It is exceptional for a primarily critical contribution to have a major impact on the philosophical literature, but such was clearly G.A. Cohen’s achievement with the sustained objections to John Rawls’s theory of justice that he mounted over more than a decade, and ultimately brought together in his book Rescuing Justice and Equality. In part, this outsized impact can be traced to Cohen’s extraordinarily probing and articulate way with philosophical arguments. Mainly, though, his arguments were influential simply because they raised deep and difficult questions. When Cohen challenged Rawls’s claim that principles of justice apply exclusively to social institutions, he forced us to consider the extent to which personal choices can be insulated from the demands of distributive justice. When he questioned how Rawls’s principles of justice were tailored to specific empirical conditions, he prompted a broader investigation into the relation between normative principles and empirical facts. No shoddy philosophical issues, these.

Yet there was one point on which Cohen’s critique could be somewhat frustrating. He often insisted that Rawls had misunderstood what a theory of justice was fundamentally about.^{Footnote 1} Famously, Rawls takes justice to be the first virtue of social institutions; the aim of his theory is to defend principles that tell us how, all things considered, such institutions should be organized. For Cohen, this whole setup signalled conceptual confusion. To reach his two principles, Rawls explicitly took into account values such as efficiency and feasibility—values that, although undeniably important, are clearly distinct from justice itself. Indeed, it’s only because he took these other values into account that Rawls could plausibly reach principles concerning how society should be set up all things considered. But then, Cohen asked, how could the result be principles of justice? Justice, efficiency, and feasibility are different values. We must first understand what justice is; only then does it make sense to ask how the demands of justice should be modified in light of values such as efficiency and feasibility. Throwing all these values into one big mix and claiming that, by doing so, we can reach conclusions about justice is just muddying the waters.

It should be obvious why I said that this part of Cohen’s discussion could be frustrating. It doesn’t really matter how we use the term ‘justice.’ What matters is that we apply whatever definition we settle on consistently. Rawls does this, as does Cohen. They use different definitions, of course, so they’re ultimately talking about different things. But there the story ends, it seems. If Cohen prefers to use the phrase ‘rules of regulation’ to refer to Rawls’s subject matter, that’s fine.^{Footnote 2} It may still turn out that these ‘rules of regulation’ are of great interest. Perhaps it will even turn out that we can establish conclusions about these rules that help answer some of our more pressing questions about how to organize society, and that do so better than any discussion of Cohen’s abstract sort of justice could. Either way, nothing need be lost in the vocabulary adjustment.

It’s a notable merit of Kyle Johannsen’s impressive monograph A Conceptual Investigation of Justice that it confronts this problematic aspect of Cohen’s critique head-on. Johannsen aims to show that, although the dispute between Rawls and Cohen on the nature of justice is partly conceptual—although the two authors advocate using the term ‘justice’ to refer to different concepts—it is not merely verbal, since issues of philosophical substance underlie the disagreement.^{Footnote 3} At the most fundamental level, Johannsen suggests, Rawls and Cohen disagree about the conception of fairness that should feature in our theorizing about justice: Rawls favours a strictly procedural conception, Cohen a substantive one. That’s where their conceptual disagreement stems from. Equating justice with the outcome of the original position, as Rawls urges us to do, is plausible only if we’re operating with a strictly procedural conception of fairness; on a more substantive conception, Johannsen argues, the equation simply can’t hold.

This suggests the possibility of a particularly forceful line of argument against Rawls. What if we could show that his strictly procedural conception of fairness is too weak for his own purposes? What if we could show that a plausible construal of the veil of ignorance can only be reached if we bring in something like Cohen’s conception of fairness? Then, as Johannsen sees it, we would have shown that Rawls can’t equate the outcome of the original position with justice. And, as a bonus, we would have shown that his view must take on board much more of luck egalitarianism than he himself would have ever allowed—to the point where we might consider amending his famous slogan to ‘justice as luck egalitarian fairness.’

