Claims about what “justice requires” are pervasive in political philosophy. Here are just a few familiar examples: “Justice requires better-off individuals to help those who are worse off” (Arneson Reference Arneson1999: 225). “[J]ustice requires a social ethos that inspires uncoerced equality-supporting choice” (Cohen Reference Cohen2008: 127). “[J]ustice requires not simply looking across the world for other fellow species members who are entitled to a decent life [but also] … at the other sentient beings with whose lives our own are inextricably and complexly intertwined” (Nussbaum Reference Nussbaum2006: 406). Similar examples abound. But what exactly is being claimed when it is said that “justice requires” that we arrange social life in these ways? What does it mean to say that implementing or maintaining some particular institutional arrangement is a “requirement of justice?”
Political philosophers have begun to devote considerable attention to such meta-normative questions (i.e., questions about how to understand the nature of claims about what justice requires) in addition to and alongside substantive questions (i.e., questions about the content of what justice requires). Prominent examples include recent debates about whether (and, if so, how) claims about the requirements of justice are: fact-sensitive or fact-insensitive (Cohen Reference Cohen2008); realistic or unrealistic (Estlund Reference Estlund2020); compliance-sensitive or compliance-insensitive (Murphy Reference Murphy2000); practice-dependent or practice-independent (Sangiovanni Reference Sangiovanni2008); subject to some kind of feasibility requirement (Southwood Reference Southwood2019); and necessarily enforceable (Blake Reference Blake2001). Such theorising has helped to produce greater meta-normative self-awareness on the part of political theorists: an enhanced understanding of what we are up to when we are engaged in substantive theorising about what justice requires and a clearer sense of the criteria of success for such theorising (see Sen Reference Sen2006).
Our aim in what follows is to add to this burgeoning meta-normative inquiry by showing that there is a highly significant ambiguity in claims involving the “requirements” of justice. The ambiguity in question is a particularly striking instance of an ambiguity that John Broome (Reference Broome2013: ch. 7) has observed in the context of claims involving other requirements such as the requirements of rationality, but it appears to have been overlooked by political philosophers in connection with claims involving the requirements of justice. Such claims may pick out either one of two categorically distinct and non-coextensive kinds of requirement that we shall call 1) requirements-as-necessary-conditions for justice and 2) requirements-as-demands of justice.
Failing to notice the ambiguity is not only liable to distort our normative thinking about politics, resulting in us talking past one another. It also makes us vulnerable to certain kinds of normatively consequential errors. One is mistakenly drawing inferences about what justice demands of us from claims that certain states or societies are not just. Another is mistakenly drawing inferences about what states or societies are (or would be) just from claims to justice does not demand of states or societies that they do certain things. Paying greater attention to the distinction between the two different kinds of requirements and the ways in which they come apart is also helpful in resolving (or at least clarifying) a number of other notoriously murky meta-normative debates, especially certain debates about realism and idealism in political philosophy. Or so we shall argue.
I. Broome on two kinds of requirements
In order to explain the ambiguity and why it matters, it will be helpful to begin with John Broome's (Reference Broome2013: ch. 7) seminal discussion of requirements in general.
According to Broome, there are at least two quite different notions of a requirement that may be picked out by claims such as “F requires that p” and “p is a requirement of F.” First, such claims may involve what Broome calls “property requirements” and what we shall call “requirements-as-necessary-conditions.” Such claims are to be interpreted as holding simply that p is a necessary condition for instantiating the property of F. Here are some examples:
(1) Healthiness requires that we have sufficiently oxygenated blood.
(2) Evilness requires that we have bad motives.
Such claims are to be interpreted as follows:
(3) Having sufficiently oxygenated blood is a necessary condition for instantiating the property of being healthy.
(4) Having bad motives is a necessary condition for instantiating the property of being evil.
Second, claims to the effect that “F requires that p” may alternatively involve what Broome calls “source requirements” and what we shall call “requirements-as-demands.” Such claims are to be interpreted as holding that F encompasses certain demands of us that include the demand that p. Here are some examples:
(5) The law requires that we pay our taxes.
(6) Etiquette requires that we not slurp our soup.
Claims of this sort cannot be correctly interpreted in terms of some necessary condition for being the law or being etiquette. Rather, they are to be interpreted as
(7) The law demands of us that we pay our taxes.
(8) Etiquette demands of us that we not slurp our soup.
The nature (and logic) of requirements-as-demands is quite different from that of requirements-as-necessary-conditions. Whereas requirements-as-necessary-conditions are alethic, requirements-as-demands are deontic, involving oughts or obligations to do things. As such, they necessarily have addressees: individual or collective agents of whom (or of which) the relevant demands are being made. Thus, part of what it means to say that the law requires us to pay our taxes is that we ought or are obliged (by law) to pay our taxes. By contrast, it is clearly not any part of what it means to say that evilness requires that we have bad motives is that we ought or are obliged to have bad motives.
