that justificatory role. This is a contractualist condition. And so relational egalitarians typically use a social contract to identify principles of distributive justice. They are typically contractualists.

What is the source of the principles’ reason-giving force? Principles are supposed to have their force or authority because of their adoption in a social contract or their survival of a “reasonable rejection” test. Since the rationale for requiring that principles be adopted in a contract or subjected to a reasonable rejection test is ultimately that those whose relations will be regulated by the principles have mutual second-personal authority, the reason-giving force of the principles is said ultimately to be founded on that authority relationship. Their justification and force are therefore said ultimately to be second-personal.^{Footnote 11} My own view is that the argument for the second-personal authority of principles of justice is controversial at several points.^{Footnote 12} But I shall overlook those difficulties here and grant, for purposes of exposition, that the argument goes through. What matters for those purposes are the ways in which third- and second-personal views of justification differ, and the reasons that relational egalitarians — who are said to hold the latter view — embrace contractualism.^{Footnote 13}

As we have just seen, relational egalitarians think principles of justice play a central role in an important social and political practice: the practice in which mutually authoritative persons justify distributions to one another.^{Footnote 14} That, we might say, is what they think principles of justice are for.^{Footnote 15} Principles of justice are justified by their suitability to play that role. Seeing this, we can say with somewhat more precision what it means to claim — as we saw Anderson does — that a concern with stability is “internal” to contractualism: we can think of contractualism as imposing an acceptability condition on principles according to which principles would be accepted in a contract only if they are able to play that role over time, so that the practice of justification is stable. Since the principles are the object of an agreement by those who engage in the practice, we can re-express this condition by saying that parties are to finalize their agreement on principles only if they have good reason to believe that the agreement they reach can be stable.^{Footnote 16}

This acceptability condition may well affect the content of the principles that would be adopted.^{Footnote 17} If it does, then the viability of a contractualism which imposes the condition depends upon there being principles which both match our considered convictions about what justice demands and can be the object of a stable agreement. I want to ask what claims must be true if there are such principles and whether those claims are internal to contractualism in any robust sense of “internal.” To pursue these questions, it will help to look into the conditions of stable agreements. To do that, it will help to return to the distinction between a third-personal conception of justification on the one hand, and contractualism and relational egalitarianism on the other.

That distinction is sometimes drawn so as to suggest that third-personal justifications cannot support principles that play the interpersonal role I have said principles of justice play in relational egalitarianism.^{Footnote 18} But I believe that this suggestion depends upon restrictive assertions about the content of premises on which third-personal justifications rely, assertions that do not follow from the notion of third-personal justification itself.^{Footnote 19} To show that it is possible to combine a third-personal conception of justification of principles of justice with the view that equality is relational, it suffices to show that a thinker has combined them without inconsistency. In A Theory of Justice, Rawls seems to have done exactly that. He says of the argument from the original position that it “aims eventually to be strictly deductive. . . . We should strive for a kind of moral geometry with all the rigor which this name connotes.”^{Footnote 20} This sounds like an aspiration to justify principles by an argument the force of which depends entirely upon its soundness, hence to provide them a third-personal justification. Yet as Anderson correctly notes,^{Footnote 21} Rawls is a relational egalitarian, and one of the points of relational egalitarian principles is to place relations among citizens on a footing of democratic equality.^{Footnote 22}

Some of the axioms of Rawlsian moral geometry would presumably concern the free and equal status of persons, the importance of identifying principles that are mutually justifiable to persons so conceived, and the reasons the original position is the appropriate means of identifying such principles. And so the contents of the axioms might themselves be thought to show that the Rawls who aspired to offer arguments that are geometric in their rigor still had a fundamentally second-personal or interpersonal conception of justification. But even if that is so, it does not imply that the Rawls of A Theory of Justice did not have a third-personal ideal of justification as well. It might instead be taken to show that contractualist and third-personal conceptions of justification are not mutually exclusive so that, as I have said, it is possible to offer both kinds of justification for the same principles.

