1. The Problem Stated
Today I’m hoping to make you puzzled about a problem that has puzzled me on and off over the years. Misery loves company, I suppose — though the problem doesn’t actually puzzle me at the moment, because at the moment I think I’ve got a solution to it. But I’ve thought this before, and found myself deceived; so I’m not breaking out the champagne just yet.
The problem is this: why does justice have good consequences?
By “justice” I mean the moral system of rights, or more precisely, the virtue concerned with respecting such rights. By “good consequences” I mean not optimal consequences, nor exceptionlessly good consequences, but at least a reliable tendency to produce good consequences, both for oneself and for others. More precisely, to say that justice has good consequences is to say that a policy of respecting people’s rights will ordinarily promote, or at least not require great sacrifices of, the well-being of three groups: those whose rights are being respected, those doing the respecting, and third parties.
The question is: why should this be so?
There are two simple answers to this question. If either of them were true, there would be no puzzle. But I don’t think either of them is true.
2. The No-Such-Explanandum Solution
Why does justice have good consequences? One simple answer would be: it doesn’t. When Socrates refused to act unjustly, he annoyed his neighbours and got himself killed. Hence his commitment to justice apparently brought little benefit either to himself or to others. Think, too, of Rawls’ famous characterisation of justice:
Justice is the first virtue of social institutions, as truth is of systems of thought. A theory however elegant and economical must be rejected or revised if it is untrue; likewise laws and institutions no matter how efficient and well-arranged must be reformed or abolished if they are unjust. Each person possesses an inviolability founded on justice that even the welfare of society as a whole cannot override. For this reason justice denies that the loss of freedom for some is made right by a greater good shared by others. It does not allow that the sacrifices imposed on a few are outweighed by the larger sum of advantages enjoyed by many. Therefore in a just society the liberties of equal citizenship are taken as settled; the rights secured by justice are not subject to political bargaining or to the calculus of social interests. … These propositions seem to express our intuitive conviction of the primacy of justice. 
If the requirements of justice are as aloof from consequentialist considerations as Rawls suggests, then there is obviously no guarantee that justice will even tend to promote on-balance beneficial consequences. So why suppose that the fact I’m trying to explain is a fact at all?
But the no-such-explanandum response is unconvincing. Even if we grant that justice does not produce benefits for everyone in every instance, it seems undeniable that people are by and large better off under conditions of justice. As general policies, rights-violating activities such as murder, robbery, rape, arson, and the like certainly have a negative impact on society, and in most cases end up causing trouble for the perpetrator as well.
Moreover, those who disagree about which policies are just almost always disagree about which policies are beneficial as well. Just think of such currently divisive social issues as: abortion, affirmative action, economic regulation, environmental protection, free trade, military intervention overseas, redistributive taxation, slavery reparations, minimum wage legislation, drug prohibition, the integration of women and/or gays into the armed forces, gay marriage, immigration, pornography, gun control, genetic engineering, intellectual property. For each of these issues, consider which side of the debate you think has a greater claim of justice on its side. Then consider which side of the debate you think would have better consequences. I strongly suspect that, for most people, there will be few if any major discrepancies between the two lists. Regardless of our views on the content of justice, we all seem to expect the implementation of justice to have good consequences.
3. The Indirect-Consequentialist Solution
Our efforts to resist this first simple solution might seem to drive us straight into the jaws of a second: if justice and good consequences tend to go together, perhaps this is simply because justice is founded on consequentialist considerations. Admittedly, justice requires us to stick to certain principles regardless of the consequences; but this might well be because there will be better results in the long run if we treat certain principles as inviolable than if we are too ready to revise them on a case by case basis. A principled commitment to respecting rights is better for society, because people will feel more secure, and will be able to engage in more long-term planning and social coordination, if they know what moral claims they can count on having honoured. A principled commitment to respecting rights is also better for the agent, because one is better off in the long run if one cultivates a reputation for being someone who can be trusted to behave justly, and the most effective way to cultivate such a reputation is to inculcate in oneself a principled commitment. Such suggestions have a long history, reaching back to Epicurus, Hobbes, Hume, and Mill, and figuring more recently in the work of Axelrod, Gauthier, and Yeager. 
But such an indirect-consequentialist solution has a fatal flaw; it runs afoul of the principles of praxeology. Praxeology is the study of those aspects of human action that can be grasped a priori; in other words, it is concerned with the conceptual analysis and logical implications of preference, choice, means and ends, and so forth. The basic principles of praxeology were first discovered by the Greek philosophers, who used them as a foundation for ethics. This approach was further developed by the Scholastics, who extended praxeological analysis to the foundations of economics and social science as well. In the nineteenth and twentieth centuries, this approach to social science was rediscovered by the philosophical economists of the Austrian School, who first dubbed it praxeology.  A crucial praxeological distinction is that between consumer’s goods, which meet human needs directly, and producer’s goods, which are valued for their usefulness in producing or obtaining consumer’s goods. As I have written elsewhere:
Whatever I choose, I choose either as a consumer’s good (a first-order good) or as a producer’s good (a higher-order good). Utilitarianism of any sort regards morality as a producer’s good, a means of producing happiness; but indirectutilitarianism maintains, in effect, that the most effective way to promote happiness is to treat morality as if it were a consumer’s good, even though it isn’t one. But is it really possible to adopt the attitude that indirect utilitarianism recommends? When I choose morality “as if” it were a consumer’s good, either it really becomes a consumer’s good for me, or else it remains a producer’s good and I am only pretending. There is no third possibility.
Suppose it does become a consumer’s good for me. In that case, I am no longer a consistent utilitarian, since in my actions I reveal a preference for morality as an end in itself. [Utilitarians sometimes recommend] treating a principle as inherently binding at the everyday level while recognizing its contingency on utilitarian outcomes at the reflective level …. but doesn’t this just amount to advising us to form inconsistent preferences? And if the preferences on which I ordinarily act do treat morality as a consumer’s good, in what sense can it be said that I really regard it as a producer’s good? On the other hand, suppose that morality remains a producer’s good for me. [E]very action embodies a means-end scheme … [E]ven when I choose to act morally, my choice commits me to rejecting morality in counterfactual situations … where immorality would be a more effective means to the end, and this commitment is a blot on my character now. (Hence the Kantian insistence on the importance of maxims rather than actions.)
