I. Ideal Theory and the Circumstances of Politics

Inspired by Rawls’s work, many political philosophers aim to develop an ideal theory of justice, or at least, part of one. Ideal theorizing has been criticized on a number of grounds, however. In this section I focus on what I think is one of the most serious challenges to it, which has emerged from a realist perspective, namely, that ideal theory is founded on the assumption that there is convergence on principles of justice, and that this assumption not only deprives ideal theory of any practical relevance but also betrays a misunderstanding of the nature of the political: disagreement over conceptions of justice is part of the enduring circumstances of politics; ^{Footnote 5} insofar as ideal theory assumes convergence on a particular conception, it denies a constitutive feature of the political. ^{Footnote 6}

I want to avoid getting drawn into an unprofitable discussion of the nature of ideal theory, so I will simply stipulate that the aim of an ideal theory of justice is to give an account of a society that is fully just, subject to the constraint that this account be capable of playing some role in guiding political action, with the precise nature of that role standing in need of further elucidation. ^{Footnote 7} I will assume that a fully just society is one in which the basic institutions realize whatever the correct or best principles of justice are that apply to them, and in which its members comply with whatever the correct or best principles of justice are that apply to their behavior. I will also assume that because ideal theory has the ultimate aim of playing a role in guiding political action, it must employ a feasibility condition of some kind. Although there is room for reasonable variation in terms of the formulation of that condition, ^{Footnote 8} an account of a fully just society must at least be compatible with hard constraints — that is, constraints on our capacity to realize principles that are part of the human condition because they are rooted in physical laws or other unchangeable features of the natural environment (such as scarcity of resources), or in aspects of human nature that are fixed. ^{Footnote 9}

The realist challenge that I am interested in does not deny that there may be correct or best-defended principles of justice. Instead it argues that disagreement over principles of justice will persist even under favorable conditions by appealing to empirical claims about the sources of disagreement over these principles, in particular, to the way in which disagreement is rooted in what Rawls calls “the burdens of judgement” ^{Footnote 10} and in a variety of nonrational causes. A realist who presents the challenge I am envisaging holds that there are no grounds for thinking that citizens will come to endorse the same principles solely on the basis of reasoned argument, nor that the nonrational causes that influence their judgments about what principles to accept will lead them to converge on the same set. ^{Footnote 11} From the claim that disagreement over political principles will persist, and is a source of conflict between citizens, he or she concludes that one of the main purposes of political theory is to guide us in the face of it, or at least to enable us to reflect upon its normative significance. ^{Footnote 12} A vision of a fully just society where principles of justice are universally shared might provide reasons for action for philosopher-rulers who inspire devotion and whose authority is unquestioned, but it does not provide reasons for action for real states or ordinary citizens living in a democratic society mired in deep and often fractious disagreement.

In response to this challenge it might be argued that ideal theorizing is not intended to generate practical recommendations by itself. Rather, its main purpose is to enable us to develop a nonideal theory that can lead to such recommendations. That may be true, but it doesn’t answer the challenge I am considering, at least when it is presented in its most powerful form. Defenders of ideal theorizing about justice think that it can play a role in guiding political action because the vision of a fully just society that it is in the business of working out can serve as a target at which to aim, or at least as a standard against which to judge the extent of injustice, ^{Footnote 13} which, when combined with knowledge of what is feasible and the probabilities that different courses of action will achieve their aims, together with an assessment of the likely costs of each, can provide reasons for political action, such as a reason for favoring one particular reform over another. The realist position I am considering maintains that a vision of a fully just society in which everyone converges on the same principles of justice cannot coherently serve as a target, or even as a standard that guides action, because disagreement is part of the enduring circumstances of politics.

One reply here might be to challenge the realist’s factual prediction: for example, one could try to show that disagreement to a large extent has its roots in nonrational causes, the power of which are likely to diminish over time as our understanding of the issues improves. I have doubts about the prospects of success of a response of this kind. Alternatively, one might defend a milder version of the claim that convergence is achievable that is more plausible empirically. For example, in Political Liberalism Rawls suggests that under favorable conditions citizens will tend to converge on a set of reasonable liberal conceptions of justice. ^{Footnote 14} Although citizens may continue to disagree over which reasonable liberal conception to endorse, it might be thought that they will each come to hold one such conception. But that claim could also be regarded as empirically implausible, unless the notion of a reasonable liberal conception is understood in a way that makes it very inclusive. ^{Footnote 15} And even if it is understood inclusively, it nevertheless seems likely that even under favorable conditions a significant minority of citizens will endorse unreasonable illiberal (or nonliberal) conceptions of justice.