This is the most ambitious strand of argument running through Johannsen’s book. It’s an exciting and original attempt to unearth an unsuspected, potentially devastating power in Cohen’s offensive against Rawls. But I don’t believe the attempt succeeds. At a conceptual level, I don’t think the possibility of equating justice with the outcome of the original position ultimately depends on our conception of fairness in the way that Johannsen suggests. And, at a more substantive level, I suspect that grafting Cohen’s luck egalitarian conception of fairness onto Rawls’s framework wouldn’t yield the fruit that Johannsen promises; if anything, it would probably kill the plant. Before I say more about these qualms, let me explain briefly why Johannsen thinks that Rawls’s conception of fairness lands him in trouble.

The idea of fairness plays a key role in justifying our conception of the veil of ignorance, which itself crucially shapes the choice of the principles that takes place in the original position.^{Footnote 4} Johannsen thus strikes at the very heart of Rawls’s theory when he contends that this whole framework can’t get off the ground on the conception of fairness that Rawls himself favours. The idea, I take it, is the following. Rawls wants to operate with as thin a conception of fairness as possible—a conception closely linked to minimal ideas of reasonableness and of reciprocity.^{Footnote 5} But, Johannsen contends, such ideas can’t justify the conception of the veil of ignorance that Rawls ends up endorsing. In particular, the thin conception of fairness can’t decide between two ways to construe the veil of ignorance. The first is the one that Rawls favours: the veil conceals all information about particular matters, and we think of the parties in the original position as choosing based on individual rationality. On the second construal, the veil would let through some information about particulars—information about which talents a person has or about her social position, for instance—but on the condition that the parties use that information in a reasonable manner.^{Footnote 6}

How are we to choose between these two possibilities? Johannsen thinks there is an obvious way to break the tie in favour of the first option, and thus to reach Rawls’s favoured conception of the veil of ignorance: we just have to take on board the luck egalitarian conception of fairness. As he puts it,

In short, Johannsen believes that the strongest version of Rawls’s theory is one that construes fairness along luck egalitarian lines. This doesn’t mean that Rawls’s theory just morphs into luck egalitarianism, or at least that’s not what Johannsen thinks; the concession concerns only the way in which the original position is set up. The principles we obtain from the original position may change slightly, of course, but the result is supposed to be a strengthened version of Rawls’s theory, not a complete metamorphosis.

As I’ll explain below, I’m not sure that Johannsen’s proposed luck egalitarian graft can take; and, if it does take, I’m not sure it can be stopped from completely taking over the theory. But first I want to raise some questions about Johannsen’s claim that the disagreement between Rawls and Cohen boils down to whether we should adopt a procedural or a substantive conception of fairness. There is something amiss in the way Johannsen characterizes the issue. To begin, note that his claim can’t be about how we should understand fairness in general; what’s at stake here is specifically how we should understand fairness for the purposes of laying out the original position.^{Footnote 8} The question is then, what does the contrast between substantive and procedural fairness amount to in that specific context? As I understand it, it refers to the manner in which a given conception of fairness is deployed—not to its content per se. Basically, if we apply our conception of fairness as we lay out the original position, and if we are satisfied that whatever comes out of this choice procedure must be acceptable to all, then we are applying our conception procedurally. By contrast, if we use our conception of fairness to evaluate the outcome of the original position—if we view the outputs as directly answerable to our standard of fairness—then we are applying it substantively.

Rawls’s hope in setting up the original position is that we can do so in a way that allows our application of the idea of fairness to be mainly procedural. If we succeed, the original position will do the work for which it is intended. But presumably that is a hope that we may also entertain if we are drawn to a luck egalitarian conception of fairness: we may still hope that our conception of the original position will be sufficiently powerful to allow us to apply our idea of fairness procedurally.^{Footnote 9} It will, of course, be possible to ask whether the outcome of the original position lives up to our standards of fairness, and thereby to deploy the idea substantively. But that’s also something we can do on Rawls’s own view—indeed, it is something we must do to reach reflective equilibrium.^{Footnote 10} There is an important qualification, however. On a properly Rawlsian view, we should do this only to the extent that the judgements of fairness to which we appeal qualify as considered judgements. Since this won’t be the case for most of our judgements of fairness, it follows that, on either version of the view, the substantive deployment will remain relatively modest.