It has often been claimed that indirect utilitarianism is unstable, and must collapse either into direct utilitarianism on the one hand or into “rules fetishism” on the other. This can be interpreted as a psychological claim about the likely results of trying to maintain a utilitarian attitude, in which case its truth or falsity is an empirical matter. By transposing the familiar stability objection into a praxeological key, however, what I’ve been trying to show is that indirect utilitarianism is not just causally but conceptually unstable. If I treat morality as a consumer’s good, I must reject utilitarianism on pain of inconsistency; if I treat morality as a producer’s good, I thereby exhibit a moral character or disposition that utilitarian considerations themselves condemn. But I must treat morality in one way or the other; hence utilitarianism is praxeologically self-defeating. The praxeologist cannot be a direct utilitarian, since praxeological reasoning itself shows us that the utilitarian’s goal depends on social cooperation, which in turn requires the kind of stable and consistent commitment to principles that a direct utilitarian cannot have. Nor can the praxeologist be anindirect utilitarian, since praxeological considerations force a choice between treating morality as a producer’s good (in which case we’re back with direct utilitarianism) and treating it as a consumer’s good (in which case utilitarianism prescribes its own rejection). We may have utilitarian reasons for adopting moral commitments, but once we have adopted them, we can no longer regard them as resting on purely utilitarian foundations — because so regarding them would alter their status as commitments. 
If indirect consequentialism is praxeologically incoherent, we cannot accept the indirect-consequentialist solution to the question of why justice has good consequences. But then we are left with a puzzle: the same conduct is both just and beneficial, but it’s not just because it’s beneficial. So is it just an extraordinarily fortunate coincidence that justice and benefit tend to go together?
4. The Rawlsian Solution
The Rawlsian theory of justice might seem to provide a way out of this puzzle. Justice as fairness is not a purely consequentialist theory, but it incorporates consequentialist concerns. In particular, the second principle of justice authorises departures from socioeconomic equality, so long as those departures make everyone better off than they would be under equality. The beneficial effects of various social orders will thus be taken into account in determining the justice of those social orders. So it’s no coincidence that justice and good consequences tend to go together. On the other hand, justice as fairness is not a purely consequentialist theory, and so is apparently not vulnerable to the charge of praxeological incoherence. Is our problem solved?
Unfortunately, no. For as Rawlsian theory requires, the semi-consequentialist second principle is lexicographically posterior to the decidedly non-consequentialist first principle. In any conflict between the two principles, the first one trumps the second. But whether the first principles has good consequences will then be a contingent empirical matter. Likewise, the frequency of conflicts between the two principles, and thus the frequency with which the second principle gets trumped, will presumably also be a contingent empirical matter. Rawlsian theory thus provides no guarantee that justice as fairness will even tend to produce good consequences.
It might be replied that the first principle too is semi-consequentialist, because the contractors behind the veil of ignorance choose it for consequentialist reasons. This is true. But the informational constraints on that choice are so severe that there is little reason to expect even a rough correlation between ex ante judgments of benefit behind the veil and ex post judgments of benefit after the veil has been lifted. Rawlsian justice as fairness leaves the correlation between justice and good consequences a mystery.
5. First Digression: Counterfactuals and Moral Knowledge
What if it turned out that — contrary to what we now believe — the principles we identify as just are, or would be, the cause of disastrous social consequences? The consequentialist must say that these principles should then be abandoned; the deontologist must say that obedience to these principles would still be morally required. On the other hand, suppose it turned out that although the principles of justice we favour do have the best social consequences, they do so only at the cost of, say, treating people as mere means, and that some other set of principles, while socially disastrous, far better exemplifies the ideals of fairness and respect for persons. In that case, it seems, the deontologist must say that the principles should then be abandoned; the consequentialist must say that they would still be morally required.
It appears, then, that the disagreement between consequentialists and deontologists turns on the question of which counterfactual situations are relevant to moral justification. Both sides can agree that certain principles of justice a) have good consequences, and b) express respect for persons. But for the consequentialist, if (a) were to be falsified, those principles would thereby be overturned, whereas if only (b) were falsified, they would not be. For the deontologist, however, it is the falsification of (b), not of (a), that would overturn the principles.
Or so it seems. But in fact things are not so simple. In real life, one rarely finds members of either camp relying solely on a single set of considerations. It is a rare moral or political polemic indeed that does not include both consequentialist and deontological arguments.
Why is this? One might think the reason is purely strategic. Most people are unlikely to find the deontological case for a given course of action compelling so long as they believe it would have terrible consequences; likewise, they are equally unlikely to find the consequentialist case compelling so long as they believe that the action violates human dignity, or equality, or liberty. But while a combination of consequentialist and deontological arguments is most assuredly the best rhetorical strategy for persuading people to accept one’s views, I don’t think it’s mainly for rhetorical reasons that would-be persuaders combine both sorts of considerations. On the contrary, the persuaders combine both sorts of considerations precisely because they sharewith the persuaded a reluctance to accept one without the other. Whatever they may say officially, most consequentialists would be deeply disturbed to discover that their favoured policies slighted human dignity, and most deontologists would be deeply disturbed to discover that their favoured policies had disastrous consequences.
This fact has often led each camp to suspect the other of hypocrisy. The consequentialists say: “Look at all the effort you deontologists put into trying to show that abiding by your principles won’t have counterintuitively disastrous consequences. For example, notice how eager contemporary Kantians are to distance themselves from Kant’s claim that it’s wrong to lie to a murderer at your door. Obviously, you deontologists implicitly regard harmful consequences as potential falsifiers of your theory; you’re really crypto-consequentialists, not sincere deontologists with the courage of your convictions.” 
And the deontologists can reply in kind: “Look at all the effort you consequentialists put into trying to show that your theory doesn’t license counterintuitively unjust actions. For example, notice how quick contemporary utilitarians are to insist, via such devices as rule-utilitarianism, that they are not committed to sacrificing one innocent person to save ten others. Obviously, you consequentialists implicitly regard sanctioned rights-violations as potential falsifiers of your theory; you’re really crypto-deontologists, not sincere consequentialists with the courage of your convictions.”