Do ideal theorists need to assume convergence on a vision of a fully just society and, if so, does this assumption create a problem for them in the way that some realists think it does? It might seem that ideal theorists cannot avoid making an assumption of this kind given that they aim to identify the principles that would govern a fully just society, and a fully just society is one in which there is full compliance with principles of justice. Strictly speaking, however, full compliance with principles of justice does not require there to be convergence on those principles. The citizens of a state may comply with a principle, or a law that embodies it, despite disagreement with it: they may regard this principle or law as binding on them, perhaps because they regard the procedures that led to its adoption as authoritative; or they may recognize self-interested reasons for complying with the law, for example, they may be deterred from breaking it by the prospect of being punished. ^{Footnote 16} Viewed in this light, the Rawlsian conception of a well-ordered society in which citizens actually endorse the principles that govern it goes beyond full compliance and incorporates an element that is unnecessary for an ideal theory of justice.

In response it might be said that a society in which citizens not only comply with the correct principles of justice but also endorse them would be more just, so an ideal theory of justice must assume convergence. In such a society everyone would act justly, that is, they would act from the true motive of justice; no one would comply with principles of justice merely because of the threat of sanctions. Indeed it might be thought that Rawls assumes convergence in a well-ordered society precisely because he wants to give an account of a fully just society that could be stable for the right reasons. But even if we suppose that a society in which citizens act from the motive of justice would be more just because it would be stable for the right reasons, this does not seem to imply that a fully just society requires convergence on the very same principles. A society might be fully just even though some comply with just laws and policies merely because they recognize the just authority of the decision-making procedures that govern them and despite believing that the majority of their fellow citizens are mistaken in their view that these laws and policies are just.

Insofar as an ideal theory does assume convergence, it would be best to regard this as an idealization — that is, as an assumption that may not hold true in any actually existing society, or indeed any feasible society, but which is made in the process of building a theory in order to simplify its construction. ^{Footnote 17} For example, Rawls assumes that all members of a society are “rational and able to manage their own affairs” and that the subject matter of a theory of justice is the basic structure of society “conceived as a closed system isolated from other societies.” ^{Footnote 18} The potential problem with these assumptions is not that they describe a state of affairs that is hard or impossible to bring about (or that would require a massive violation of individual rights in order to do so), but rather that they may have the effect of creating too large a gap between the theory and the societies to which it is meant to apply for it to be capable of playing a role in guiding political action. ^{Footnote 19} So does the assumption of convergence create too large a gap between the theory and the circumstances to which it is meant to apply for it to be capable of doing so?

This idea of action guidance is vague: let us say that an ideal theory cannot play any role in guiding us if it cannot provide us with any reasons for action whatever available empirical knowledge is combined with it. There are several ways in which a principle that is part of an ideal theory might provide us with a reason for action when combined with empirical knowledge: for example, it may provide a reason here and now for performing an action it prescribes, or it may provide a reason for bringing about a state of affairs in which we could perform that action. And there are different ways in which a principle might provide a reason for action without that reason being conclusive: there may be several principles in play in a given context that give us conflicting reasons for action; and these principles may reflect or express different values, or they may reflect or express different aspects of an internally complex value. But so long as an ideal theory when combined with empirical knowledge can provide some reason for action in the circumstances we now confront, then that would seem a sufficient response to the charge that it cannot play a role in guiding us, even if this reason is outweighed by others. ^{Footnote 20} So how might an ideal theory that assumes convergence on principles of justice play such a role?

Even if disagreement and conflict that is rooted in it are a constitutive feature of politics, political activity need not always be limited to maintaining order in the face of it. There is generally room in modern democracies for the pursuit of other ideals and ordinary citizens as well as politicians need to form a view concerning what policies they should advocate and support. This is where an ideal theory of justice seems to have a potential role to play in guiding action, while at the same time deepening our understanding of the issues and making possible a fuller justification of our practical decisions. For example, an ideal theorist can work out in more depth the visions of a just society that citizens may already have within their sights and assess their strengths and weaknesses. The conclusions she reaches, when combined with empirical judgments concerning what is politically feasible, what arrangements are likely to be stable over time, and what costs would be incurred in instituting those arrangements — together with reflection upon the normative implications of noncompliance — may then provide her with a reason to support or reject particular policies, a reason to defend those policies in public debate, and a reason to vote for particular representatives in elections. Ideal theory, when it is founded on an assumption of convergence on principles of justice, is limited in the guidance it can provide for dealing with persistent and widespread disagreement over what justice requires when that disagreement leads to noncompliance with laws and policies. But ideal theorists would regard the inquiry into the proper principles for governing a society where that occurs as part of nonideal rather than ideal theory.