Rawls’s attraction to a third-personal ideal of justification is not just suggested by his express hope that the principles of justice could be established more geometrico. It is also suggested by his hope that they could be established by proofs akin to those in the theory of price^{Footnote 23} and the theory of games.^{Footnote 24} The proofs to which he refers establish certain combinations of actions as equilibrium points. Rawls’s reference to price theory is more extended, but one of his turns of phrase suggests that he thought of common acceptance of the principles as a game-theoretic equilibrium — more specifically, as a Nash equilibrium — among the parties in the original position. For a Nash equilibrium is a strategy-combination in which each player’s strategy is her best response to the strategy played by the others. And speaking of parties in the original position, Rawls says “I should like to show that these principles are everyone’s best reply, so to speak, to the corresponding demands of the others.”^{Footnote 25} Language he uses elsewhere suggests he thinks that principles would be supported by a Nash equilibrium outside the original position. For speaking of members of the well-ordered society, he says “The plan of life which [calls for preserving the sense of justice] is [each person’s] best reply to the similar plans of his associates.”^{Footnote 26}

Thus while the Rawls of A Theory of Justice provides a contractualist justification of his principles, I believe he also hoped to provide a third-personal justification.^{Footnote 27} The passages in which he expressed that hope show that he thought his principles define an equilibrium point which the social contract helps to identify. Seeing that the social contract is a device for identifying an equilibrium point suggests conditions under which agreements are stable.

II. An Argument Connecting Justice and Stability

To say that the agreement reached in a social contract would be stable is to say that the terms of the agreement would be generally honored by parties to it, but there are at least two ways in which the agreement can be maintained. The agreement could be what Rawls called “inherently stab[le]”: when an agreement on principles of justice is inherently stable the agreement is to be stabilized by the uncoerced activity of the members of that society, who act from the principles for their own sake. Its stability is therefore to be contrasted with stability that depends upon an outside force, such as a Hobbesian sovereign.^{Footnote 28}

Since Anderson quite understandably takes Rawls’s theory as paradigmatically contractualist, I shall assume that the kind of stability that she claims is internal to contractualism is inherent stability. I take her argument for the claim that it is internal to run roughly as follows.

Anderson says contractualism is the view that:

1. (1) “Principles of justice are whatever principles free, equal, and reasonable people would adopt to regulate the claims they make on one another.”^{Footnote 29}

I believe she thinks that

1. (2) Free, equal and reasonable people would agree to principles to regulate the claims they make on one another only if those principles do not impose unreasonable demands on any of the parties to the agreement.

According to Anderson, contractualists also think that if the principles could not be universally complied with for their own sake, even if all parties to the agreement were reasonable — if some reasonable parties to the agreement had to be forced to comply with it — then that shows that the agreement imposes unreasonable demands on some of them.^{Footnote 30} So contractualists think that:

1. (3) If an agreement on some principles to regulate claims could not be inherently stable, then those principles impose unreasonable demands on at least one party to the agreement.

Step (4) is equivalent, by contraposition, to:

1. (4) If principles to regulate claims made on one another do not impose unreasonable demands on any of the parties to the agreement, then an agreement on those principles can be inherently stable.

Steps (2) and (4) imply:

1. (5) Free, equal, and reasonable people would agree to principles to regulate the claims they make on one another only if an agreement on those principles can be inherently stable.

Steps (1) and (5) imply that:

1. (6) Principles that free, equal, and reasonable people would adopt to regulate claims made on one another are principles of justice only if an agreement on those principles can be inherently stable.

Because she thinks contractualists move from their defining commitment (1) to (6) by steps they accept, Anderson concludes that:

If I have interpreted Anderson correctly, then the crucial step in the argument for C is the claim that contractualists accept (3). Since (3) is equivalent to (4), we can ask what must be assumed by contractualists who think that inherently stable agreements are possible by asking whether contractualists accept (4) and, if they do, why. To sharpen these questions, it will be helpful to have a clearer idea of what is meant by saying that an agreement is “inherently stable,” when demands are and are not “reasonable,” and under what conditions terms which are reasonable “can” be inherently stable.

III. Self-Enforcement and Inherent Stability: A Question about the Fourth Step

The phrase “inherently stable” is not, to my knowledge, a technical term used in the formal study of equilibria; instead it is a Rawlsian term of art. I said an inherently stable agreement is one that is maintained by the uncoerced activity of members rather than by a third-party enforcer such as a Hobbesian sovereign. Agreements that are stabilized without third-party enforcement are sometimes described as “self-enforcing.”^{Footnote 31} So we can recast (4) as:

• (4') If principles to regulate claims made on one another do not impose unreasonable demands on any of the parties to the agreement, then an agreement on those principles can be self-enforcing.