If these points are correct, they cast doubt on Johannsen’s contention that adopting the luck egalitarian conception of fairness (or any substantive conception of fairness, for that matter) must lead us to reject Rawls’s claim that justice just is what comes out of the original position. The reasoning here is not as detailed as it might be, but I take the main idea to be that the principles chosen in the original position can’t fully live up to the demands of fairness, since they are the result of a compromise that takes account of other values; accordingly, if we’re operating with a conception of fairness that allows for a direct evaluation of the outcome of the original position, then we can’t possibly view this outcome as identical with justice.^{Footnote 11}

This can’t be quite right. As we just saw, Rawls’s view of fairness is also used to evaluate the outcome of the original position directly. Although our main deployment of the idea of fairness comes when we design the original position, we must also confront its outcome against our considered judgements of fairness to achieve reflective equilibrium. So it can’t be Rawls’s conception of fairness that pushes him to view the outcome of the original position as principles of justice. Rather, it’s because he starts from the understanding that principles of justice are rules that social institutions should follow, all things considered, that he takes the outcome of the original position to consist of such principles. If he started from the view that justice is identical with equality or fairness, as Cohen does, then he would think of the outcome of the original position differently. In spite of Johannsen’s best efforts, then, I find myself drawn back to the traditional construal of the disagreement between Rawls and Cohen about the nature of justice—and to the worry that Cohen’s arguments on this point fail to get at anything of substance.

Even if I am right about all this, Johannsen could still be correct to claim that understanding fairness along luck egalitarian lines makes for an overall stronger version of Rawls’s theory. But I have misgivings about this part of his argument as well. Let me start with a very general one. As I just explained, I don’t think the disagreement between Rawls and Cohen really turns on how substantive or procedural our conception of fairness should be. A different way of understanding the contrast would be more accurate, but also less flattering for the luck egalitarian side. What motivates Rawls to adopt what Johannsen sees as an excessively thin conception of fairness is not that he wants to keep the conception procedural; it’s that he wants to keep the view as uncontroversial as possible. And he proposes to do that by tailoring his conception of fairness to the problem at hand—that is, by ensuring that his conception of fairness does not commit him to more than necessary.

The key thought receives a particularly clear formulation in T.M. Scanlon’s illuminating discussion of Rawls’s views on justification:

This underscores a crucial respect in which Rawls’s conception of fairness is circumscribed: he is asking what would be a fair procedure for choosing principles of justice as between persons who disagree about certain specific matters. This is a much more modest aim than trying to put forward a general theory of fairness.

A possible rejoinder is that, although Rawls’s aim has the advantage of being modest, the luck egalitarian conception of fairness has the advantage of being systematic. And, of course, as we saw already, Johannsen would add that, in any case, we need a stronger conception of fairness to reach the conclusion Rawls wants. Once we are clear about how Rawls uses the idea of fairness, however, it becomes hard to see what force these objections have. The question that allegedly poses a problem for Rawls is whether allowing the parties in the original position knowledge of their talents and social position would be fair. But once we interpret the idea of fairness to mean fair as between people who disagree about how economic goods are to be distributed, I am unsure what the difficulty is supposed to be: that allowing such knowledge wouldn’t be fair seems entirely uncontroversial. Certainly—and this is the main point here—the claim by itself is considerably less controversial than the full-fledged luck egalitarian theory of fairness.