Realists are likely to argue that my response to the challenge posed by intractable disagreement over principles of justice fails to engage at a sufficiently deep level with it. Indeed this appears to be what lies behind Matt Sleat’s insistence that realism is not simply a version of nonideal theory. ^{Footnote 21} For him, the idea that there needs to be a division of labor between ideal and nonideal theory misses the point that we don’t need ideal theory to reflect upon the political problems that confront us, and that in assuming away disagreement ideal theory fails to appreciate that such disagreement is a constitutive feature of the political. I shall return to the issue of whether we need ideal theory to justify our practical conclusions and, if we don’t, what its point might be, later in this essay. With regard to the nature of the political, ideal theorists can accept that disagreement is the stuff of politics without subverting what they do: they can insist that ideal theory can provide practical guidance when combined with knowledge about what is feasible, while conceding that it is nonpolitical in an important way. ^{Footnote 22} In order to be plausible, the envisaged division of labor between ideal and nonideal theory would involve giving the latter a considerable degree of autonomy in relation to the former, and not merely because nonideal theory will need to pay attention to empirical research in order to identify what courses of action are feasible. ^{Footnote 23} Deep disagreement is not merely a constraint on the realization of principles of justice, but raises very different kinds of normative questions. When nonideal theory takes it into account, the question for it would not simply be: “What way of proceeding in the face of the fact that people will inevitably disagree with the principles delivered by an ideal theory will enable us to make the most progress toward a fully just society, in the light of what is feasible and the costs of implementation?” It would also need to focus on questions such as the following, and the way in which they arise in particular contexts: How should we respond to those fellow citizens who hold (what we believe to be) unreasonable conceptions of justice and what procedures should be constructed to deal with such disagreements? What reasons, if any, can be given to those who do not regard political procedures as authoritative to comply with the outcomes of those procedures when they make “demands for legitimation”? ^{Footnote 24} How should we respond to those who, we believe, unreasonably reject those procedures?

II. Principles and Their Consequences

Even if some forms of theorizing about justice had no practical relevance, they might nevertheless be thought to provide intellectual or philosophical illumination, in much the same way that within metaphysics or epistemology a theory of truth or a theory of knowledge might do so. ^{Footnote 25} Ideal theory, when it is defined in the way that I have done, must have practical relevance if it is to fulfill its purpose, but we might follow Alan Hamlin and Zofia Stemplowska in drawing a distinction between ideal theory and the theory of ideals, with the latter encompassing abstract normative or evaluative theorizing that is not constrained in any way by a need to play some practical role. ^{Footnote 26} Accordingly, my aim in this section is to defend, against a different realist challenge, the coherence and value of theorizing about ideals when that simply aims to help us to understand an ideal better. This form of theorizing might enable us to identify principles for ranking logically possible states of affairs in terms of how just they would be, but it would do so without any regard to the practical implications of these principles for the society we inhabit, what institutions would be required to realize them, or the feasibility of implementing them.

Some realists, such as Edward Hall, would reject this mode of theorizing because they maintain that we cannot properly assess normative principles without considering what they entail, and in order to assess them “we must consider how principles map on, or apply to, the factual contexts which they aim to govern.” ^{Footnote 27} Elizabeth Anderson, although not a realist in any conventional sense, goes further in one respect, maintaining that ideals need to be put into practice before they can be properly assessed. In her view, ideals offer solutions to problems that we experience and they need to be tested by “seeing whether they solve the problems for which they were devised, settle people’s reasonable complaints, and offer a way of life that people find superior to what they had before.” ^{Footnote 28}

Hall is surely correct that some principles, particularly those designed for the immediate purpose of governing institutions and policy, cannot be properly assessed without considering the effects of putting them into practice. His contentious claim is that this is true more generally of all normative principles. ^{Footnote 29} Indeed, Hall’s thesis when it is understood as a universal claim about normative principles has an air of paradox to it, for how are we to assess the effects of putting a principle into practice without appealing to more fundamental normative principles that are not assessed in this way? Surely an assessment of the effects of putting a principle into practice will at the very least involving determining whether it treats people as equals, whether, for example, it is consistent with the more fundamental principle that each person’s life matters equally; yet we cannot insist without circularity that this more fundamental principle be assessed in terms of the consequences of putting it into practice.

Even if I am right that there are fundamental principles that cannot be assessed by considering the effects of implementing them, we might think that these principles can always be interpreted differently, and that selecting the best interpretation of them will involve considering those effects. If it is proposed that the best interpretation of the principle that we should treat people with equal concern and respect is a luck egalitarian principle according to which no one should be worse off than another through no fault of her own, then we cannot assess whether this interpretation is correct except by considering the consequences of implementing it. Indeed it might be thought that even principles such as this one will be indeterminate until they are embodied in institutions and policy, lending support to Anderson’s more radical contention that we need to test ideals by actually putting them into practice. For this reason it might even be regarded as a mistake to think of institutions simply as the means through which principles are realized since institutions are always partly responsible for giving content to principles. William Galston expresses this idea in the following terms: “institutions provide arenas within which abstract concepts of principles and aims (rights, the general welfare) are worked up into concrete conceptions. As such, they help define the community’s purposes, rather than simply putting prior understandings into effect.” ^{Footnote 30} This line of thought perhaps receives some support from Wittgenstein’s considerations concerning rule-following. ^{Footnote 31} Although his remarks on rule-following are complex, and their import is a matter of scholarly debate, his observation that any rule can be interpreted differently, and his apparent view that what determines the correct interpretation of it are the practices in which it is embedded, have clear resonances here.