What are we to make of the fact that each side appears to regard the considerations advanced by the other side as crucial to moral justification? To be a consequentialist advocate of X is to believe that, so long as X still had good consequences, X would be justified even if it were discovered that X exemplifies contempt for persons; yet most consequentialists would regard such a discovery as seriously weakening the case for X. Likewise, to be a deontological advocate of X is to believe that, so long as X still exemplified respect for persons, X would be justified even if it were discovered that X had bad consequences; yet most real-life deontologists would regard such a discovery as seriously weakening the case for X.
It seems, then, that each side is committed to giving inconsistent answers to the question “What if it turned out that X failed to meet the other side’s standards?” To be on either side is to be committed to regarding such a counterfactual situation as irrelevant to the justification of X; but in actual practice, few if any advocates of X (for any X) do regard it as irrelevant.
One approach to solving this problem is to invoke a distinction popular among the early Natural Law theorists: that between a principium essendi and a principium cognoscendi. 
principium essendi of X: that in virtue of which X is so.
principium cognoscendi of X: that in virtue of which X may be recognised as being so.
For example, sandalwood has a distinctive smell by which it can easily be identified; that smell is a principium cognoscendi of sandalwood. But that smell is not what makes sandalwood what it is, so it is not sandalwood’s principium essendi; theprincipium essendi of sandalwood would presumably be something like its biochemical microstructure. But although sandalwood is not defined by its smell, in the absence of that smell we would be justified in doubting that a given sample of wood is really sandalwood. (Of course a thing’s principium essendi will ordinarily be one of its principia cognoscendi as well; the point is simply that the class of principia cognoscendi will be broader.)
The fact that moral theorists in both camps rely on consequentialist and deontological considerations alike, not just to convince others but to convince themselves, suggests that members of each camp implicitly regard both sorts of considerations asprincipia cognoscendi of moral justification; and there is no inconsistency in doing this while at the same time regarding only one sort of consideration as the principium essendi. Just as a sandalwood-detector can take a certain smell as a reliable sign of the presence of sandalwood without taking that smell to be the essence of sandalwood, so deontologists and consequentialists can take beneficial consequences and respect for persons, respectively, as reliable signs of moral justification, though not as its essence. 
It now becomes clear that the counterfactual questions we have been worrying about are seriously ambiguous. Consider the following sort of example, familiar from the Kripke-Putnam literature on identity. Most of us believe that water is a compound of hydrogen and oxygen molecules; water just is H2O. But that view commits us to the consequence that if there were no hydrogen and oxygen molecules (and thus no H2O), there would be no water. Yet in fact, if we were to find out that our scientific theories are all wrong and that in reality there are no such things as hydrogen and oxygen molecules, we would not conclude that water is nonexistent; instead, we would conclude that we had been mistaken in identifying water with H2O. Does that make our position inconsistent? No. It’s important to distinguish between metaphysical and epistemic counterfactuals:
metaphysical counterfactual: What if it were actually the case that p?
epistemic counterfactual: What if I were to come to believe that p?
The question “what if it turned out that H2O is nonexistent?” is ambiguous as between “what if it were actually the case that H2O is nonexistent?” (in which case the answer is that water, being identical with H2O, would also be nonexistent) and “what if we were to come to believe that H2O is nonexistent?” (in which case the answer is that we would give up our belief that water and H2O are identical). One evaluates the metaphysical counterfactual from the standpoint of one’s actual beliefs: the nonexistence of H2O would mean the nonexistence of water, given that we are right in identifying the two; but since the case for acknowledging the existence of water is older and stronger than the case for acknowledging the existence of H2O, we would almost certainly take any evidence against the existence of H2O as evidence against the identity of water with H2O, rather than as evidence against the existence of water.
Likewise, given the available evidence, a detective investigating a murder may conclude that the butler is the murderer, and this likewise commits the detective to believing that if the butler had not been present at the time of the murder, the murderer would not have been present either. But, all other things being equal, the detective will probably treat any new evidence that the butler was not in fact present, not as tending to show that the butler committed the murder from a distance, but rather as tending to show that someone other than the butler committed the murder. Hence it would be foolish to object to the detective: “Oho, so you think the butler and the murderer are identical? But the claim is absurd, for it commits you to attributing extraordinary powers to the butler, should he turn out to have been five thousand miles away at the time of the murder.”
Similarly, in defending the divine command theory of ethics, Robert Adams relies on the Kripke-Putnam conception of identity in order to rebut the objection that divine command ethics requires us to abandon morality entirely if it should turn out that God is nonexistent. While I hold no brief for the content of Adams’ metaethical theory, his remarks about its form offer useful insights. Adams writes:
The thesis that wrongness is (identical with) contrariety to a loving God’s commands must be metaphysically necessary if it is true. That is, it cannot be false in any possible world if it is true in the actual world. For if it were false in some possible world, then wrongness would be non-identical with contrariety to God’s commands in the actual world as well, by the transitivity of identity, just as Matthew and Levi must be non-identical in all worlds if they are non-identical in any. … If it is necessary that ethical wrongness is contrariety to a loving God’s commands, it follows that no actions would be ethically wrong if there were not a loving God. This consequence will seem (at least initially) implausible to many, but I will try to dispel as much as I can of the air of paradox. It should be emphasized, first of all, that my theory does not imply what would ordinarily be meant by saying that no actions are ethically wrong if there is no loving God. If there is no loving God, then the theological part of my theory is false; but … in that case ethical wrongness is the property with which it is identified by the best remaining alternative theory. 
Adams wishes to endorse the following claims:
I. Ethical wrongness = contrariety to the commands of a loving God.
II. Ethical wrongness exists.
III. A loving God exists.
IV. If (III) were false, then (I) would be true and (II) would be false.
V. If (III) is false, then (I) is false and (II) is true.
The way Adams puts things invites us to suppose that the crucial difference between (IV) and (V) lies in the fact that the former, but not the latter, is expressed in counterfactual form. But it may be more helpful to see (IV) and (V) as expressing different sorts of counterfactuals – viz., metaphysical and epistemic counterfactuals, respectively. (IV) makes a counterfactual claim about the world, from the standpoint of Adams’ present beliefs. (V), on the other hand, makes a counterfactual claim about how Adams would revise the rest of his belief system upon the falsification of (III).