The connection of self-enforcing agreements with “inherently stable” ones suggests why those agreements do not need to be enforced by a third party. It suggests that these agreements remain stable because stability inheres in the agreements themselves or, more precisely, in their terms. This happens when the terms of an agreement define an equilibrium of some kind — as we saw that Rawls thought his principles would — for then no one has sufficient incentive to deviate. Indeed, in the game theoretic literature, self-enforcing agreements are often characterized simply by reference to the structure of their payoffs, rather than as agreements that do not require third-party enforcement.^{Footnote 32}

Once we understand self-enforcing agreements in this way, we can see why it is generally thought^{Footnote 33} that:

Suppose that the terms of an agreement define a Nash equilibrium just in case those terms do not impose unreasonable burdens on anyone. Then (N) would imply that:

Step (4′) would then follow immediately, since an agreement which is self-enforcing can be self-enforcing, and we could see why contractualists accept it.

This line of argument for (4′) will not work because the supposition on the basis of which the argument moves from (N) to (4′) is false. Some equilibria may entail that at least one party to the agreement plays a strategy which imposes unreasonable demands on her. Rawls recognizes as much, for shortly after his remark implying that the principles chosen in the original position will define a Nash equilibrium he says “the fact that a situation is one of equilibrium, even a stable one, does not entail that it is right or just.”^{Footnote 34} But the failure of the argument from (N) to (4′) suggests another and more promising argument. Suppose it could be shown that:

Then, again, (4′) would follow immediately and, again, we could see why contractualists accept it. Indeed, it may be thought that Rawls arrives at (4′) in precisely this way. For shortly after he implies that the terms of agreement reached in the original position will define a Nash equilibrium, he says

Can (N⁺) be shown? That depends upon what further conditions a Nash equilibrium must meet if the strategy combination that is in equilibrium is not to impose unreasonable burdens on any of the players.

There may be a number of solution concepts that strengthen the demands of a Nash equilibrium by adding “no unreasonable burdens” conditions. One that seems especially promising is the concept of a “payoff-dominant Nash equilibrium.” A payoff-dominant Nash equilibrium is a Nash equilibrium in which every player receives a higher payoff than she would have had she played any other strategy open to her.^{Footnote 36} It may seem clear that an equilibrium satisfies that condition — an equilibrium in which each player achieves the most she could get — does not impose an unreasonable burden on any of the players. Moreover, it may seem especially plausible that an agreement on terms defining a payoff-dominant Nash equilibrium will be self-enforcing since no one would have a sufficient reason to defect. So interpreting the phrase “a Nash equilibrium that does impose unreasonable burdens on any of the parties to the agreement” in (N⁺) as “a payoff-dominant Nash equilibrium” seems to be a sufficiently strong construal of the phrase to be plausible. It also seems strong enough to make (N⁺) true if any construal of it can.

Unfortunately, Robert Aumann has shown — on grounds we shall look at in a moment — that so construed, (N⁺) is false, since not all agreements to play payoff-dominant Nash equilibria are self-enforcing.^{Footnote 37} The argument to (4′) from (N⁺) so construed therefore fails. But Aumann also argues that agreements to play payoff-dominant Nash equilibria can be self-enforcing. This argument offers the contractualist a different way of defending (4′), since (4′) follows from the variant of (N⁺) that Aumann has shown to be true — the variant that results from replacing “is” in the consequent of (N⁺) with “can be.”

IV. Self-Enforcement and Inherent Stability: A Rawlsian Defense of the Fourth Step

Aumann’s argument is important, for seeing under what conditions payoff-dominant Nash equilibria can be self-enforcing brings to light some of the contractualist’s reasons for endorsing (4′).

Very roughly, Aumann argues that for an agreement to play a payoff-dominant Nash equilibrium to be a self-enforcing agreement, the fact of agreement must be what brings it about that players play the set of strategies that are in equilibrium. If they defect from an agreement after reaching it, then the strategies were not played and the agreement was obviously not self-enforcing. If they cooperate after reaching the agreement but would have cooperated anyway, the agreement was superfluous. There are, Aumann says, two kinds of difference an agreement can make, hence two kinds of cases that support (4′) by establishing the variant of (N⁺). The agreement can change the information available to the players, as when a pre-play agreement enables those playing “Battle of the Sexes” to coordinate their strategies. Or the agreement can change the payoffs, so that the strategy combination that is in equilibrium becomes the one that all the players favor where it was not before.