In the case of requirements-as-demands, if the thing that is required fails to occur, then it follows that we have done something wrong or impermissible. For example, if we don't pay our taxes, it follows that we have violated a legal demand (to pay our taxes) and are guilty of committing an act that is wrong or impermissible in the sense of being unlawful. Or, again, if we slurp our soup, then it follows that we have violated a demand of etiquette and are guilty of committing an act that is wrong or impermissible in the sense of being a breach of etiquette.
By contrast, in the case of requirements-as-necessary-conditions, if the thing that is required does not obtain (i.e., if it is not the case that p), then it merely follows that the relevant property F is not instantiated. Thus, for example, if our blood is not sufficiently oxygenated, it follows that we don't possess the property of being healthy; and if we don't have bad motives, it follows that we don't instantiate the property of being evil. But it clearly does not follow that we have done anything wrong or acted impermissibly in any way.
The instances of “F requires that p” considered so far, while involving a locution that is potentially ambiguous, do not themselves seem ambiguous because there seems to be a unique correct interpretation of them. (Specifically, “healthiness requires that we have sufficiently oxygenated blood” and “evilness requires that we have bad motives” seem uniquely correctly interpreted as involving requirements-as-necessary-conditions (and not requirements-as-demands). “The law requires that we pay our taxes” and “etiquette requires that we don't slurp our soup,” in contrast, seem uniquely correctly interpreted as involving requirements-as-demands (and not requirements-as-necessary-conditions).) But certain other claims to the effect that “F requires that p” may be correctly interpreted as involving either requirements-as-necessary-conditions or requirements-as-demands. Consider the following:
(9) Morality requires that we refrain from telling malicious lies.
(10) Rationality requires that we believe that Q if we believe that P and we believe that if P then Q.
On the one hand, such claims may be correctly interpreted as involving requirements-as-necessary-conditions, i.e.,
(11) Refraining from telling malicious lies is a necessary condition for our instantiating the property of being moral.
(12) Believing that Q if one believes that P and one believes that if P then Q is a necessary condition for our psychology instantiating the property of being rational.
That is because there is a (normative) property of being moral to which “morality” may be correctly interpreted as referring, such that we fail to instantiate it if we tell malicious lies. Likewise, there is a (normative) property of being rational to which “rationality” may be correctly interpreted as referring, such that we fail to instantiate it if we fail to believe that Q in spite of believing that P and believing that if P then Q.
On the other hand, such claims may also correctly be interpreted (and indeed, according to Broome, in the case of claims about what “morality requires” are typically more naturally interpreted) as involving requirements-as-demands, i.e.,
(13) Morality demands of us that we refrain from telling malicious lies.
(14) Rationality demands of us that we believe that Q if we believe that P and we believe that if P then Q.
That is because morality and rationality are also like law and etiquette in encompassing – and indeed being (partly) constituted by – certain demands understood as deontic claims involving oughts or obligations. We are obliged to refrain from telling malicious lies; if we tell a malicious lie, then we will have done something impermissible. Similarly, we ought to believe that Q if we believe that P and we believe that if P then Q; if we fail to believe that Q in spite of believing that P and believing that if P then Q, respectively, then again we will have gone wrong.
This naturally raises the question: In the case of requirements such as those of morality and rationality where there are both requirements-as-necessary-conditions and requirements-as-demands, what is the relationship between requirements-as-necessary-conditions and corresponding requirements-as-demands?
One view holds that they are necessarily coextensive. That is to say, anything that is a necessary condition for instantiating the relevant property must be something that is also demanded of us, and anything that is demanded of us must also be something that is a necessary condition for instantiating the relevant property. However, such a view is hard to maintain. To be sure, it is quite plausible (though not wholly uncontroversial) in the case of some requirements, such as the requirements of rationality. Anything that is a necessary condition for being fully rational is also plausibly something that rationality demands of us, and vice versa. But in the case of other requirements, such as the requirements of morality, it is highly questionable. For there are plausible candidate necessary conditions for instantiating the property of being moral that arguably are not accompanied by any corresponding demands of morality.
Here is just one example. Consider requirements involving acting on the basis of certain motives, such as the following:
(15) Morality requires that we help other people out of a concern for their well-being.
Interpreting (15) as involving a requirement-as-necessary-condition gives us
(16) Helping other people out of a concern for their well-being is a necessary condition for instantiating the property of being moral.
Suppose that we always help other people, not out of a concern for their well-being, but on the basis of some purely self-serving motive. It follows that we fail to instantiate the property of being a moral person. This strikes us as hard to deny.
Now consider a claim involving a corresponding requirement-as-demand, i.e.,
(17) Morality demands of us that we help other people out of a concern for their well-being.
