One classic way to think about justice is to think about it in terms of utopia. On this picture, political theorising is about outlining the best possible society we could live in. While this way of thinking about justice sometimes gains and sometimes loses in popularity, I don’t believe it ever has quite lost its intuitive attraction. Appeals to some desirable end state for society play an important role for many contemporary anarcho-capitalists, socialists, and conservatives. (For the latter, the utopia is usually in the past.)

Many philosophers nowadays claim that this chiliastic way of thinking about justice is problematic, as it gives us little guidance for our actual societies. In particular, it’s become common to object that this fruitless type of “ideal” theorising characterises much of modern political philosophy, in particular Rawls’s theory of justice. There are different types of objections here, but I wish to focus on one particular line of objection advanced by Amartya Sen (in The Idea of Justice, Harvard University Press, 2009) and Gerald Gaus (in The Tyranny of the Ideal, Princeton University Press, 2016).

Sen’s Dilemma

We can start from Sen, who develops the problem succinctly. Sen distinguishes two ways of thinking about justice, “transcendental” and “comparative” judgments. A transcendental judgment outlines what the best or perfectly just society looks like. A comparative judgment, on the other hand, is a pairwise comparison between different types of society, ranking them in terms of their relative justice. What we need to solve practical problems, Sen claims, are comparative judgments: we need to be able to say, for example, which of two or more social reforms we ought to choose.

Sen goes on to argue that transcendental judgments are neither sufficient nor even necessary for comparative judgments about justice. However, he claims, mainstream liberal philosophers like Rawls are primarily in the business of providing transcendental theories of justice. If true, this would be a severe indictment of standard ways of thinking about justice. We can formulate the dilemma through three propositions (my formulation):

(1) Philosophical theories of justice ought to be action-guiding.

(2) To be action-guiding, a theory of justice needs to provide us with comparative judgments.

(3) Transcendental judgments are neither necessary nor sufficient for comparative judgments.

Let’s grant the first and second premise. It’s true, of course, that not all theorising in political philosophy needs to be directly, or even indirectly, action-guiding. Thinking about political utopias might be interesting, exciting and valuable all by itself, without being action-guiding. But if this would be all that political philosophy did, it would certainly not merit the attention it currently gets. Defending political philosophy by denying (1) would be a Pyrrhic victory.

The second premise also strikes me as true, though there is a way to overstate what a theory of justice must do. Gaus, for example, says that a theory ought to deliver a “complete” ordering of different social worlds (Tyranny of the Ideal, 40), but this surely is too demanding. Developing a theory of justice which allowed us to judge even the smallest social changes, and put all possible kinds of society into a complete and transitive ordering, would be a super-human task. I tend to think that a successful theory of justice could be quite coarse-grained—e.g., it needs to help us decide between a laissez-faire state or a welfare state, but it needn’t be helpful for deciding between different types of welfare state. Either way, it is clear that a theory of justice ought to provide some comparative judgments.

Are transcendental judgments sufficient?

So this brings Sen’s dilemma down to the last premise. Sen offers a simple, but interesting argument why transcendental judgments aren’t sufficient for comparative judgments. (I set questions of necessity aside.) One of the great strengths of Gaus’s book is that he tries to formalise the problem at this point, which I also find useful, and I will follow Gaus loosely. Let’s say that the object of evaluation in a theory of justice are “social worlds” (Gaus’s term). These social worlds can be actual or possible, feasible or infeasible, past or present.Contra Gaus, I prefer to think of them as social histories, so as to account for non-patterned theories of justice.

In particular, what a theory of justice should provide is an ordering of these social worlds in terms of justice, or more generally, social desirability. Call this a J-ordering of these worlds, and write a>_Jb just in case a is more desirable than b. A theory of justice can deliver more or fewer such judgments. A theory which ranks every possible social world against every other possible social world is complete. As I hinted above, usually significantly less than completeness will suffice in a theory of justice, but I’ll assume completeness for simplicity.