Aumann says little about the second kind of case, but he presumably has in mind cases in which an agreement makes a strategy combination a payoff-dominant Nash equilibrium. Recall that I introduced the concept of payoff-dominance to make the antecedent of (N⁺) precise by identifying a kind of equilibrium that does not impose unreasonable burdens on any of the players. Some contractualists will think a weaker solution concept suffices. They will want to show that agreements on terms that are in their favored kinds of equilibrium can be self-enforcing, for this will enable them to infer (4′). Aumann’s second kind of case suggests how they might show that, and hence why they might endorse (4′). They can show it by showing that an agreement on terms that define their favored kind of equilibrium can change the parties’ payoffs, so that ex post the parties all have sufficient incentives to abide by the terms of the agreement.

There are at least two ways an agreement can make such a change. One way is for the agreement to attach new penalties to defection from strategies that are in the contractualist’s favored equilibrium. But if those penalties must be enforced by a third party, then the agreement concerns cooperative rather than noncooperative games and is, in any case, not self-enforcing after all. Another way an agreement can change the payoffs players attach to their strategies is, crudely put, by changing the players. Less crudely, it will make such a change if, as a result of having reached the agreement, players all become the kind of persons who prefer strategies that are in the contractualist’s favored kind of equilibrium to those that are not. In that case, the agreement is self-enforcing. Provided the terms of the agreement do not unreasonably burden any of the parties, the possibility that an agreement can change payoffs in this way establishes (4′) and shows why contractualists might endorse it.

Rawls thinks that the agreement reached in the original position changes parties’ payoffs in the second way.^{Footnote 38} For the agreement reached there is an agreement on principles of justice to regulate society’s basic institutions. Rawls argues that citizens who live under just institutions will acquire a sense of justice, which is a regulative desire to act from the principles of justice for their own sake. He also thinks that just institutions shape citizens’ views of their own good by affecting the payoffs they attach to the strategies open to them. More specifically, they shape parties to the agreement into the kind of person who “recogni[zes] that advantages gained in conflict with [the principles agreed to] are without value.”^{Footnote 39} As a result, no matter how much their views of the good may differ in other ways, these views are all “congruent” with justice^{Footnote 40} and none of them “is moved by rational considerations of good not to honor what justice requires.”^{Footnote 41} Of course assurance problems still must be overcome: members of the well-ordered society will actually do what justice requires only if they have the assurance that others, or sufficiently many others, are themselves committed to honoring the principles in political life. Just institutions would therefore encourage citizens to provide that assurance by complying with the guidelines of public reason when debating and voting on fundamental questions.^{Footnote 42}

Because the terms of the agreement reached in the original position can educate members of the well-ordered society in these ways when those terms are institutionalized and publicized, the agreement can make a difference in the second of the two ways Aumann says an agreement can make a difference. It is because the terms can make a difference in that way that Rawls implies the stability of the agreement inheres in them and why he says that the agreement reached in the original position could be “inherently stable.”^{Footnote 43} It is also why he thinks that the agreement reached in the original position can “[generate] its own support,”^{Footnote 44} why he would think that that agreement could be self-enforcing, and why he would endorse (4′).

Suppose that other contractualists endorse (4′) on the same grounds that Rawls does.^{Footnote 45} I have also supposed that they accept:

1. (1) “Principles of justice are whatever principles free, equal, and reasonable people would adopt to regulate the claims they make on one another.”

and

1. (2) Free, equal, and reasonable people would agree to principles to regulate the claims they make on one another only if those principles do not impose unreasonable demands on any of the parties to the agreement.

Given the way I have parsed “inherently stable,” (2) and (4′) imply:

1. (5) Free, equal, and reasonable people would agree to principles to regulate the claims they make on one another only if an agreement on those principles can be inherently stable.

And as we saw, (1) and (5) imply:

1. (6) Principles that free, equal, and reasonable people would adopt to regulate claims made on one another are principles of justice only if an agreement on those principles can be inherently stable.

The way the contractualist moves from (1) to (6) seemed to support the desired conclusion:

Now that we have seen that the argument for C relies on the grounds adduced for (4′), I want to ask about what additional assumptions are needed to support the conclusion.