The structure of Adams’ beliefs about ethical wrongness is analogous to the structure of our beliefs about H2O. We believe a) that water and H2O are one, and thus b) that if there were no H2O there would be no water. Yet we also believe c) that we can identify instances of water in our environment. At present, these beliefs do not involve us in any inconsistency. If, however, we were to come to believe (rightly or wrongly) that H2O is nonexistent, we would be forced to choose between accepting (a) and (b) on the one hand, and accepting (c) on the other; and if it is true d) that under those circumstances we should keep (c) and reject (a) and (b), then the answer to the question “If it should turn out that there is no H2O, does that mean that the existence of water must be rejected?” is yes if taken in metaphysical sense (b) and no if taken in epistemic sense (d).
H2O is the principium essendi of water; its stereotypical surface features such as transparency, potability, colourlessness, odourlessness, freezing at 273.15° K, and boiling at 373.15° K are not water’s principium essendi, but do constitute an important principium cognoscendi of water. But both these beliefs — the belief that H2O is water’s principium essendi and the belief that the surface features are water’s principium cognoscendi — are in principle open to revision; and if they come into conflict, there is no guarantee that the latter, rather than the former, belief will be revised.
The same applies, I suggest, to our puzzle about justice. The consequentialist, for example, believes the following propositions:
1. Conduct X has good consequences.
2. Conduct X expresses respect for persons.
3. Having good consequences is a principium essendi of justice.
4. Expressing respect for persons is a principium cognoscendi of justice.
5. Conduct X is just.
Suppose the consequentialist ceases to believe (2). It follows that she can no longer consistently believe both (4) and (5); she must reject one or the other. If she holds on to (1) and (3), she must keep (5) and reject (4). But nothing requires her to hold on to (1) and (3). It is equally open to her to hold on to (4) and to reject (3) and (5). It all depends to which beliefs she is most strongly committed; and the H2O example shows that we are not necessarily more strongly committed to beliefs aboutprincipia essendi than to beliefs about principia cognoscendi.
All of the above applies mutatis mutandis to the deontologiost, who believes:
1. Conduct X expresses respect for persons.
2. Conduct X has good consequences.
3. Expressing respect for persons is a principium essendi of justice.
4. Having good consequences is a principium cognoscendi of justice.
5. Conduct X is just.
If the deontologist ceases to believe (2), she must choose between (4) and (5); and there is no guarantee that (5) will win. Believing that X is the foundation of Y does not commit one to making one’s belief in X the foundation of one’s belief in Y; this is to confuse the ground of explanation with the ground of knowledge — or, in Aristotle’s terms, to confuse what is best known in itself with what is best known to us. In short, it is to confuse principia essendi with principia cognoscendi.
But why should we expect consequentialist and deontological criteria to go together, even for the most part? Once we have accepted one set of considerations as the principium essendi of justice, what justifies us in granting a rival set of considerations the status of principia cognoscendi? As a reflective equilibrium theorist, I don’t think we necessarily need to have an explanation for the concurrence of deontological and consequentialist criteria in order to be justified in believing in such a concurrence.  Still, it would be surprising if no such explanations were forthcoming; indeed, the absence of such explanations where one expected to find them might well lead us to revise our belief that the criteria do coincide.
So the question remains: if good consequences are not the principium essendi of justice, why are they among its principia cognoscendi? What’s the connection?
6. The Theistic Solution
One obvious answer is to appeal to a divine creator who secures the correlation. This hypothesis takes two forms. The first is that the deontological principles of justice were laid down by the creator with an eye to their beneficial effects. So it’s no surprise that duties track benefits. But in that case, is it God’s choice, or the reasons for God’s choice, that constitute the principium essendi of justice? If the former, we run into the usual problems of divine command morality. If the latter, we’re back with indirect consequentialism, which we’ve already seen to be praxeologically incoherent.
The second form flips the direction of explanation. The deontological aspects of justice are its principium essendi; but since God is benevolent and omniscient, she has arranged the causal laws of the universe so that humankind will be rewarded for just conduct. So it’s no surprise that benefits track duties.
Notoriously, this position involves a number of difficulties. It will suffice to name just one: any deity powerful enough to arrange a rough concurrence of justice and benefit could presumably have arranged a more precise one than that which we currently enjoy; her failure to do so must thus be explained, and any such explanation, to the extent that it is successful, is likely to make even the rough concurrence mysterious once more.
In other words, let x signify the degree to which justice and benefit currently coincide. Conceivably, the degree might have been higher or lower than x. So why isn’t it x + 1? If no explanation can be given as to why x + 1 is too high a level for God to be willing, or able, to produce it, then the theistic solution simply exchanges one mystery for another. But if, instead, such an explanation can be given, it will be hard to see why, if x + 1 is too high, x isn’t too high as well, and so the existing level of concurrence between justice and benefit will still be in need of explanation. Either way, the theistic solution appears to fail.
7. The Evolutionary Solution
Perhaps evolution, the currently fashionable substitute for God, might be pressed into service here. After all, evolution is not supposed to be omnipotent or benevolent, so the failure to produce a perfect concurrence between justice and benefit might be more easily explainable.
According to the evolutionary solution, since beings who cooperate with each other tend to be more successful than beings who don’t, both biological and cultural evolution will favour those with cooperative dispositions, and so we will tend to find plausible those principles that urge us to behave in cooperative rather than predatory fashion. The survival value of justice cannot, of course, be its principium essendi — that would land us back in indirect consequentialism. But we might instead assume, à la Adam Smith, that our psychological dispositions to approve and disapprove are the principium essendi of justice; so if those dispositions have been shaped by biological and cultural evolution, it’s no surprise that justice and benefit tend to coincide.