But even if fundamental principles, such as the principle that we should treat people with equal concern and respect, admit of different reasonable interpretations, there are some interpretations that are ruled out as unreasonable because they cannot be reconciled with our intuitions about what abiding by this fundamental principle involves. We can often probe these intuitions by considering imaginary cases, and in this way give more content to the principle. Indeed, specific interpretations of the fundamental principle that we should treat people with equal concern and respect, such as the luck egalitarian principle, might also be refined and assessed in this way. Elizabeth Anderson is skeptical about thought experiments when assessing ideals and principles, on the grounds that evaluative judgement involves expressing feeling, and we are not particularly good at predicting our feelings about states we have not yet experienced. ^{Footnote 32} While accepting that our intuitions sometimes differ from what we expect when we are confronted by real rather than imaginary cases, it seems to me that we should not exaggerate this truth. No doubt it gives us some reason to worry about the assessment and refinement of principles solely by considering imaginary outcomes, but it does not give us reason to think that this way of proceeding has no legitimate role to play.

Indeed, consideration of imaginary outcomes may be the best way, or even the only way, of determining whether it is the case that a principle has some force but is outweighed by another principle, or whether instead it is an incorrect principle and should be rejected. ^{Footnote 33} That a principle would have bad consequences if implemented is rarely a conclusive reason for rejecting it. Consider the principle “from each according to his ability, to each according to his need.” Hall argues that if the implementation of this socialist principle of justice would require an ethos that could only be sustained by repressive state intervention because it goes against the grain of human nature, then this would give us sufficient reason to reject the principle. ^{Footnote 34} But there are other conclusions that might be drawn here. For example, we might suppose that this principle needs to be balanced against another principle of justice that requires us to respect basic liberties (or to maximize equal basic liberties) and which is outweighed by the latter. If that were the correct diagnosis, the socialist principle would still possess independent force. Whether this is the correct diagnosis could be “tested” by asking what the principle would imply in relation to a range of cases, some of which are unlikely to obtain, where it could be implemented without repressive state intervention perhaps because human beings were differently motivated, and seeing whether its implications for those cases match our considered judgments. It is hard to see why these reflections would be incoherent, even though such circumstances might be far removed from “the factual contexts” that the principle is to govern.

The idea that some legitimate forms of normative theorizing, including theorizing about justice, might involve identifying principles and then testing them against our intuitions about imaginary examples in a way that is independent of any consideration of how those principles could be realized through institutions, has the implication that when we turned to that question we might find that it is impossible in practice ever to realize these principles, and that they could never provide reasons for individual or collective action. This might make us start to wonder whether theorists engaged in this activity are investigating a rather different concept from those whose work is motivated by practical concerns. ^{Footnote 35} Indeed, could it be argued that those engaged in this purely intellectual activity are no longer in the business of theorizing justice at all, even if they use that term? It might be claimed that the point of the concept of justice is to guide political action, in the sense distinguished earlier, which would seem to imply that a judgment cannot be a judgment of justice unless it can guide political action. This might even be taken to support the view that adequate principles of justice in particular must meet some sort of feasibility condition, even if normative or evaluative theorizing in general is not subject to such a condition. ^{Footnote 36}

This is a difficult argument to finesse since we need some way of justifying the claim that the concept of justice has an inherently practical purpose. It is most plausible when it is understood as a claim about our ordinary practice of using the term, for example, when we say “these welfare reforms are unjust,” or “the protesters were dealt with unjustly” and so on, where the point of making these claims is to persuade us of the importance of some action in response. ^{Footnote 37} But even if we can show that the concept of justice has a practical purpose in our ordinary use of it, and that theorizing about the requirements of justice is inherently an exercise in practical reason, it is hard to know what follows from it. ^{Footnote 38} If the identity of a concept is fixed by its point, it might seem to follow that those who defend normative or evaluative principles that can give no practical guidance, whether now or in the foreseeable future, must be employing a different concept. But that would be misleading if it was then taken to imply that there is no connection at all between “justice” in its practical sense and “justice” in its purely theoretical sense. Although it seems unobjectionable to suppose that theorists may be engaged in very different activities — some are seeking to understand “justice as such” in a way that is independent of any concern for what is feasible and the costs of implementation, whereas others are seeking to provide principles of justice to govern a possible society — we cannot avoid the issue of what the relationship is between these two forms of activity. Regarding these forms of activity as focused on different concepts that happen to be denoted by the same term would encourage the view that they are entirely unrelated, yet this seems mistaken. ^{Footnote 39} It will be denied by many of those who see themselves as trying to understand justice as such, for they will argue that this form of reflection is logically prior, because without it we won’t fully understand the principles that are being proposed for governing an actual or possible society, or be able to justify these principles fully. If they can establish this claim, it would in effect turn the tables on the realist.