V. Some Questions about the Rawlsian Defense

Much depends upon what is meant by saying that a claim is internal to philosophical view such as contractualism. I take contractualism to be a closely related family of views whose defining claims bear a strong family resemblance. Perhaps the defining claim of some versions of contractualism are theses about principles of right, and not just principles of justice; perhaps the defining claim of others is a conjunction of a thesis about moral wrongness and a thesis about moral motivation.^{Footnote 46} But if (1) is not a defining commitment of all forms of contractualism, it seems to be entailed by contractualism’s defining commitment. Since entailment by defining commitments seems sufficient for internality, I grant that (1) is internal to the view. Moreover, since (1) requires that contracting parties adopt principles freely and (2) seems to express a test on free adoption, I grant that (2) is internal to it as well. We saw earlier that the crucial step in the argument for C is (4), which I have recast as (4′). Is (4′) also internal to contractualism, and if so, in what sense “internal”?

The Rawlsian defense of (4′), which I have supposed for now that other contractualists endorse, appeals to psychological claims to show that we can become the kind of persons that institutions which satisfy the terms of social contract would encourage us to be. If those claims were internal to contractualism, then (4′) would seem to be also, and the argument for C would seem to go through. But these claims do not seem to follow from the claims that I have granted are internal, since (1) and (2) do not seem to entail any psychological claims at all. So it is not obvious that those claims are internal.

It may be replied that this criticism, even if sound, does not show that C itself is false, and that a concern for inherent stability is not internal to contractualism. At most it shows that one set of assumptions about how inherent stability is achieved is not internal to contractualism — in one strong sense of “internal.” What is internal to contractualism, it may be insisted, is the assertion that there are some true psychological claims or other that support (4′).

That assertion does not, however, seem to be entailed by contractualism either, since denying it does not seem to be inconsistent with affirming (1) and (2).

To see this, consider Rawls’s two principles. Let us suppose that in the original position, those principles are preferred to the principle of average utility on the grounds Rawls appeals to in his “ends in themselves” argument.^{Footnote 47} The parties in the original position represent free, equal, and reasonable persons. Rawls’s principles therefore satisfy the condition expressed by (1). Since I am supposing for the sake of argument that (1) is true, they are principles of justice. Moreover, the original position is a contract situation that embodies conditions we think it reasonable to impose upon arguments about principles of justice. Its conditions are such that parties to the agreement cannot avail themselves of information about their own resources or abilities to gain bargaining power that we intuitively regard as giving them an unfair advantage. Because the contract situation is fair, the terms of the agreement reached are fair: they do not impose unreasonable burdens on anyone, and our considered judgments seem to support this conclusion. So Rawls’s principles also satisfy (2).

But there does not seem to be any inconsistency in further supposing that the forces of social learning which are supposed to engender a sense of justice and to bring about congruence of the right and the good are not successful enough that the principles agreed to in the contract are regularly honored in society. Perhaps one of the principles that would be agreed to requires that material inequalities be quite narrow or even nonexistent. It seems possible that citizens develop some desire to act from that principle for its own sake, but that that desire will not be strong enough regularly to overcome citizens’ contravening desire to acquire more than the principle allows. And so while we might suppose the laws of human psychology to be such that parties to the agreement will develop a sense of justice, we might also hold that the agreement lacks the motivational power needed to stabilize agreement on the principles for the long run. It also seems possible that the laws governing formation of views of the good allow only for weak congruence. Citizens’ conceptions of their own good would then be sufficiently congruent with justice that they would like to be less materialistic than they are — indeed, they would like to be less materialistic because they recognize that the principles of justice do not impose unreasonable burdens on them — but they remain powerfully tempted to acquire more than justice allows. We may then suppose that those who benefit from permissible inequalities accede to temptation. They accumulate resources and use them to influence the political process, securing legislation that enables them to expand and perpetuate their gains well beyond what justice allows.

The argument for C could still be vindicated if the assertion that there are true —not necessarily Rawlsian — psychological claims which support (4′) were said to be internal to contractualism, not because the assertion is entailed by (1) and (2), but because it, along with (1) and (2), is one of contractualism’s signature commitments. I shall assume that this is how the argument is to be vindicated. In the next section, I shall draw out three implications of vindicating C this way.