A transcendental judgment about justice is one which identifies the optimal social world u, such that u≥_Jx for any x. Let’s assume that the J-ordering has one, and only one, such element.One of Sen’s pet peeves is the possibility that there might be no optimal, but merely maximal elements. An element a is maximal in case there is no x such that x>_Ja. This is especially a possibility if the J-ordering isn’t complete. It should be clear that knowing the optimal element of the J-ordering gives us little information. As an analogy, if you know who the fastest runner is you still know nothing about the relative order of all other runners. In the same way, if transcendental theories of justice merely identified the optimal social world, they would appear to be rather useless.

However, there is an intuitive way how we might try to derive the full J-ordering from knowing its optimal element. After all, if we know what the perfectly just world looks like, we might think that a world is just to the degree that it resembles the perfectly just world. Gaus develops this idea extensively (Tyranny of the Ideal, 51-5), and I will follow him loosely. What we need is a distance metric d over social worlds. Such a metric measures how similar different social worlds are. If a is more similar to b than to c, then we write that d(a,b)<d(a,c). If u is the perfect social world, then we might suggest that

a>_Jb if and only if d(u,a)<d(u,b)

Put in a graph, we would be in the following situation—what Gaus calls a “Mount Fuji optimization landscape” (Tyranny of the Ideal, 62):

In such a world, the closer we get to the perfect society, the more just our world is. If we are at point B, then moving to A (which is more similar to the perfect world, U) will increase justice. Every further step we take will increase justice. If this were true, transcendental theories of justice would be sufficient to deliver comparative orderings.

It’s not of course easy to develop a convincing measurement of “distance between social worlds”. We can already see a lot of philosophical issues lurking beneath the surface. But in any case, the transcendental theorist could claim, these are epistemic difficulties. There is no principled obstacle to transcendental theories of justice being useful.

The Problem of Second-Best

Unfortunately, as both Sen and Gaus argue, this is a mistaken hope. The problem is that the second-best might be very dissimilar from the best. Sen offers a non-moral example:

a person who prefers red wine to white may prefer either to a mixture of the two, even though the mixture is, in an obvious descriptive sense, closer to the preferred red wine than pure white wine would be (Idea of Justice, 16)

If red wine is not available, we do not instead want a mixture of (say) 90% red wine and 10% orange juice. We will instead prefer a very different drink altogether.

This is also plausible when we think about the perfectly just society. In the perfectly just society, there is presumably no crime, no poverty, and no racism. So there might also be no need for punishment, or redistribution, or affirmative action in our perfectly just society. In the presence of crime, poverty, and racism, however, it is not true that reducing punishment, or redistribution, or affirmative action will make our society more just, even if doing so would make our political institutions more descriptively similar (in some respect) to the ideal political institutions.Gaus’s examples can be found in Tyranny of the Ideal, 14-5, 63-5.

Examples like these are well-known from other areas of philosophy as well. One such vexing problem arises in virtue ethics. Assume that you know how the perfectly virtuous person would act. Should you, who are imperfect, then act as the perfect person would? The answer is no. For example, perfect people have no problem with anger, so they do not need to take anger management classes. But perhaps you do have such a problem, in which case you should take the classes. Much more can be said in the face of this issue, but the seemingly simple advice “make yourself as similar to the perfect person as you can” seems to be insufficient. Sometimes making yourself more similar to the perfect person will make you worse.

Rugged Landscapes

Going back to our problem, it seems likely that we find ourselves in what Gaus calls a “rugged” landscape:

Consider, in particular, the choice characterised by points A, B and U: moving closer to a perfectly just society might make it less just, while moving away from it might make it more just.

Sen claims that in the face of this choice, an ideal-oriented theory of justice is useless. Given that the ideal U is presumably not within our reach, we need to be able to make local comparisons. But as distance to the ideal is not a reliable indicator for justice, an ideal-based theory of justice is pointless.

Sen briefly considers “conglomerate” theories, which provide us with both transcendental and comparative judgments of justice (Tyranny of the Ideal, 16-7). Sen rejects such theories, as their transcendental element would simply be redundant. Similarly, Sen accepts that even the transcendental theorist might sometimes have some insights into comparative questions. But these are mere by-products of an endeavour which is ultimately misguided.

Some Remarks

Sen’s dilemma raises a host of interesting issues, and it is one of Gaus’s book’s great merits to work it out clearly and in a formally sophisticated version. I have some ideas about the issue, but here I merely wish to highlight some initial thoughts and problems.