III. Feasibility and the Justification of Action-Guiding Principles

Why might it be thought that the full justification of action-guiding principles relies on first identifying fundamental principles of justice that provide us with an understanding of justice as such, without any concern for what is feasible or what costs would be incurred in implementing them? G. A. Cohen has argued that ultimate normative principles, including ultimate principles of justice, are not constrained by facts about human incapacity. When a fact about human incapacity “is said to exclude a principle because it can’t be obeyed, we may then ask what we should say about the putatively excluded principle on the counterfactual hypothesis that it could be obeyed. And it is only when we thus clear the decks of facts about capacity and get the answer to that counterfactual question that . . . we reach the normative ultimate.” ^{Footnote 40} In other words, we reach ultimate normative principles in these cases by abstracting from what is feasible, employing a chain of counterfactual reasoning that involves asking questions such as, “If it were possible to implement this principle, would justice tell us to do so?” or “If it were possible to implement the principle, would an injustice be involved in not implementing it?” The relevant ultimate principles do not express or entail categorical ought statements; rather they express or entail statements such as “if it is possible to do A, then you ought to do A.” ^{Footnote 41} Presumably Cohen would say much the same about the costs of implementing a principle: if someone objects to a principle on the grounds that implementing it would be costly, then in order to identify the correct ultimate principle we should ask questions such as, “If implementing this principle could be done in a way that was costless, would justice favor doing so?”

What follows from this? It might seem to be a short step to the conclusion that the identification of the ultimate principles that are required in order to justify the principles — or as Cohen would have it, rules of regulation — that apply in particular circumstances would always require us to “clear the decks” of all facts about what is feasible and what costs would be incurred in implementing them. ^{Footnote 42} That does not follow, however. In some cases the ultimate principle that is needed to justify an appropriate rule of regulation in a given set of circumstances may have to be identified in the light of what is feasible, for there are some ultimate principles that apply precisely because of facts about what is feasible. Consider, for example, the principle that everyone should be allowed to have the amount and quality of material resources that they want. If someone objects to this principle on the grounds that it is impossible in practice to satisfy because there are not enough material resources to go around, Cohen’s method of interrogation would have us ask, “if it were possible to allow everyone to have the amount and quality of material resources that they want, would this be what was morally required?” The answer then might well be “yes”: if material resources were simply unlimited, now and in the future, then what objection could there be to this principle? So the principle that everyone should be allowed to have the material resources they want is an ultimate principle that applies on condition that resources are bountiful. But if we want to identify the ultimate principles of distributive justice that apply to our world and help us to devise rules of regulation for it, then we have to do so on the assumption that resources are scarce.

More generally, it seems to me that “clearing the decks” of facts about what is feasible is not the way to identify a number of the ultimate principles that ground the rules of regulation relevant to us and the societies in which we live. Consider another example. Suppose that in circumstances where it is possible to meet everyone’s needs, justice requires us to do so. But imagine that it is impossible to meet everyone’s needs. The principle that we ought to do so does not provide us with any guidance. Even if it should not be rejected as an ultimate principle on those grounds alone, we nevertheless need some further ultimate principle to guide us in determining and justifying the rule of regulation that applies under these circumstances where the possibilities are limited, and which explains and justifies how we are meant to balance conflicts between, for example, helping the neediest by meeting some of their needs as opposed to bringing more people up to the level at which all their needs are met. The lesson to be learnt here is that there are some ultimate principles the applicability of which depends on facts in the sense that these facts are a condition of the application of the principle. In such cases, we arrive at the ultimate principle by keeping in view these facts rather than clearing them away. This may help to deflect some of the criticism of Cohen’s work that has come from realists: even if ultimate principles of justice are not constrained by facts about what is feasible, some of these facts may be relevant because they are conditions for the application of ultimate principles. ^{Footnote 43}

Even when the full justification of the rules of regulation that apply in a given set of circumstances does require us to abstract from what is feasible and identify the principles that would apply if those circumstances did not limit our options, it is often possible to achieve an adequate justification of what rules of regulation apply without doing so. First, as Amartya Sen points out, we can make judgments about what is better without making judgments about what is the best, ^{Footnote 44} and these judgments may often be adequately justified without reference to ultimate principles. Furthermore, as John Dunn argues, even when we cannot identify the best outcome, we may be able to identify the worst possible outcome (for example, nuclear obliteration), and the need to avoid it may provide us with an adequate justification for a particular policy. ^{Footnote 45}

Second, we may be able to show that in the circumstances we confront, a particular rule of regulation follows from any of a number of principles, in which case there may be no need to probe further. For example, strict egalitarian, prioritarian, and sufficientarian distributive principles may entail the same rule of regulation to govern the redistribution of resources in some circumstances where options are limited.