Many of us will baulk at (17). Demands involve oughts or obligations. Yet, as Ross famously put it: “That action from good motive is never morally obligatory follows ... from the Kantian principle, which is generally admitted, that ‘I ought’ implies ‘I can’. It is not the case that I can by choice produce a certain motive (whether this be an ordinary desire or a sense of obligation) at a moment's notice, still less that I can at a moment's notice make it effective in stimulating me to act” (Ross Reference Ross2002: 4–5). Of course, we might reject the Kantian principle – or at least Ross's interpretation of it. But suppose we accept it. Perhaps there is some morally significant property – what Kant called “moral worth” – such that our actions only possess it if they are performed from certain motives. But morality cannot demand of us that we act from a certain motive. It follows that morality cannot demand of us that we help other people out of a concern for their well-being. (17) is false. Requirements-as-necessary-conditions and requirements-as-demands are thus not coextensive in the case of the requirements of morality.
II. Two kinds of requirements of justice
Having introduced the distinction between requirements-as-necessary-conditions and requirements-as-demands in general, we are now in a position to see how that distinction bears on claims involving justice in particular. We shall argue for two main theses. The first thesis is that claims about the “requirements” of justice are straightforwardly ambiguous between claims involving requirements-as-necessary-conditions for justice and requirements-as-demands of justice. Call this the ambiguity thesis. The second thesis is that, given a certain very plausible assumption, these two different kinds of requirements of justice may come apart in at least one direction (i.e., there are some genuine necessary conditions for justice that do not entail corresponding valid demands of justice). Call this the non-coextension thesis. Let us take each of these in turn.
A. In defence of the ambiguity thesis
Start with the ambiguity thesis. Consider the familiar Rawlsian claim that
(18) Justice requires that states arrange their social institutions in a way that advantages the least well off.
It seems clear, first, that this may be correctly interpreted as involving a requirement-as-necessary-condition, i.e.,
(19) Arranging their social institutions in a way that advantages the least well off is a necessary condition for states to be just.
(19) appears to be a perfectly legitimate interpretation of (18). If someone were to argue on the basis of (18) that the Australian state does not have the property of being a just state since it does not arrange its social institutions in a way that advantages the least well off, we would not bat an eyelid. To be sure, we might reject the argument as unsound: we might reject either the Rawlsian view of what justice requires; or we might reject the empirical claim about the Australian state. But we would hardly accuse the argument of invalidity. The argument seems to involve a perfectly valid inference. Yet the validity of the inference obviously depends on the legitimacy of interpreting (18) as (19).
It also seems clear that (18) may be correctly interpreted as involving a requirement-as-demand, i.e.,
(20) Justice demands of states that they arrange their social institutions in a way that advantages the least well off.
Suppose someone were to argue that the Australian state is guilty of violating a demand of justice, or committing an injustice, or doing something wrong or impermissible as a matter of justice in virtue of not arranging its social institutions in a way that advantages the least well off. Again, this would seem to involve a perfectly valid inference. But the inference is only valid if we interpret (18) as (20).
Claims involving what “justice requires” may correctly be interpreted in either way – as requirements-as-necessary-conditions or as requirements-as-demands. Furthermore, there is no obvious way of explaining away the apparent ambiguity (e.g., by saying that such claims are vague or context-sensitive or under-specified). Therefore, it would seem that we have good reason to conclude that “justice requires” is indeed ambiguous between those two interpretations.
Indeed, the ambiguity seems especially striking in the case of the requirements of justice. As mentioned above, some of Broome's examples, such as claims involving the requirements of morality, are typically more naturally interpreted as involving requirements-as-demands. There will, in consequence, typically be little danger of confusion or misunderstanding in the case of those claims. In the case of claims involving the requirements of justice, by contrast, neither the requirements-as-necessary-conditions interpretation nor the requirements-as-demands interpretation seems more natural than the other. On the contrary, both interpretations seem equally natural.
B. In defence of the non-coextension thesis
Even if we are right that claims about what “justice requires” are indeed ambiguous between claims involving requirements-as-necessary-conditions and claims involving requirements-as-demands, for all that we have said so far it could still be the case that the two sets of requirements are perfectly coextensive. Perhaps the requirements of justice are like the requirements of rationality: anything that is a necessary condition for justice is also something that justice demands of us, and vice versa. Perhaps justice demands of us that we do what is necessary for justice to obtain in each and every case. For example, according to one view, demands of justice simply are (or at least are to be explained in terms of) necessary conditions for justice.
Against this, we shall argue that, given a certain plausible assumption, there are going to be cases where there are true claims about what “justice requires” in the sense of genuine necessary conditions for justice that do not entail corresponding true claims about what “justice requires” in the sense of what it demands of us. We focus on cases where satisfying a particular necessary condition for justice would make things more unjust overall.