1. As I have highlighted in my description of the problem of second-best, I would group it together with a long line of similar problems elsewhere in philosophy. In these other areas, however, it’s not clear whether the problem is a fatal one. In virtue ethics, for example, Rosalind Hursthouse has suggested that the view suggests a wealth, perhaps even over-abundance, of action-guiding advice. This comes in the form of virtue-based rules such as “be brave”, “be kind”, “don’t be greedy” and so on. Similarly, even if identifying the ideal might not be directly action-guiding, knowing about the ideal might suggest a number of such non-ideal principles by which we can evaluate the justice of non-perfect social worlds.

The Gaus-Sen line of argument would be bolstered if, in addition to the theoretical line of argument they advance, it could be practically shown that a theory of justice like Rawls’s (say) cannot provide us with any analogous kinds of principles. It’s commonplace nowadays to claim that it doesn’t, but the proof is likely to be in the detail.

2. Gaus states the problem in a slightly different way from Sen, a complication I have so far set aside. Gaus appears to accept that an ideal theory evaluates the justice of different social worlds both on the basis of their “fundamental inherent features” and their distance to the ideal, where neither is reducible to the other (Tyranny of the Ideal, 10-1). He calls such theories “multidimensional”.

One worry I have is that Gaus’s multi-dimensional theories are instable. If a multi-dimensional theory can already build on the “fundamental inherent features” of the social worlds it evaluates, why does this not suffice? It’s useful to consider some possible analogies. Assume that whether someone is a good priest is a function of both the degree to which they resemble Christ, and the degree to which they have the various virtues. This would be a multi-dimensional theory. But the problem is obvious: it’s hard to imagine how these two measurements can come apart. Can you have all the virtues but be very much unlike Christ, or the inverse? But it would seem that you simply resemble Christ the more you are virtuous, as Christ (or who-have-you) is the paradigm of perfect virtue.

Here’s a second example. Could the quality of a novel be a function of both the degree to which it resembles (say) Ulysses, and the degree to which it is enjoyable, well-written, etc.? I am not sure this would ever be a good theory. More importantly, it seems to me, one of the two elements will tend to crowd out the other. First, we might compare a novel to Ulysses simply because Ulysses is a paradigm of the qualities we value in a novel—in which case “similarity to Ulysses” drops out as an independent factor. Second, we might care about a novel being enjoyable, well-written, etc., simply because these are good proxies of a novel being similar to Ulysses, which we ultimately care about. In this case, it’s not really the intrinsic features of a book which make it good.

3. Gaus is likely to respond that such theories are no longer ideal (or multi-dimensional) theories, as they make reference to an ideal explanatorily redundant. Instead, such theories merely use the description of an ideal social world as a heuristic device to develop the philosophical principles by which we ought to evaluate different social worlds. Fair enough. In this case, the Sen-Gaus critique would then only apply to theories of justice in which reference to the ideal social world is an explanatorily ineliminable part of those theories.

But note that this raises the bar of proof. It’s not objection enough against a theory of justice that it centrally invokes the notion of the perfectly just society (or the well-ordered society, or what have you). We must also show that such a theory of justice cannot be restated without such a reference. I suspect that many of the theories which Sen and Gaus allege to be transcendental theories on closer inspection turn out to be mixes of purely transcendental, transcendental-as-heuristic, and comparative bits of theorising. Either way, I again expect that the proof is in detailed attention to the work of particular philosophers. This is not to denigrate the Sen-Gaus line of objection, but to note its incompleteness.

4. I have also talked little about the alternative, “comparative” approach to thinking about justice. This is partially because there is probably no unified method that comparative theorists will pursue. One thing useful to highlight, however, is that a comparativist need not be a particularist. It’s easy to get the impression that someone who develops a non-transcendental theory of justice is also someone who disavows general principles of justice—Sen’s repeated emphasis on “pairwise comparisons” sometimes supports this impression.

But this would be a mistake. A non-comparative theory of justice might be quite principled and general. Take a simplistic, libertarian theory of justice, on which the injustice in a society is a weighted function of the number of rights violations in it. This is a perfectly general, abstract theory, but it isn’t a transcendental theory of justice, as no reference is made to a perfectly just society (or at least, any such reference would be eliminable).