Third, we can reason analogically on a case by case basis to conclusions that, even though they are not backed by a deductively valid argument containing an ultimate principle as a premise, may nevertheless be adequately justified. Consider, for example, the analogy that Judith Jarvis Thomson used in her well-known article on the permissibility of abortion in which you are asked to imagine waking to find yourself hooked up to a famous violinist who has a fatal kidney ailment, and given your blood group you are the only person who can keep him alive. ^{Footnote 46} If he is unplugged from you now, he will die, but in nine months he will have made a full recovery. Thompson argues that even if the violinist has a right to life, it is morally permissible for you to disconnect him. She concludes by analogy that even if a fetus has a right to life, it is morally permissible for a pregnant woman to abort it. The crucial thing to note here is that Thompson’s argument is not deductively valid, and nor does it have as one of its premises a general moral principle: it simply appeals to the idea that the two cases are relevantly similar. To the extent that its conclusion is justified, its justification relies on the considered judgment that we make in relation to the case of the violinist, and on the two cases being relevantly similar, rather than on an appeal to principles — and of course many of the objections to it have involved drawing out relevant differences between the cases. Insisting that the analogy provides a justification for the conclusion only if we can identify the principles that underlie it would seem to me to be mistaken. ^{Footnote 47}

The search for “ultimate principles” through abstracting from all considerations concerning what is feasible may also take us a very long way from where we are now. Sometimes we may want to say that it leads us to principles for beings unlike us, or for a society unlike ours, rather than to ultimate principles for us or the society in which we live. Galston makes much the same point when he says that reflection on political values that is motivated only by intellectual concerns, and is a purely truth-seeking activity, may be the political theory equivalent of science fiction. ^{Footnote 48} He is perhaps more skeptical than I am about the value of this kind of activity. In my view it is nevertheless an important and worthwhile approach, for it may deepen our understanding of the concept of justice by telling us what justice would be like if we were very different beings from what we are, or if we had kinds of information or technology that will always be unavailable to us. The way in which it deepens our understanding of the concept of justice is no different in the relevant respects from the way in which some of the bizarre thought experiments that philosophers of mind engage in may deepen our understanding of the concept of personal identity, for example, imagining that a person wakes up with all of Guy Fawkes’s memories, or that we split a person’s brain in two and transplant each half into a new body, or that we “download” all of a person’s memories and “upload them” into a new brain. Reflection on these cases may not provide us with any additional help when it comes to judging whether the person in front of us is the same person we met up with ten days ago, but they do nevertheless have the potential to further our understanding of the nature of personal identity.

In this respect I am also disagreeing with David Miller when he challenges what he calls the “Starship Enterprise view of political philosophy.” ^{Footnote 49} Unlike Galston, Miller in the end thinks that this view of political philosophy is incoherent. He thinks that in order to be coherent, political theory must operate under at least some feasibility constraints and make some assumptions about what human beings are like and the circumstances under which they are living. On this issue, I am at least partly on Cohen’s side. I think he is right that we can always coherently and meaningfully ask: What principle would apply if human nature was differently constituted — if we had different capacities or inclinations, for example — or what principle would apply if the world was very different — if there were unlimited resources in it, for example — and I believe that in doing so we learn something about the concept of justice, even if we don’t identify (more) ultimate principles for us, constituted as we are and subject to various hard constraints. ^{Footnote 50} Indeed by considering what would be just in a very different society from ours where we were differently constituted, we may learn how various aspects of our society or our natures are important conditions for the application of those principles that do apply to us.

IV. On the Alleged Priority of Ideal Theory

Cohen’s argument is designed to show that the identification of ultimate normative principles sometimes at least requires us to abstract from facts about what is feasible, including facts about human incapacity that may have the status of hard constraints. But there is a less radical argument available that is intended to show that ideal theorizing which abstracts from most or all “soft” constraints — that is, constraints that are malleable to some degree — must be prior to nonideal theorizing that addresses the issue of what we should do now in the face of those constraints. According to this view, we can be justified in thinking that we are making genuine progress toward a fully just society only if we are in possession of an ideal theory that tells us what justice would require in the absence of most or all soft constraints. It constitutes a response to those such as Sen who think that practical conclusions about what we should do now can be justified without any clear understanding of full justice because we can be justified in thinking that a state of affairs is more just or less unjust without knowing what a fully just state of affairs would be. (If it were successful, it would also cast doubt on my argument above, that we can adequately justify what principles apply in the circumstances that confront us even if we have not successfully identified the ultimate principles that underpin them.)

The key argument here is developed by John Simmons, building on Rawls’s conception of the relationship between ideal and nonideal theory. ^{Footnote 51} He argues that any defensible nonideal theory of justice must have a transitional rather than merely comparative character. Its transitional character means that it requires for its application some reasonably well worked-out theory of perfect or full justice. ^{Footnote 52} Even if we can say of two states of affairs C and D that D is more just than C, this does not suffice to show that we ought to move to D since a move to D might block the path to a fully just society or make that transition much more difficult. Under these circumstances, it may be the case that we ought to take one step back in order to take several steps forward. Simmons concludes that we need an ideal theory that provides us with a conception of full or perfect justice that sets an ultimate target in order to be able to determine whether or not there is a problem of path dependence of this kind. He argues that for similar reasons we need an integrated conception of a fully just society, rather than simply a conception of what it would mean for a society to be fully just along one dimension, for example, in terms of gender justice. Without an integrated conception of a fully just society, we will not be able to tell whether a reform along one dimension that leaves other dimensions unaffected, and which makes a society more just when judged in purely comparative terms, nevertheless takes us further away from perfect justice understood as an integrated ideal.