Here is such a case from David Estlund:
Suppose that a virulently racist citizenry would furiously and successfully resist fully equal civil rights, and these are genuinely, but not terribly, unequal at present. … [T]here is virtually no chance of success [in instituting such rights] (lower the chance as much as you need to in order to accept this assumption), and overwhelming likelihood (again, set this where you need to) that it would lead to civil war, still with no hope of equal civil rights at the end. Indeed, we might add the supposition that in the imagined conditions the most likely outcome of such a civil war is destruction of even the modest degree of equality in the status quo ante (Estlund Reference Estlund2020: 23).
Now consider the claim that
(21) Justice requires that Estlund's society institute fully equal civil rights.
Suppose, first, that we interpret this as involving a requirement-as-necessary-condition, i.e.,
(22) Instituting fully equal civil rights is a necessary condition in order for Estlund's society to be just.
Such a claim seems very hard to deny. As Estlund puts it, the racist society he invites us to imagine “is patently not a just society” (Estlund Reference Estlund2020: 23). Part of the reason why it is not a just society is precisely that there are no equal civil rights. Even if the society were to make certain other improvements with respect to justice, including perhaps becoming less racist, it would not be a just society unless and until it had also instituted equal civil rights.
Now suppose that we interpret (21) as involving a requirement-as-demand, i.e.,
(23) Justice demands of Estlund's society that it institute fully equal civil rights.
Such a claim is highly questionable. For it would imply that justice demands of Estlund's society that it institute fully equal rights in spite of the fact that doing so will, as a matter of fact, result in a far greater setback to justice. This is hard to maintain. It is very tempting to assume that justice cannot demand of us that we do what will as a matter of fact make things (much) more unjust overall. Given this assumption, it follows that justice cannot demand of Estlund's society that it institute fully equal rights in spite of the fact that doing so is a necessary condition in order for the society to possess the property of being a just society. Thus, we have an apparent example of a necessary condition for justice that is not accompanied by a corresponding demand of justice.
Before proceeding any further, however, we want to consider an important challenge.Footnote 1 This is that the non-coextension thesis is mistaken and that we have been seduced into thinking otherwise as a result of having mislocated the source of the ambiguity in claims about what “justice requires.” We have been assuming that the ambiguity derives from the “requires.” That is, there are two distinct things we mean by “requires:” one that consists of being a necessary condition, the other that consists of being the content of a (deontic) demand. But there is an alternative view. According to this alternative view, the ambiguity derives instead from “justice.” That is, there are two distinct things we mean by “justice:” one that consists of being a potential property of particular things (such as a state or society) and that is concrete and relative (in that it can be instantiated to a greater or lesser extent); the other that consists of being a kind of abstraction or ideal and that is capable of making demands of (concrete) particulars such as states or societies. Moreover, justice as a property and justice as an ideal are related as follows: the ideal of justice demands the instantiation of certain properties in particulars. If this alternative view is correct, it suggests that claims about what “justice requires” will always unclearly mix these interpretations together. And it suggests that we are wrong to think that true claims about what justice demands of us can come apart from true claims about necessary conditions for justice. Anything that is a necessary condition for justice to obtain will presumably be a property of the ideal of justice, and vice versa. If the demands of justice are just demands to instantiate properties of the ideal, then it follows that the demands of justice and necessary conditions for justice will be necessarily co-extensive after all.
We are sceptical of this alternative view. In particular, we are sceptical that it succeeds in offering a plausible account of what justice demands of us. One problem is that it fails to generate a plausible account of demands in general. It is natural to assume that any plausible account of the demands of justice must be generalisable to other demands. But there are certain other demands that plainly cannot be understood as demands of an ideal given that there simply is no relevant ideal (that could be construed as making demands). Consider, for example, the demands of law. As we noted above, the law makes bona fide demands of us: that we pay our taxes, refrain from driving a car without a seatbelt, and so on. But such demands cannot plausibly be interpreted in terms of some “ideal of the law” that is making demands of us. There is no such ideal. The law is not an ideal but, rather, a special and distinctive set of requirements-as-demands. (Saying exactly what makes them special and distinctive is, of course, a hard question that we shall leave to others.) It would be odd, we suggest, to hold that the demands of justice should be understood in a radically different way from the demands of law.
Suppose, however, that we set aside this worry and focus squarely on the demands of justice. Unlike the law, there is, plausibly, an ideal of justice (that we might potentially think of as making demands of us). Nonetheless, we suggest that it is deeply implausible, for reasons we have already seen, that the demands of justice are going to line up perfectly with an ideal of justice. Recall the case of Estlund's society, where instituting fully equal civil rights will result in a far greater setback to justice overall. While instituting fully equal civil rights is surely an aspect of any plausible ideal of justice, it is not plausible that justice demands of us that we do what will make things much more unjust overall. Rather, justice is presumably going to demand of us that we refrain from instituting such rights, at least for now, while presumably working to alter the underlying social conditions in light of which instituting the rights will have these pernicious consequences.