VI. The Reach of Contractualism

First, the assertion I have supposed is among contractualism’s commitments concerns terms of agreement that do not impose unreasonable demands. The proof that some terms do not impose unreasonable demands cannot simply be that an agreement on them can be self-enforcing. Rather, if contractualism is to be plausible, contractualists must provide arguments for two different conclusions. One argument must show that terms of agreement do not impose demands that we intuitively regard as unreasonable, so that they satisfy (2). The other must show that an agreement on those terms can be self-enforcing, so that the terms satisfy (4′).^{Footnote 48} And so if the argument laid out above for C is sound, then one of the defining commitments of contractualism — which does not follow from its other defining commitments — is that there are true psychological claims which connect the reasonability of terms of agreement with the possibility of an agreement on those terms that is inherently stable. Thus I take contractualists to have an independent, if generally unarticulated, commitment to the claim that human beings can be shaped so that our payoff structure favors honoring rather than defecting from an agreement on principles which satisfy (2), at least when they know others face similar payoffs.

That psychological claim is about moral motivation and its development. But now note that contractualists could make any of a variety of assumptions about how moral motivation is elicited and sustained. In Section III, I tried to give some indication of the assumptions Rawls makes. There is a reading of Kant, most prominently associated with John Hare, according to which Kant made very different assumptions. Briefly put: on Hare’s reading, Kant thinks that there is a “moral gap” between the demands of principles of right and our ability reliably to act on them. He reads Kant as saying we can develop and sustain the motivation we need freely to act from the moral law only with the help of divine grace.^{Footnote 49}

My aim is not to defend the point Hare finds in Kant. Nor is it to defend Hare’s reading of Kant, which I lack the textual knowledge to do. My point, rather, is that the claim which Hare imputes to Kant is not inconsistent with contractualism as such. Contractualists could support (4′) by making Rawlsian psychological assumptions, but they could also support it by making the assumptions Hare finds in Kant.

If Rawlsian psychological assumptions were internal to contractualism, then perhaps there would be a good case to be made that they are internal to relational egalitarianism as well. But we need not take those claims as internal to contractualism. Not doing so opens the possibility that relational egalitarians could endorse the same variety of assumptions that are compatible with contractualism. Some make Rawlsian assumptions and argue that just institutions can sufficiently encourage effective motivations to honor the demands of relational egalitarianism. But it seems consistent with relational egalitarianism to deny that we can honor those demands and relate to one another as equals without the help of grace. This means that there can be — and if Hare is right, there may well have been — forms of relational egalitarianism that draw on very strong religious, rather than very strong naturalistic, assumptions. A history of relational egalitarianism would have to be written with this possibility in mind.

Second, I take it most contractualists endorse the secular form of contractualism, so that their defining commitments are (1), (2), and the assertion that there are true naturalistic psychological claims which support (4′). Since contractualists must make some such claims if they are to show that a just society is possible, such claims are of vital importance to the contractualist’s position. But how are these claims to be vindicated? Arguments might be brought forward to make a given set of psychological claims plausible, but it is hard to see how any such claims could be conclusively established. The disparity between their centrality to the contractualist’s position and the strength of the grounds that can be adduced for them suggests that the claims can be described as articles of the contractualist’s moral faith.^{Footnote 50}

In calling these claims articles of moral faith, I do not mean to imply that either they are untrue or that they have religious content. Nor do I mean to imply that they are unreasoned, since there can be defenses of reasonable faith. I am, rather, making a point about their importance to the contractualist moral position, their logical independence from the contractualist’s other defining commitments, and the fact that grounds for reasonable doubt about those commitments cannot be entirely removed.^{Footnote 51} If this characterization is right then — since one way to reject contractualism is to reject the assertion that (4′) can be supported — one way to reject it is to reject the contractualist’s moral faith.

There are at least two forms this rejection could take.

I said in Section III that Rawls’s argument for (4′) depends upon the claim that institutions which satisfy the terms of the agreement reached in the original position can effectively encourage citizens’ desire to act from those terms and can shape them so that they judge doing so to be congruent with their good. Since those terms are terms of relational egalitarianism, his argument depends upon the claim that institutions that comply with those terms can effectively encourage those who live under them to relate to one another as equals.