At one time I thought something like this solution might be the answer to my puzzle. I no longer think so. Rather, I now think that the evolutionary solution is vulnerable to a variant of the same objection that felled the indirect-consequentialist solution. Here’s why. Evolutionary considerations may explain why we approve of X. But on pain of indirect consequentialism, and thus of praxeological incoherence, we cannot regard such considerations as explaining why X is deserving of our approval. In approving of X, we must regard X’s merit as independent of the evolutionary process whereby we came to approve of it. But in that case, all that’s been explained is why we came to approve non-consequentially of something that in fact has good consequences, not why it’s justice that plays that role. For all that the evolutionary story tells us, it’s still a cosmic coincidence that what has survival value is the disposition to approve of the very thing that actually merits our approval.
8. Second Digression: Ends and Means
What sort of value does justice have? Is justice to be valued as a means, as an ultimate end, or neither?
Some deontologists might plump for the latter option: neither. Rights are not goals to be pursued, either as ends in themselves or as means to further ends; rather, they are side-constraints on our pursuit of goals. But it’s difficult to make sense of this idea praxeologically. If justice is neither one of our ultimate ends, nor a means to one of our ultimate ends, what reason could we have to care about it?
Suppose, then, that justice is an ultimate end — one that serves no further value beyond itself. Then either it is our sole ultimate end, or it is one among others. But it would be very odd to have justice as one’s sole ultimate end, as though respecting people’s rights were the one and only goal worth pursuing. Such an end would radically underdetermine the shape of one’s life.
If justice is an ultimate end, then, it must be one among others. But in that case, how is it to be integrated with our other ultimate ends? Do we make trade-offs when ultimate ends conflict? Or do we look for some way of conceiving of our ultimate ends so that conflicts are impossible? In either case, we seem to be asking how to fit justice into the broader goal of an integrated lifeplan – what the Greeks called eudaimonia. But then we are no longer treating justice as an ultimate end; justice now serves the more inclusive end of eudaimonia.
Hence justice must be understood as a means, not as an ultimate end. But here again there are two options; justice is either an external means or an internal means. An external means bears a causal or instrumental relation to its end, while an internal means bears a logical or constitutive relation to its end. If Freud is right, then my motive in writing this address was to win “fame, fortune, and the love of women.” This would be an example of an external means. (What the causal mechanism is I’m not sure.) By contrast, playing this particular chord here — Kevin was supposed to provide me with a calliope at this point, or at the very least with two elephants who would trumpet in different keys, but I guess you’ll have to use your imagination — playing this particular chord here is an internal means to playing the Moonlight Sonata. I’m not playing the chord as an end in itself; the chord’s value to me lies in its contribution to the whole sonata. So the chord is a means — but not an external means. One test for the difference is to see whether it makes sense to wish for the end without the means. It makes sense to say, “I wish I could achieve fame, fortune, and the love of women without having to compose this Presidential Address,” because the means and the end are logically separable; but it doesn’t make sense to say, “I wish I could play the Moonlight Sonata without having to play all these notes.” Just these notes, played in just this sequence, constitute the Moonlight Sonata; there’s nothing we could count as playing the Moonlight Sonata without playing the particular sequence of notes of which it is composed.
Now if the value of justice lies in its being an external means to some end, then it makes sense to wish for the end without having to use the means — in which case we’re entangled once again in the same sort of paradox that afflicts indirect consequentialism. Indeed, I think any theory that sees justice as a solution to a problem, or sees rights as a device for protecting people’s interests, is in danger of running afoul of the same paradox, so long as the means and the end are treated as logically separable — in which case many of liberal rights theory’s most ardently anti-utilitarian thinkers, from Rawls and Dworkin to Rand and Rothbard, are skating on thin ice over a utilitarian abyss.
Treating justice solely as an external means is inconsistent with the kind of counterfactually stable commitment that justice must involve in order to function successfully even as an external means. Hence the value of justice can ultimately lie only in its being an internal means as well. This is essentially the Platonic and Aristotelean view: justice is an internal means to eudaimonia. Since nothing counts as that end in the absence of that means, no counterfactual stability problem arises.
Just about every paper of mine sooner or later features Aristotle descending in a contrivance at some crucial point in the plot, and now seems as good a time as any. If the structure of justice turns out to be Aristotelean, perhaps the solution to our larger puzzle will be so as well.
9. Victory, Part I: The Unity-of-Virtue Solution
Aristotle and other Greek philosophers (e.g., Socrates, Plato, the Stoics) accept a principle known as the unity of virtue. This principle is sometimes described as holding that one can’t (fully) possess any one virtue without possessing them all; but in fact that is merely a corollary of the fundamental principle, which is that one can’t specify the content of any one virtue independently of the content of all the other virtues.
Consider the following example. The Hilton Beachfront Inn is burning down, and Eric Marcus is trapped under a gigantic pumpkin-coloured beach umbrella. I could rush in and try to save him, but at considerable risk to myself. One might think of courage as counseling me to take the risk, and prudence as counseling me not to take the risk; but from an Aristotelean perspective, this would misdescribe the situation. The virtue of courage does not require us to take any and all risks, but only those risks that are worth taking; facing a danger worth running away from is no more admirable than running a way from a danger worth facing. Taking stupid risks is not admirable, and so is incompatible with what virtue requires. Likewise, the virtue of prudence does not require us to save our skins at all costs; we have a prudential interest not just in the length of our lives but in their quality, where quality of life depends, in turn, not just on material comforts but on whether we are living a life worthy of admiration and respect. Hence saving Eric is not courageous if it is imprudent; and letting Eric die is not prudent if it is cowardly. What courage requires of me in this instance cannot be determined independently of determining what prudence requires of me, and vice versa; the contents of the two virtues are specified reciprocally, via mutual adjustment. That is why I cannot possess one virtue fully without possessing them all: virtues require counterfactual stability. I do not count as fully courageous unless I can be counted on to do the courageous thing inevery situation, which in turn requires that I be a reliable assessor of which risks are worth taking; but which risks are worth taking might sometimes depend on the requirements of prudence, or justice, or loyalty; to the extent that I am imprudent, or unjust, or disloyal, I cannot be counted on to assess those risks properly in such possible or actual situations, and so I will not be fully just. (Since not all the virtues will be relevant to every individual choice, nothing in the unity-of-virtue thesis, so far as I can see, rules out the possibility of possessing different virtues to different degrees – being, say, more reliably just than reliably courageous. 80% courage might be compatible with 65% generosity; but 100% courage requires 100% generosity.)