It is perhaps worth mentioning one final possibility. In evaluating the alternative view, we have been assuming that the properties whose instantiation are demanded by the ideal of justice are supposed to be themselves properties of the ideal. But perhaps ideals may demand the instantiation of properties that are not properties of the ideal, and vice versa. If so, then it might be that the alternative view can make room for demands of justice such as the demand in Estlund's society to refrain from instituting fully equal civil rights. This will generate a more plausible account of the demands of justice. The problem is that it will hardly undermine the non-coextension thesis. On the contrary, the non-coextension thesis is straightforwardly vindicated if properties of the ideal of justice just are necessary conditions for justice but some other properties altogether are what is demanded by justice.
In sum, we have good reason to think, not merely that our talk about the “requirements of justice” is often ambiguous between claims involving requirements-as-necessary-conditions and claims involving requirements-as-demands, but also that the two different kinds of requirements of justice are not coextensive. The requirements of justice therefore afford us a particularly striking example of Broome's ambiguity. They are like the requirements of rationality (and unlike the requirements of morality) in that it is equally natural to interpret them as involving either requirements-as-necessary-conditions or requirements-as-demands. But they are like the requirements of morality (and unlike the requirements of rationality) in that those two are non-coextensive. At the very least, there are some true claims involving requirements-as-necessary-conditions of justice that do not entail corresponding true claims involving requirements-as-demands of justice.
III. The payoffs of disambiguating
We have suggested that claims about what justice “requires” are often straightforwardly ambiguous between claims involving two categorically distinct, non-coextensive kinds of requirements. We now want to explain why this matters – to enumerate some of the payoffs, as it were, of disambiguating.
A. Avoiding talking past one another
The first and most obvious payoff is that it is a useful antidote to the prospect of talking past one another or pseudo-disagreements. In order for debates about what justice requires to involve genuine disagreements, theorists must obviously be engaging with the same question. Failing to notice our ambiguity carries the risk that theorists who appear to be disagreeing with one another are in fact implicitly engaging with different questions – questions about requirements of different kinds – and, hence, not really disagreeing with one another at all. Paying attention to the ambiguity helps to avoid such pseudo-disagreements or merely verbal disputes (see Chalmers Reference Chalmers2011) and to focus attention on matters where we really do disagree.
Consider the debate over open borders. Formulating this as a debate about whether “justice requires that states embrace open borders” or whether “open borders are a requirement of justice,” as is often done, makes it sound as if there is a single question regarding which proponents and opponents of open borders disagree. But as we have seen there are, in fact, at least two questions. Moreover, we suggest that proponents and opponents of open borders may not have in mind the same questions and, hence, may be talking past another. On the one hand, proponents of the open borders view will often be best understood as implicitly addressing the question of what justice requires in relation to immigration, in the sense of what are the relevant requirements-as-necessary-conditions of justice in that realm. That is, they are to be understood as claiming that states embracing open borders is a necessary condition in order for justice to obtain; and any state that fails to do so (or any society whose state fails to do so) is for that reason less than just. Given the non-coextension thesis, however, this does not mean that justice necessarily demands of actual states that they embrace open borders.
For example, at one point Joseph Carens describes his case for the open borders view as “an exploration of what justice … entails with respect to immigration” (Carens Reference Carens2013: 301, italics added). He also writes: “[i]f a just world had states, they would be states with open borders” (Carens Reference Carens2013: 287). Such language suggests that, at least in these passages, Carens may have in mind necessary conditions of justice. By contrast, many opponents of open borders such as David Miller (Reference Miller2016) are best understood as concerned with the question of what justice requires in the sense of what it demands of us. They deny that justice demands of states that they embrace open borders; and they can do that without taking a stand on the question of what is necessary in order for justice to obtain. Indeed, Miller himself does not appear to be especially interested in such questions in general, often expressing scepticism about such theorising on the grounds that it reflects a kind of pointless or even detrimental unworldliness (see Miller Reference Miller2013).
B. Avoiding drawing mistaken inferences
Another important payoff of disambiguating is that it helps to guard against drawing certain mistaken inferences that might otherwise seem to be in perfectly good order. Failing to notice the ambiguity in our talk about what justice “requires” makes us vulnerable to such mistaken inferences in our thinking and failing to spot them in others'. This is significant, in part, because it makes us vulnerable to certain normative errors: errors about what justice demands of us and/or errors about what is necessary in order for justice to obtain. But it is also significant inasmuch it makes us vulnerable to certain meta-normative errors. It risks lending a spurious appeal to certain arguments for significant claims about the nature of the requirements of justice.