Someone who objects to Rawls’s version of contractualist liberal egalitarianism might grant that we can identify norms which the institutions and practices of very small societies must satisfy so that their members relate on a footing of equality. And she might grant that those institutions and practices can effectively encourage members of those societies to act from those norms. So she might grant that in the case of very small societies, reasonability and stability coincide. If the norms for such societies define what Rawls calls “the morality of association,”^{Footnote 52} then she grants that agreements on the morality of association can be inherently stable. The objector I am now imagining is concerned with the kind of large modern societies to which Rawls’s theory is addressed. In these cases, she denies that reasonability and stability can coincide. She may grant that there are terms of agreement of which it is intuitively plausible to say that they would, if satisfied, establish relations of equality among citizens of large modern societies. And she might grant that those terms would establish equal relations because they satisfy (2). But she might deny that the basic institutions of such societies can effectively encourage their citizens to relate to one another as equals. That is, she might doubt the Rawlsian grounds for (4′) by doubting Rawls’s moral faith in the educative power of the basic structure of modern societies. Indeed, that is one, though just one, of G. A. Cohen’s reasons for rejecting Rawlsian contractualism.^{Footnote 53}

A different objector might also grant that it is possible to identify principles that would satisfy (2), but would go on to deny that human beings are capable of freely relating to one another as equals regardless of social circumstances. This objector doubts the grounds for (4′) by doubting Rawls’s moral faith in the possibility of human goodness.^{Footnote 54}

Note that neither objector is committed to the conclusion, which Cohen and other luck egalitarians endorse, that there are principles of justice which cannot be the terms of inherently stable agreements. Neither objector is committed to this conclusion even if she accepts (1) and (2) and thinks it possible to identify some principle which satisfies (2). For (2) expresses a necessary condition of acceptance in a social contract. It is open to the objector to hold that there are other conditions for acceptance which principles that satisfy (2) must also satisfy but which the identified principle does not. Indeed, she could consistently hold that (5) expresses such a condition, provided she takes (5) to be a basic commitment of contractualism rather than a commitment that depends upon (4′) and the claims that are said to support it. In that case, the objectors would hold that though (1) is true and we can identify principles which satisfy (2), we cannot identify any which satisfy (1). She would then hold that there are no principles of justice for the basic structure of modern society. The only terms we can hope to live by are those of a stable modus vivendi. This is the position of political realism.

In the previous section, I argued that what I am now calling contractualism’s articles of moral faith are among its fundamental commitments. I remarked at the outset that we can gain insight into the structure of philosophical theories by trying to understand them contrastively. The contrast with political realism suggests that the moral faith of contractualism is not just fundamental to it but is also distinctive of it. Contractualism may be distinguished from some other views by the interpersonal form of justification expressed in (1). But that at least some forms of contractualism are also distinguished by faith in the claims about human psychology that are appealed to in support of (4′).

Finally, a persistent puzzle for contractualism is how its adherents — who by definition accept (1) — can consistently be relational egalitarians in domestic politics while endorsing realism or nonegalitarianism internationally.^{Footnote 55} That question seems especially pressing because it is easy enough to think of egalitarian principles of global justice that seem, intuitively, not to impose unreasonable burdens on anyone, and so satisfy (2). The foregoing discussion of political realism suggests an answer.

Contractualists accept (1) and (2), and can grant that there are principles about global distribution that satisfy the condition (2) expresses. But they might deny that there are psychological truths that would support the claim that principles of global justice satisfy (4′). They could deny it by denying that we have any idea how to construct institutions that could make us the kind of persons who would reliably act from such principles. More specifically, they could point out that the psychological claims which support (4′) in the domestic case postulate educational effects which depend upon certain features of domestic institutions — such as the obvious pervasiveness of their effects, the frequency and publicity of their justifications, their influence on civic education, and the public culture that has grown up around them. They could deny that international institutions, in their present state of development anyway, could have those features. If they also accept (5) on grounds other than (4′), they could thereby deny that egalitarian principles which purport to be principles of global justice could be the objects of an agreement which is self-enforcing. They could therefore deny that those principles are in fact principles of global justice because, though they satisfy (2), they fail to satisfy (5), and so do not satisfy (1).

This is consistent with contractualists affirming that principles governing domestic distribution can satisfy (2) and be the objects of self-enforcing agreements. In short, it is open to contractualists to say: reasonable moral faith in the educative power of institutions extends no further than national borders. It may be contended that justice extends beyond those borders, so that we in wealthy countries have duties of justice to the global poor. If that is right, then the justice we owe them would then have to be different in point and perhaps in kind from justice which aims at relational equality — the difference marked by the fact that we do not know how to institutionalize its principles so as to render an agreement on them self-enforcing.