Simmons is surely correct that path dependence matters. Indeed, it matters even for those who think ideal theory is unnecessary or deeply flawed for some reason, for there may be circumstances in which we can feasibly create a state of affairs that would be more just but we should nevertheless not do so because we would then risk much greater injustice in the future. ^{Footnote 53} But there are a number of possible lines of response to Simmons’s claim that a proper appreciation of the problems created by path dependence establishes the priority of ideal theory. I will focus on three. He partially anticipates the first of these, but it nevertheless shows that there is considerable scope for nonideal theorizing without the possession of anything resembling a worked out ideal theory. It relies on pointing out that our social scientific knowledge is highly limited and we do not have the ability reliably to predict what impact efforts to make a society more just will have in the longer term. ^{Footnote 54} Indeed, our knowledge of the future tends to become more uncertain the more distant it is from the present. We are rarely in a position to know that we could get closer to full justice by taking a path that involved making society less just in the short- to medium term. Evidence one way or the other will be very hard to come by, yet it is bound to affect any justification for choosing a path that makes society less just in the short- to medium term on the grounds that it brings us closer to full justice.

Simmons acknowledges some of these points but concludes from them that when our knowledge of the future is insufficient, all we can do is muddle through and then “it may seem acceptable to cross our fingers and just accept whatever comparative gains in justice we can get or single-mindedly attack some particular, salient injustice.” ^{Footnote 55} But it seems to me that this is the situation we face in the majority of cases when we are contemplating the justice of reforms. I would draw the more radical conclusion that in the absence of strong evidence — evidence that is likely to be unavailable to us —– that a reform that produces a significant comparative gain in the short term is highly likely to make things much worse in the long run, then we are justified in supporting it or pursuing it, and indeed are often morally required to do so.

Second, suppose that full justice will not be achieved even though it could be achieved because, say, there are powerful people with a vested interest in preventing it from being achieved. ^{Footnote 56} Under these circumstances, why think that it is full justice that should serve as the target, rather than some other standard that there is some probability we could achieve? Why would it be justified, say, to move further away from what is the best we will achieve on the grounds that doing so would nevertheless leave open the bare possibility of moving closer to full justice? Or suppose that full justice will not be achieved in the foreseeable future. Why should we choose a course of action that keeps open the possibility of achieving it in the distant and unforeseeable future, even though it takes us further away from the best we can achieve in the foreseeable future?

Perhaps these reflections give us grounds for thinking that ideal theory should be conducted in the light of a different feasibility condition, namely, that principles should be compatible with what could be achieved in the best of foreseeable circumstances, rather than with what could be achieved when constrained only by human nature and physical laws. But this brings me to the third problem, namely, how are ideal theorists to justify the particular feasibility condition that they are to employ when constructing the account of full justice that is to serve as the moral target or the standard against which reforms are to be judged? Indeed different ideal theorists place different feasibility conditions on the adequacy of normative principles. For example, Rawls supposes that principles of justice are adequate, that is, are candidates to be adopted within ideal theory, only if they could be realized in a manner that is stable for the right reasons under favorable circumstances, that is, under the best of foreseeable conditions. ^{Footnote 57} This seems to be more restrictive: it would seem to rule out a principle that is compatible with hard constraints but where we have good reason to think that it could not be realized under the best of foreseeable conditions, given the nature of the soft constraints that exist here and now. ^{Footnote 58} Allen Buchanan, in contrast, argues that ideal theory should meet the condition of “moral accessibility,” which involves satisfying three conditions: first, it must be feasible to implement it, so it must be “compatible with human psychology, human capacities generally, the laws of nature, and the natural resources available to human beings”; second, there must be a “practical route from where we are now to at least a reasonable approximation of the state of affairs that satisfies its principles”; third, the state of affairs that it specifies as ideal must be attainable without unacceptable moral costs. ^{Footnote 59}

Simmons defends Rawls’s conception of ideal theory against Buchanan’s on the grounds that Buchanan’s is in tension with regarding the principles of ideal theory as ultimate standards, for Buchanan’s version entails that those principles will change as facts about moral accessibility change and it does not allow us to condemn arrangements that there is no practicable way of reforming. ^{Footnote 60} But this response cuts no ice against those who hold that ideal theory should set a moral target that is accessible from where we are now, rather than identify standards to be used in assessing courses of action, and judging the injustice involved in states of affairs, even when there may be no practicable means of reform from our current position. It is hard to see why we should think that there is a uniquely correct account of the feasibility condition that the principles of justice defended within ideal theory should meet. There may be limits on what can count as a defensible feasibility condition that are imposed by ideal theory’s action-guiding role, but these will permit different reasonable interpretations of that condition, perhaps as a result of different views concerning how accessible the target it sets needs to be, or perhaps as a result of contrasting conceptions of how ideal theory should guide action, that is, whether it should set a target at all or merely provide a standard compatible with hard constraints that is to be used in assessing different courses of action and states of affairs, irrespective of whether that standard is unattainable now or even in the foreseeable future. Because ideal theories may reasonably interpret the feasibility condition differently, they may incorporate different soft constraints and — if they see themselves as setting a target — employ different timeframes for its application.