First, consider drawing inferences about what justice requires from claims that certain states or societies are not just. For example, take a society where the social institutions permit and indeed contribute to extensive inequality of opportunity. Let us suppose that such a society is for that reason not just. Now suppose that we infer from this that justice requires the society to reform the social institutions in question so that they no longer permit and contribute to such inequality of opportunity. Such an inference is perfectly correct if the requirement in question is supposed to be a necessary condition for justice. That is, we can correctly infer that justice “requires” the society to carry out such institutional reform in the sense that this is a necessary condition for the society to instantiate the property of justice. But the inference is plainly mistaken if the requirement is a demand of justice. That is, we cannot correctly infer that justice demands of the society that it carry out the relevant institutional reform, that the society is going wrong with regard to justice insofar as it fails to do so. Justice may or may not demand this of the society; it does not follow from the fact that the society in question is not just that justice necessarily demands of the society that it remove the injustice. To suppose otherwise would be either a) to confuse the logic of requirements-as-demands with that of requirements-as-necessary-conditions or b) to mistakenly assume that requirements-as-necessary-conditions necessarily entail corresponding requirements-as-demands.
Second, consider drawing inferences about what states or societies are (or would be) just (say, for the sake of reductio) from claims that justice does not require states or societies to do certain things. For example, take the claim that justice does not require societies to eradicate all forms of discrimination because doing so would lie beyond the motivational capacities of many of the society's members (see Estlund Reference Estlund2020: 114). Suppose that we infer from this claim that a society that fails to eliminate all forms of discrimination “might nonetheless count as fully just” (Estlund Reference Estlund2020: 114). Such an inference is, of course, correct if we interpret the claim that “justice does not require societies to eradicate all forms of discrimination” as involving a requirement-as-necessary-condition. But we are not entitled to draw the inference if the claim is interpreted as involving a requirement-as-demand. As we have seen, requirements-as-necessary-conditions do not entail corresponding requirements-as-demands. So, it is perfectly coherent to think both (a) that justice does not demand of a society that it eradicate all forms of discrimination and yet (b) that it is a necessary condition for a society to be just that it eradicate all forms of discrimination. But if (b) is the case, then obviously a society that has failed to eradicate all inequalities cannot count as fully just.
Here is an example of such an argument from David Estlund's recent book, Utopophobia: On the Limits (If Any) of Political Philosophy. A central aim of Estlund's book is to argue against a certain kind of realism constraint on principles of justice. Estlund argues that, in order to be valid, such principles must not “require more of people than we have robust reason to expect in light of certain deep and pervasive human motives” (Estlund Reference Estlund2020: 113). Notice that the language that Estlund uses here clearly indicates that he is interpreting the realism constraint as applying to principles of justice understood as requirements-as-demands rather than requirements-as-necessary-conditions. As we have seen, the former but not the latter may require things “of people.” Estlund writes:
[A] common tenet of many liberal conceptions of justice requires some version of entrenched practices of hiring, school admission, and eligibility for many other social goods without invidious regard to applicants’ race, gender, sexuality, religion or certain other categories. Call this feature of social justice non-discrimination. At the same time, many will argue that this is a standard that humans, as we know them to be now and expect them to be in any plausible future, cannot bring themselves to conform to. Grant, then, (only) for the sake of argument, that people are unlikely to conform in anything approaching a complete way to such a principle of non-discrimination. If the standard of justice itself is bent to fit this aspect of humans’ crooked shape, then a social world with these expectable levels of invidious discrimination might nevertheless count as fully just. This is implausible on its face (Estlund Reference Estlund2020: 114).
Estlund's objection is as follows: If the aforementioned realism constraint were valid, then it would entail that a society in which there is rampant discrimination because the members of the society cannot bring themselves not to discriminate counts as just. But it is not the case that such a society counts as just. So the realism constraint is invalid.
This argument, however, is clearly guilty of making the second kind of mistaken inference we mentioned above. That is, it makes the mistake of thinking that if justice does not demand of a society that it do something, then a society that does not do that thing can count as fully just. But it plainly does not follow since doing something can be a necessary condition for a society to be just even when justice does not demand it of a society. All that follows is that such a society is not violating a valid demand of justice. So Estlund's intended reductio fails.
C. Clarifying certain other meta-normative debates
Yet another payoff of disambiguating is that it provides us with a way of clarifying and thereby helping to resolve certain rather murky meta-normative debates that loom large in contemporary political theory concerning realism versus idealism. We shall briefly mention two such debates.Footnote 2
i. Ideal and non-ideal theory
Consider, first, the ideal/non-ideal theory debate. Ideal theory involves theorising about requirements of justice on the basis of maximally favourable assumptions: e.g., that the requirements will attract perfect compliance, that societies are “well ordered,” and so on. Non-ideal theory involves theorising about the requirements of justice in a way that makes only more realistic assumptions. These two modes of theorising can produce radically different kinds of requirements. Consider irregular migration. As Carens put it:
if a[n ideally] just world requires open borders (as I contend), the problem of irregular migration disappears in a[n ideally] just world. But what to do about irregular migration is one of the most urgent practical moral issues that we face today. Moral reflection about justice [thus] requires more than simply constructing an ideal picture of justice and then seeing how close we can approximate that picture in the real [non-ideal] world (Carens Reference Carens2013: 306).