The unity-of-virtue thesis also implies that the requirements of the various virtues cannot conflict. Nowadays even such enthusiastic Hellenophiles as Williams, Nussbaum, and MacIntyre tend to dismiss this claim as unduly optimistic. So it will certainly seem, if we follow the modern habit of labeling any desirable character trait as a virtue. But virtues are principles of choice; to say that courage, or loyalty, or temperance requires a certain course of action is to say that we ought to follow that course of action; and ought implies can. The Greeks are not committed to the claim that all things worth wishing for are compossible, but only to the claim that all things worth aiming for are compossible. The requirement to integrate our aims into a compossible system plays a role in determining the content of the aims; it is in that sense that all goods, including the virtues, turn out to be means, internal or external, to eudaimonia.
If the contents of the virtues are specified by reciprocal determination, it follows that the content of justice is partly specified by, inter alia, such virtues as prudence and benevolence — virtues that have among their chief concerns the production of benefits, for oneself and for others respectively. This does not make justice a consequentialist notion, since the direction of determination runs both ways; what counts as a beneficial consequence will be partly determined by the requirements of justice. For the concept of benefit is in reciprocal determination with the concepts of prudence and benevolence, which in turn are in reciprocal determination with the concept of justice. Thus justice and benefit are brought into reciprocal determination with one another.
On this view, human welfare (whether individual or general) and justice are conceptually interrelated, with neither concept being basic but each depending in part on the other (and all the other virtues) for its content, just as Aristotle defines virtue and human flourishing in terms of one another. Since (for reasons pointed out, in rather different ways, by John Rawls and Bernard Williams) principles of justice that imposed unreasonable and excessively self-sacrificial demands on moral agents would be unfair, there are reasons of justice for attempting to accommodate self-interested concern. And since a way of life that did not allow agents to regard themselves as admirable, or their lives as tracking genuine value, would not be worth living, there are self-interested reasons for attempting to accommodate justice. These considerations yield a version of ethical constructivism (though not of an anti-realist variety) in which neither the concept of justice nor the concept of welfare has a completely determinate content independently of the other, but the optimal conception of the good life is constructed out of the mutual adjustment of such concepts. It turns out, then, that justice and benefit are each a partial principium essendi of the other.
(One important implication of this approach is to recast the debate between Rawls and his critics (e.g., Nozick and Sandel). Rawls maintains that the correct principles by which to assess political and social institutions are those that self-interested contractors could rationally agree to under fair bargaining conditions; his critics argue that there are substantive moral values that hold prior to and independently of the social contract procedure, and that these values should constrain the principles that result. But if, as my approach maintains, neither the self-interest of the contractors nor the independent moral values can be fully specified without reference to the other, then each side has adopted an excessively absolutistic position, and a basis for compromise through mutual adjustment emerges.)
We can now see our way, apparently, to a solution to the problem of why justice has good consequences. It isn’t just a happy coincidence; rather, justice and benefit are conceptually entangled; their internal conceptual dynamic drives them into alignment with one another. Semi-deontological considerations of justice play a role in determining what counts as a good consequence; semi-consequentialist considerations of benevolence and prudence play a role in determining what counts as just. Hence it is only to be expected that justice should tend to coincide with benefit, both for oneself and for others. The reason justice does not ordinarily require great sacrifice of anybody’s welfare, is that any conceptions of justice and of welfare on which the former required constant sacrifice of the latter would demand revision of one or the other or both.
At one time I thought the unity-of-virtue solution was the complete and final answer to my puzzle. Unfortunately, there’s a wrinkle. I think the unity-of-virtue solution is a partial answer; in particular, the fact that justice and benefit are conceptually interconnected explains why we implicitly tend to assume that the two will go together. But the unity-of-virtue solution explains the concurrence of justice with benefit by showing that the content of each notion has been adjusted to bring it into conformity with the other. This solution gives us no reason, however, to expect any concurrence between the prima facie contents of justice and benefit, before they have been mutually adjusted. (And there must be some such prima facie contents. If, prior to mutual adjustment, no virtue had any content at all, there’d be no basis for such adjustment to begin.) Should there turn out to be even a rough concurrence (not just between the adjusted contents but) between the prima facie contents of justice and benefit, the unity-of-virtue thesis could offer no explanation for this. That much of the concurrence would remain a mysterious coincidence.
Unfortunately — well, fortunately for humankind, I suppose, but unfortunately for my theory, which is surely more important — there does appear to be such a concurrence. If we tried to specify the content of benefit without bringing in considerations of justice, or virtue generally, we would get something broadly like long-run preference-satisfaction. If we tried to specify the content of justice without bringing in considerations of beneficial consequences, we would get something like libertarianism.
My argument for this latter claim is that, absent consequentialist considerations, the libertarian conception of equality in authority — the kind of equality enjoyed in a Lockean state of nature — answers better than any of its rivals to the basic Kantian demand to treat persons as ends in themselves rather than as mere means. As I have written elsewhere:
[T]he equality that Locke and Jefferson speak of is equality in authority: the prohibition of any “subordination or subjection” of one person to another. Since any interference by A with B’s liberty constitutes a subordination or subjection of B to A, the right to liberty follows straightforwardly from the equality of “power and jurisdiction.” …
[S]ocioeconomic equality and legal equality both fall short of the radicalism of Lockean equality. For neither of those forms of equality calls into question the authority of those who administer the legal system; such administrators are merely required to ensure equality, of the relevant sort, among those administered. Thus socioeconomic equality, despite the bold claims of its adherents, does no more to challenge the existing power structure than does legal equality. Both forms of equality call upon that power structure to do certain things; but in so doing, they both assume, and indeed require, an inequality in authority between those who administer the legal framework and everybody else.
The libertarian version of equality is not circumscribed in this way. [E]quality in authority entails denying to the legal system’s administrators – and thus to the legal system itself – any powers beyond those possessed by private citizens …. Lockean equality involves not merely equality before legislators, judges, and police, but, far more crucially, equality with legislators, judges, and police. …
The case against socioeconomically egalitarian legislation is … an egalitarian one; for such legislation invariably involves the coercive subordination or subjection of dissenting individuals to the taxes and regulations imposed by government decision makers, and thus presupposes an inequality in authority between the former and the latter.