So the worry is that, by bringing in luck egalitarian fairness, as Johannsen advises us to do, we would be appealing to a principle that is far more controversial than the conclusion it is intended to support. This kind of move can be warranted, of course—sometimes, the greater systematicity is worth it. But it’s unclear why that would be the case when the task is to set up the original position. We need to determine what restrictions on the information available to the parties in the original position would be fair to all members of society. If it’s uncontentious that the availability of information about talents and social positions would be deemed unfair by people who hold different views about how economic goods should be distributed in their society, then it’s unclear what there is to gain from bringing in a more systematic theory of fairness.^{Footnote 13}

I have been pointing to some very general reasons for being sceptical of Johannsen’s claim that Rawls’s theory would be strengthened by the addition of the luck egalitarian conception of fairness. I now want to argue that the proposal runs into a more specific and immediate problem, namely, that the luck egalitarian conception of fairness can’t actually do the work that would be required from it in the context of Rawls’s theory. Let me explain. Johannsen’s contention is that luck egalitarian fairness would clearly yield the more robust veil of ignorance we seek, since “[o]ne’s talents and social position are largely traceable to brute luck.”^{Footnote 14} But if we really take the logic of luck egalitarianism seriously, can we just stop there? Don’t we need to ask to what extent exactly talents and social positions are traceable to brute luck? And shouldn’t the parties in the original position have information about their social position precisely to the extent that it results from their choices? If this information is completely blocked, aren’t we allowing people who’ve made all the wrong choices to benefit from the careful thinking and self-control of those who’ve made all the right ones—an affront to everything that luck egalitarians stand for?

Admittedly, figuring out the contribution of choice to a person’s lot in life is no easy matter, but it would be odd for luck egalitarian sympathizers to find comfort in that thought. In any case, let’s put that worry to one side: let’s assume that we can rely on a group of solid, Oxford-trained experts. We still encounter a serious problem: if the parties in the original position are granted access to this kind of information, then it becomes unclear what principles they can agree on. If they know the extent to which their choices have improved or worsened their social position, then those whose choices generally turned out well will presumably favour more responsibility-sensitive principles. Those whose choices turned out badly, however, will want to push in the opposite direction. How an agreement could be reached is hard to see—unless, of course, we assume that the parties in the original position adopt luck egalitarian fairness as their principle of choice. But if we do that, there will be nothing left standing of Rawls’s theory—we’ll have transformed it into a strangely ornate version of luck egalitarianism.^{Footnote 15}

Perhaps we should say, in a Rawlsian spirit, that the parties in the original position should know to what extent their social position has been affected by their choices (30%, say) but not whether it was for better or worse. This may help prevent the kind of stalemate I just described, but only at the cost of creating other problems. Most obviously, we run into a problem of coherence, since the information we propose to allow the parties to access is entirely arbitrary from a luck egalitarian point of view. After all, if the effect of choice on my position is not morally arbitrary, why should the veil of ignorance conceal any part of that information? Even if we leave that concern aside, I suspect that a problem of indeterminacy will remain, since it’s still unclear what principles the parties should choose. Johannsen seems to think that they would choose principles that are more responsibility-sensitive than Rawls allows. But it’s unclear on what grounds they would do so if we don’t assume that they themselves accept something like luck egalitarian fairness. Nor is it clear why they should all choose the same principles, given that the extent to which their social positions were affected (for better or for worse) by choice will vary considerably. If we stick to the sort of reasoning that Rawls envisages taking place in the original position, what is most likely is that we will fail to obtain a solution.

The upshot of all this is that we can’t take Rawls’s approach more or less as it is and simply plug in the principle of luck egalitarian fairness. Doing so renders the outcome of the initial situation indeterminate. And if we take the obvious way out—if we assume that the parties in the original position base their choice on the luck egalitarian conception of fairness—then although the outcome of the original position is determinate, it isn’t particularly interesting, as it just reflects the old adage ‘luck egalitarianism in, luck egalitarianism out.’ The real question is then whether we want to rest our entire theory on the luck egalitarian conception of fairness in this way. Much as I am impressed by the detailed arguments and innovative proposals that Johannsen puts forward in his thoughtful book, I remain unconvinced that this would be a wise move.