Or consider inequalities that (given human nature or at least robust motivational structures) are needed to incentivise certain valuable behaviour. Cohen writes:
If we need inequalities to ‘encourage effective performance’ then it might be folly not to have them, but it does not follow that having them is a requirement of basic justice, where a basic principle of justice is one that has application in a society where, as in Rawls's, everyone always acts justly (Cohen Reference Cohen1992: 325)
The ideal/non-ideal theory debate is concerned with explaining the nature of and relation between these two different modes of theorising about the requirements of justice.
Disambiguating requirements-as-necessary-conditions and requirements-as-demands can potentially help with both of these issues. First, it can help to provide a more illuminating account of the nature of ideal and non-ideal theory, respectively, in a way that can make sense of the key differences between them. (The rough Rawlsian gloss we gave above appears to be a mere marker, rather than constitutive, of the distinction.) The idea would be that they involve aspiring to identify two quite different kinds of requirements: requirements-as-necessary conditions and requirements-as-demands.
Ideal theory should be understood to be aspiring to identify requirements-as-necessary-conditions, which as we have seen do not entail corresponding requirements-as-demands. To be sure, it might be said that they entail corresponding ideal requirements-as-demands: demands that we do certain things insofar as we find ourselves in ideal conditions. Since conditions are typically not ideal, those are mostly counterfactual rather than actual demands, and hence they are more peripheral to ideal theory's account of the requirements of justice.
Thus, paradigmatic instances of ideal theory such as Cohen's egalitarianism and Carens’ cosmopolitanism should be interpreted as candidate claims about what is necessary in order for perfect justice to obtain. They should be interpreted as the claims that any society which is unequal or in which borders remain closed is, for that reason, not a perfectly just society. Such an interpretation can explain why the relevant maximally favourable assumptions such as the assumption of perfect compliance are entirely legitimate, indeed indispensable, when we are doing ideal theory. If we were to assume anything less favourable, such as imperfect compliance, then this would automatically foreclose the possibility of correctly identifying necessary conditions for perfect justice. At most we could hope to identify necessary condition for imperfect justice.
Non-ideal theory, by contrast, should be understood as aspiring to identify requirements-as-demands. Thus, paradigmatic non-ideal-theoretic claims to the effect that justice requires that we make special accommodation for irregular migrants or embrace some incentive-compatible form of progressive taxation should be interpreted, not as claims about what is necessary in order for perfect justice to obtain, but rather as claims about what justice demands of us. What justice demands of us, unlike what is necessary for perfect justice to obtain, is obviously going to depend on facts about our actual circumstances – including facts about how others are going to act, facts about how well-ordered our societies are, and so on.
To make unrealistically favourable assumptions about these circumstances would automatically foreclose the possibility of correctly identifying what justice demands of us in our current circumstances. At most, we could hope to identify counterfactual demands of justice: claims about what justice would demand of us if we were in less unfavourable conditions than we actually are. Such counterfactual demands can be important in all sorts of ways. They can recommend pathways toward the realization of “realistic utopias” (Rawls Reference Rawls2001: 4, 13; Wright Reference Wright2010: esp. ch. 1). Judgments about closeness of possible worlds can help us to calibrate how near or far any given non-ideal situation is from the conditions that are necessary for ideal justice. Such judgments enable us to compare the proximity of different societies to the ideal. While useful in all of those ways, such counterfactual demands clearly do not tell us what to do in the here and now.
Next, disambiguating can also help with the question of the relation between ideal and non-ideal theory. Some theorists such as Rawls (Reference Rawls1971) hold that ideal theory is explanatorily and methodologically prior to non-ideal theory. To say that ideal theory is explanatorily prior to non-ideal theory is to say that non-ideal-theoretic requirements of justice somehow depend on, and are to be explained in terms of, ideal-theoretic requirements of justice. To say that ideal theory is methodologically prior to non-ideal theory is to say that theorising about non-ideal-theoretic requirements of justice should proceed by first identifying ideal-theoretic requirements of justice and then somehow adapting or revising them in light of our actual circumstances. Others such as Sen (Reference Sen2009) and Wiens (Reference Wiens2015) deny the methodological priority thesis on various grounds: that ideal theory is not necessary in order to do non-ideal theory; that doing non-ideal theory in a way that starts with ideal theory is liable to produce serious distortions and highly implausible results; and so on.
The interpretation of ideal and non-ideal theory that we have offered does not necessarily settle this question. But it does help at least to clarify it. In the terms we have proposed, establishing the explanatory priority of ideal theory would require establishing that necessary conditions for perfect justice are explanatorily prior to demands of justice: that what justice demands of us is somehow to be explained in terms of what is necessary for perfect justice to obtain. And establishing the methodological priority of ideal theory would require establishing that we should theorise about what justice demands of us by first trying to identify what is necessary in order for perfect justice to obtain.