Nor would an anarchistic version of socialism fare any better; as long as some people are imposing redistributive policies by force or threat of force on unconsenting others, we have inequality in authority between the coercers and the coerced, regardless of whether those doing the coercing are public citizens or private individuals, and regardless of whether they represent a majority or a minority. Nor would a Hobbesian jungle, where anyone is free to impose her will on anyone else, embody equality in authority; for as soon as one person does succeed in subordinating another, an inequality in authority emerges.
The Hobbesian jungle might represent equal opportunity for authority, but in this context the libertarian favors equality of outcome. (That, incidentally, is why the right to liberty is inalienable.) Only defensive uses of force are justified, since these restore equality in authority rather than violating it. By the same token, an idealized democracy in which every citizen had an equal chance to get into a position of political power would also represent only equal opportunity for authority, not equality of outcome, and so would likewise offend against Lockean equality. To a libertarian, the saying “anyone can grow up to become president,” if it were true, would have the same cheery ring as “anyone might be the next person to assault you.” 
(Somehow that last line seems more appropriate now than when I first wrote it.)
What I claim, then, is that libertarianism represents the prima facie content of justice, considered apart from consequentialist considerations. Hence most objections to libertarianism are broadly consequentialist, even when brought by deontologists. For example, Rawls’ chief objection to libertarianism is that it would allow unfair socioeconomic inequalities. The ground of the objection is a deontological one, but what it objects to is a purported consequence of libertarianism, and so is not an objection that could derive from the demands of justice considered in abstraction from consequences.
The problem for the unity-of-virtue solution, then, is that there seems to be a rough concurrence between the prima facie libertarian content of justice, considered apart from consequences, and the prima facie subjectivist content of benefit, considered apart from justice. The social theorists of the Austrian School have shown, on praxeological grounds, how a libertarian social order constitutes an economic democracy, in which consumer preferences direct the productive resources of society through the imputation of value from consumer goods to goods of higher order.  Hence justice, as it would be conceived prior to adjustment, does a reasonably good job of producing beneficial consequences, as those would be conceived prior to adjustment. Whether one thinks that the alterations to be produced in these two concepts after adjustment would be great or small, the fact remains that there is a rough concurrence prior to adjustment, and this rough concurrence seems to require explanation. But the explanation is one that the unity-of-virtue solution cannot provide.
10. Victory, Part II: The Praxeological Solution
Perhaps Vienna can come to the aid of Athens. The praxeological economists whose work creates this problem for the unity-of-virtue solution may also provide the means of resolving it. I shall quote at length from Friedrich Hayek — because my words need some accompaniment, and Kevin still hasn’t brought that calliope:
[A]ll propositions of economic theory refer to things which are defined in terms of human attitudes toward them …. I am not certain that the behaviorists in the social sciences are quite aware of how much of the traditional approach they would have to abandon if they wanted to be consistent or that they would want to adhere to it consistently if they were aware of this. It would, for instance, imply that propositions of the theory of money would have to refer exclusively to, say, “round disks of metal, bearing a certain stamp,” or some similarly defined physical object or group of objects. 
That the objects of economic activity cannot be defined in objective terms but only with reference to a human purpose goes without saying. Neither a “commodity” or an “economic good,” nor “food” or “money,” can be defined in physical terms …. Economic theory has nothing to say about the little round disks of metal as which an objective or materialist view might try to define money. … Nor could we distinguish in physical terms whether two men barter or exchange or whether they are playing some game or performing some ritual. Unless we can understand what the acting people mean by their actions any attempt to explain them, that is, to subsume them under rules … is bound to fail. 
Take such things as tools, medicine, weapons, words, sentences, communications, and acts of production — or any one particular instance of these. I believe these to be fair samples of the kind of objects of human activity which constantly occur in the social sciences. It is easily seen that all these concepts (and the same is true of more concrete instances) refer not to some objective properties possessed by the things, or which the observer can find out about them, but to views which some other person holds about the things. These objects cannot even be defined in physical terms, because there is no single physical property which any one member of a class must possess. These concepts are not merely abstractions of the kind we use in all physical sciences; they abstract from all the physical properties of the things themselves. … [W]e do not even consciously or explicitly know which are the various physical properties of which an object would have to possess at least one to be a member of a class. The situation may be described schematically by saying that we know the objects a, b, c,…, which may be physically completely dissimilar and which we can never exhaustively enumerate, are objects of the same kind because the attitude of X toward them all is similar. But the fact that X’s attitude toward them is similar can again be defined only by saying that he will react toward them by any one of the actions a, b, g,…, which again may be physically dissimilar and which we will not be able to enumerate exhaustively, but which we just know to “mean” the same thing. …
As long as I move among my own kind of people, it is probably the physical properties of a bank note or a revolver from which I conclude that they are money or a weapon to the person holding them. When I see a savage holding cowrie shells or a long, thin tube, the physical properties of the thing will probably tell me nothing. But the observations which suggest to me that the cowrie shells are money to him and the blowpipe a weapon will throw much light on the object — much more light than these same observations could possibly give if I were not familiar with the concept of money or a weapon. In recognizing the things as such, I begin to understand the people’s behavior. I am able to fit [the object] into a scheme of actions which “make sense” just because I have come to regard it not as a thing with certain physical properties but as the kind of thing which fits into the pattern of my own purposive action. …
[A]s we go from interpreting the actions of men very much like ourselves to men who live in a very different environment, it is the most concrete concepts which first lose their usefulness for interpreting the people’s actions and the most general or abstract which remain helpful longest. My knowledge of the everyday things around me, of the particular ways in which we express ideas or emotions, will be of little use in interpreting the behavior of the inhabitants of Tierra del Fuego. But my understanding of what I mean by a means to an end, by food or a weapon, a word or a sign, and probably even an exchange or a gift, will still be useful and even essential in my attempt to understand what they do. …
From the fact that whenever we interpret human action as in any sense purposive or meaningful … we have to define both the objects of human activity and the different kinds of action themselves, not in physical terms but in terms of the opinions or intentions of the acting persons, there follow some very important consequences; namely, nothing less than that we can, from the concepts of the objects, analytically conclude something about what the actions will be. If we define an object in terms of a person’s attitude toward it, it follows, of course, that the definition of the object implies a statement about the attitude of the person toward the thing. When we say that a person possesses food or money, or that he utters a word, we imply that he knows that the first can be eaten, that the second can be used to buy something with, and that the third can be understood — and perhaps many other things. 