V. Concluding Remarks

I have claimed that different kinds of political theory are valuable and no particular type should always come first. But I am not arguing for a Paul Feyerabend-style methodological anarchism according to which anything goes. ^{Footnote 61} Different kinds of political theory can be done well or badly: an ideal theory can make idealizations that create too large a gap between it and the society to which it is intended to apply, with the result that it is incoherent because it cannot fulfill its aim of playing a role in guiding political action; reflection upon political problems may be vitiated by the neglect of important empirical evidence, and as result lead to flawed practical recommendations. Even when a type of political theory is done well, it may be misapplied, or unjustifiable conclusions may be drawn from it: we cannot straightforwardly draw conclusions about what justice requires us to do in our world from a theory about what justice would require us to do in a world that was very different from our own or in which we were very different.

The value to us of a particular kind of political theory depends in part on what question we want to address, and what we are seeking to gain from addressing it. Are we merely seeking intellectual or philosophical illumination by trying to understand “justice as such” or are we trying to find solutions to practical problems — or both? Political theory that serves a practical purpose by providing guidance on what we as individuals, or as collectives, ought to do, whether here and now or in the foreseeable future, is clearly important. But so too is political theory that is unnecessary for reaching adequately justified conclusions concerning what we should do, or that would provide guidance only in some remote possible world, or that merely serves an intellectual purpose. Political theory of this kind is valuable because it can provide us with a fuller or more complete justification for our conclusions about what we should do, or a deeper understanding of ideals such as justice. Employing David Miller’s somewhat loaded terms, we might say that there is value in doing political philosophy from the perspective of the crew of the Starship Enterprise — that is, theory that brackets all considerations of feasibility, including those that arise from human nature — instead of, or in addition to, political philosophy for earthlings, constrained by what is possible for societies of them, given where they are and what they can achieve.

Nor need there be anything wrong with political theorizing that accepts the constraints imposed by deeply entrenched political beliefs, choosing in effect to be limited by considerations of political feasibility, and asking how we should act in the face of the fact that our fellow citizens have emphatically placed some reforms off the political agenda. ^{Footnote 62} Indeed, doing so may be important for addressing questions such as “What should we do here and now?” which is a legitimate question for a political theorist. For example, political theorists might ask questions about what regulatory framework should be applied to publicly funded faith schools in Britain, given the widespread commitment to these schools among members of different religious groups and the distinctive tradition of state funding for religious schools that exists there, even if we believe that it would be better or more just if these schools did not exist. Indeed I would like to see more political theory of this kind, even though it will have a more parochial character and may not travel very well across political boundaries.

Since different types of political theory address different questions, there is a division of labor between them, and they may often complement each other. But that does not settle the issue of whether these questions ought to be addressed in a particular order of priority. As part of my defense of pluralism, I have resisted the idea that the questions raised in ideal theory concerning what principles would govern a fully just society should always be addressed first. But what of the view that nonideal theory should always come first because our theorizing should start from reflection on particular injustices, with the aim of reaching practical conclusions about what should be done to overcome them?

Sometimes the view that nonideal theory should come first is based upon a conviction about the proper role of political theory in a democratic society — in particular, the view that it ought to make a contribution to public debates about what laws and policies should be enacted. For the reasons I have given, however, it is not clear why we should think that the value of political theory consists solely in a contribution of this kind. But it might be thought that the priority of nonideal theory can be grounded in the deeper idea that we are able to understand what is just only by reference to what is unjust. David Schmidtz seems to endorse this deeper idea when he suggests that “[j]ustice is less a property than an absence of properties that make for injustice.” ^{Footnote 63} If justice is simply the absence of injustice, then we have to identify injustice before we can know what justice is. But that it is not enough to show that nonideal theory must come first, if by “nonideal theory” we mean “theory that aims to reach justified conclusions about how we should respond to injustice in the light of what is feasible under the circumstances.” For even if justice is simply the absence of injustice, it does not follow that it would be misconceived to try to describe a society that had no injustices (that is, was fully just) without first addressing the issue of how those injustices should be tackled in practice in order to make it more just. In the face of particular injustices, we might think it valuable to deepen our understanding of the nature and character of these injustices and the relationships between them by addressing the issue of what a society might look like in which they had all been eradicated and no new ones had been introduced, irrespective of whether it was feasible to bring about such a society. And we might think that this approach is worthwhile partly because we believe that the kind of understanding we would then acquire is valuable in itself and partly because we believe that if we were to acquire it, we would have more confidence in the proposals we make for reform in the light of what we judge to be feasible, and be better equipped to defend these proposals. ^{Footnote 64}