In both cases, this seems a tall order. Regarding the explanatory priority thesis: we cannot simply identify demands of justice with necessary conditions for ideal justice since, as we have seen, they come apart in a number of familiar ways. Alternatively, we might seek to explain demands of justice in terms of some more general demand to pursue ideal justice. But there are familiar reasons to think that at least the most natural ways of unpacking that idea won't work either. Thus, for example, it is not plausible that justice demands of us that we try to realise necessary conditions for ideal justice. (Doing so will often be pointless or counterproductive and sometimes disastrous in the actual world, as argued above.) Nor is it plausible that justice demands of us that we approximate necessary conditions for ideal justice. (This suggestion is vulnerable to the problem of the second-best (Goodin Reference Goodin1995).) Regarding the methodological priority thesis: it seems implausible in its own right. We take it that uncertainty and disagreement about what an ideally just society would look like are endemic. Yet they seem to be no impediment to fruitful theorising about the demands of justice here and now.
ii. Distribution over conjunction
Next, consider the question of whether claims about the requirements of justice are subject to a logic that allows us to distribute over conjunction, i.e., whether the following principle is valid:
(CD) For any state/society S and outcome O, if justice requires that (S achieve O1 and O2), then justice requires that S achieve O1 and justice requires that S achieve O2.
Imagine an intensely patriarchal society where domestic violence is not recognised as a legal offence and where, in the event that the state were to introduce laws recognising and forbidding domestic violence, the laws would not be enforced, resulting in even more instances of domestic violence taking place. What does justice require in such circumstances? Justice surely requires that the state introduce and enforce laws forbidding domestic violence. But is it also true that justice requires that the state introduce such laws when, as a matter of fact, the states would be virtually certain not to enforce the laws and where the effects of introducing and yet not enforcing the laws would actually increase domestic violence? Many of us feel conflicted in the face of such non-conjunctive claims. On the one hand, such claims might seem obviously true (and to follow straightforwardly from corresponding conjunctive claims to the effect that justice requires that states introduce and enforce relevant laws). On the other hand, such claims might seem false (or at least not obviously true), and insisting otherwise might seem to smack of objectionable high-mindedness.
Once again, disambiguating can help to render our reactions coherent. Suppose, first, that we interpret the claim that justice requires that the aforementioned state introduce anti-domestic violence laws as involving a requirement-as-necessary-condition. This is obviously true. It follows straightforwardly from the claim that justice requires that the aforementioned patriarchal state introduce and enforce such laws, combined with the fact that the conjunction-distribution principle (CD) is obviously valid as applied to requirements-as-necessary-conditions, i.e.,
(CDNC) For any state/society S and outcome O, if it is a necessary condition for justice to obtain that (S achieve O1 and O2), then it is a necessary condition for justice to obtain that that S achieve O1 and it is a necessary condition for justice to obtain that S achieve O2.
Second, suppose that we interpret the claim that justice requires that the patriarchal state introduce anti-domestic violence laws is interpreted as involving a requirement-as-demand. Given that in the patriarchal state in question the law would not be enforced and domestic violence would then only become worse, the claim is at least not obviously true (and might be false). That is because the conjunction-distribution principle CD is not obviously valid (and might even be invalid) when it is interpreted as applying to requirements-as-demands, i.e.,
(CDD) For any state/society S and outcome O, if justice demands of S that (S achieve O1 and O2), then justice demands of S that S achieve O1 and justice demands of S that S achieve O2.
Once again, then, keeping the distinction between requirements-as-necessary-conditions as requirements-as-demands clearly in mind potentially allows us to have our cake and eat it too.
Here is why distribution over conjunction might behave differently in these two cases. What justice requires-as-a-necessary-condition it obviously requires of each necessary condition. By contrast, it is at least plausible to suppose that what justice requires-as-a-demand is a matter of what is needed to make an overall improvement with regard to justice. Suppose that this is so. Now suppose that two conditions are individually necessary for justice to obtain but only jointly sufficient for an overall improvement with regard to justice. Then justice will require-as-necessary conditions each of those two conditions; but justice will only require-as-a-demand the conjunction of the two conditions.
IV. Conclusion
Contemporary political philosophy abounds, not only with substantive theorising about what justice requires, but also increasingly with meta-normative theorising about how to understand the nature of such theorising. We have sought to add to the latter by showing that claims about what justice “requires” are ambiguous between claims involving two categorically distinct and non-coextensive kinds of requirements. In addition, we have pointed to some of the ways in which the distinction is practically and theoretically significant. We suspect that our brief remarks here only scratch the surface: that in making and evaluating certain claims about what justice requires and indeed in engaging with certain questions about the nature of such claims, there are many other important ways in which political philosophers stand to gain from paying greater attention to the distinction between requirements-as-necessary-conditions and requirements-as-demands.
Open access