This is the Austrian case for claiming that the laws of economics, and of the social sciences generally, are a priori conceptual truths. Concepts like “price,” “unemployment,” “money,” and so forth are defined in terms of people’s attitudes and actions concerning such items, so it is no surprise that there should turn out to be conceptual truths about how people will behave with regard to such items. The principles of economics thus turn out to have the same status as the principles of logic and mathematics. 
If the Austrians are right, and I think they are, then a solution to our problem may be in sight. The fact that a libertarian social order tends to satisfy consumer preferences is not a contingent empirical fact; the Austrians argue at length — and I do mean at length: Mises’ Human Action and Rothbard’s Man, Economy, and State weigh in at nearly a thousand pages each — that this concurrence can be established by conceptual analysis.
But if this is so, then the concurrence requires no explanation. It makes sense to ask why there are four shrimp on my plate instead of five, because the alternative is all too conceivable. But it doesn’t make sense to ask why two plus two equals four instead of five, because the alternative is incoherent. Nothing could count as two plus two equaling five, so “Why don’t two and two make five?” is no more coherent a question than “Why isn’t MOO?” If the praxeological approach is sound, then demanding to know why the laws of social science are as they are is equally incoherent. That whose alternative is inconceivable requires no explanation.
Our initial problem, then, has turned out on closer inspection to comprise two problems, and so we have to award a double prize. One problem is: why is there a concurrence between the prima facie contents of justice and benefit? The prize for providing the solution to that problem goes to the Austrian delegation. The other problem is: why is there a concurrence between the all things considered contents of justice and benefit? The prize for providing the solution to that problem goes to the Athenian delegation.
And your prize, for having sat through my ruminations on this subject, is to go to lunch.
1. John Rawls, A Theory of Justice, rev. ed. (Cambridge: Harvard University Press, 1999), pp. 3-4.
2. Robert Axelrod, The Evolution of Cooperation (New York: Basic Books, 1984); David Gauthier, Morals By Agreement (Oxford: Clarendon Press, 1986); Leland B. Yeager, Ethics as Social Science: The Moral Philosophy of Social Cooperation(Cheltenham: Edward Elgar, 2001).
3. Among the leading texts in Austrian praxeological theory are: Carl Menger, Principles of Economics, trans. James Dingwall and Bert F. Hoselitz (Grove City PA: Libertarian Press, 1994). Ludwig von Mises, Epistemological Problems of Economics, trans. George Reisman (New York: New York University Press, 1981); The Ultimate Foundation of Economic Science: An Essay on Method, 2nd ed. (Kansas City: Sheed Andrews and McMeel, 1978); Human Action: The Scholar’s Edition (Auburn: Ludwig von Mises Institute, 1998). Friedrich A. Hayek, The Counter-Revolution of Science: Studies on the Abuse of Reason (Indianapolis: Liberty Fund, 1979); Individualism and Economic Order (Chicago: University of Chicago Press, 1948); Studies in Philosophy, Politics and Economics (London: Routledge and Kegan Paul, 1967); New Studies in Philosophy, Politics, Economics and the History of Ideas (Chicago: University of Chicago Press, 1978). Murray N. Rothbard,The Logic of Action, 2 vols. (Brookfield: Edward Elgar, 1997); An Austrian Perspective on the History of Economic Thought, 2 vols. (Brookfield: Edward Elgar, 1995); Man, Economy, and State: The Scholar’s Edition (Auburn: Ludwig von Mises Institute, forthcoming 2003). Israel Kirzner, The Economic Point of View: An Essay on the History of Economic Thought (Princeton: Van Nostrand, 1960); Perception, Opportunity, and Profit: Studies in the Theory of Entepreneurship (Chicago: University of Chicago Press, 1979). Hans-Hermann Hoppe, Economic Science and the Austrian Method (Auburn: Ludwig von Mises Institute, 1995). See also Roderick T. Long, Wittgenstein, Austrian Economics, and the Logic of Action: Praxeological Investigations (unpublished manuscript). Several of these works are available online at: praxeology.net/praxeo.htm.
4. Roderick T. Long, review of Leland B. Yeager, Ethics as Social Science: The Moral Philosophy of Social Cooperation; forthcoming in Quarterly Journal of Austrian Economics (Spring 2003).
5. This sort of argument has been frequently employed by consequentialist libertarians against deontological libertarians in the pages of the libertarian periodical Liberty.
6. See, for example, the eighteenth-century Prussian legal theorist Samuel Cocceji’s influential Dissertationes Proemiales. I am grateful to Rebecca Reynolds for bringing Cocceji’s work to my attention.
7. Does the distinction between principia essendi and principia cognoscendi correspond to Wittgenstein’s distinction between criteria and symptoms? I don’t think so. Criteria and symptoms are both principia cognoscendi; the difference is that a criterion’s status as a principium cognoscendi is logical, while a symptom’s status as a principium cognoscendi is empirical. The fact that criteria are still principia cognoscendi, and not necessarily principia essendi, is one of the crucial differences between Wittgenstein and the verificationists. (In the text I do not use the term “criteria” in any special Wittgensteinian way.)
8. Robert M. Adams, “Divine Command Metaethics Modified Again,” Journal of Religious Ethics 7 (Spring 1979), pp. 71-79.
10. Roderick T. Long, “Equality: The Unknown Ideal”; available online at: www.mises.org/fullstory.asp?control=804.
11. See in particular Mises (1998), Rothbard (1997, 2003), and Hayek (1948).
12. Hayek (1948), II. 9.
13. Hayek (1979), I. 3.
14. Hayek (1948), III. 2.
15. For a fuller defense of these claims, see the sources cited in footnote 3, and in particular my unpublished book manuscript on Wittgenstein and praxeology, an early draft of which is available at praxeology.net/praxeo.htm.