A couple of days ago I went to a talk about the distinction between ideal and non-ideal theory in contemporary political philosophy. The distinction is typically presented as a distinction between a theory that describes a perfectly or completely just society, and a theory that describes what justice requires in imperfect societies with varying levels of injustice in order to improve it (but not necessarily to make it perfect). There are more precise ways of articulating the distinction (see, for example, here), but probably the most common way of making it precise is the one that Rawls follows in his A Theory of Justice (where the terminology originates): namely, as a distinction between a theory that describes what justice requires when we can expect everyone or almost everyone to be well disposed to act justly ("full compliance" conditions, in the jargon) and a theory that describes what justice requires when we cannot expect everyone to do what justice requires, or to the full extent that justice requires ("partial compliance" theory). To use Rousseau's terminology, this would be a distinction between theories that take people "as they should be" (and can be, if the institutions of society were correctly ordered) and theories of justice that take people "as they are" (but try to suggest institutions that could improve them).
Rawls famously (or infamously, depending on whom you ask) presented his own theory as ideal ("full compliance") theory. But he did not think that ideal theory was unrealistic, or that it incorporated false assumptions about human nature. Since the development of the moral powers is dependent on institutions, and just institutions would reinforce the development of the moral powers, there is nothing "impossible" about full compliance: the right institutions would produce the maximal amount of compliance, because people who grow up in such institutions would develop the moral powers as much as possible (dispositions are "endogenous" to institutions, to use a bit of useful economic jargon). Of course, Rawls might be wrong about the effects of well-ordered institutions on human characters and dispositions, but the theory is not ideal because he assumes something false about human beings, but only because he assumes (on the basis of what he takes to be the best available knowledge about human moral development) something that does not obtain now but might obtain in a different institutional environment, namely, full compliance with the demands of justice.
Since we do not live under the best possible institutions, we need something like a theory of just transitions, i.e., a theory that prescribes or describes just changes along the path towards a well-ordered society; hence, according to Rawls, we need not just ideal theory but also non-ideal theory (which he did not propose to develop). Crucially, however, he argued that without ideal theory it would be impossible to properly develop non-ideal theory. Rawls' model of action seems to have been, like that of most philosophers, a model of making (in Arendt's terminology): we need some kind of ultimate blueprint if we are to be sure that we are moving in the right direction. But this assumption has been attacked by a number of people (as it was at the talk) who think that the (academically dominant) preoccupation with ideal theorizing has become a kind of scholasticism. The objection is basically that ideal theory is superfluous: we do not need ideal theory to tell us which injustices are important and need to be addressed immediately (e.g., slavery, which we might all agree is unjust – why would we need Rawls' theory of justice to tell us this?), and at any rate ideal theory is insufficient to tell us what to do in other cases (how does knowing what the ideal society look like help us figure out what to do about, e.g., female genital mutilation, if anything?).
There are a few obvious responses to this criticism. On one view, for example, ideal theory helps us clarify the values that should be guiding our actions in the non-ideal world: we need to be able, for example, to know what a commitment to "equality" or "freedom" really entails before we go around making changes to various existing social arrangements in their name. In this case ideal theory does not serve as a blueprint for (long-term) action, but rather as an exploration of the structure of our values and of the kinds of trade-offs among them that we should find permissible. But it occurred to me that one useful way of understanding the relationship between ideal and non-ideal theory is as a relationship between global theories of justice and local theories of justice.
A global theory of justice is a theory about the just arrangements affecting an entire social order (e.g., a country or the entire world): it prescribes changes to the entire structure of institutions that define a social order so as to make such an order just. A local theory of justice, by contrast, prescribes changes to specific institutions or practices on the background of a (e.g., a local theory of justice might be concerned with the rules of war, or with the problem of poverty in some particular context). Typically, global theories of justice will appear to be ideal relative to local theories in that global changes might seem to require many steps, each of which would be more or less likely to happen, which implies a low likelihood of eventual realization, whereas local changes might seem to require fewer steps, so they appear to be feasible. There is often a relatively clear path to implementation for the prescriptions of local theories of justice, whereas there is no such clear path for global theories. Moreover, global theories of justice may incorporate assumptions about human nature, the nature of justice, and about the possibilities of socialization, that are more uncertain than the assumptions that local theories of justice need to make; so, for example, we are less certain about whether any set of institutions can produce dispositions resulting in "full compliance" than we are that some set of institutions might produce the desired behaviour in some particular field of endeavour. From this point of view, we might assign a much lower probability of being true to any global theory of justice than we would to a local theory of justice.
But global theories of justice might still constrain local theories of justice. For example, it would count as evidence against a local theory of justice that it leads us to a situation where certain apparently desirable global changes seem impossible without injustice. Consider a concrete example discussed in the talk. Buchanan and Keohane have argued for a rule enabling some kind of preventive war, so long as it was waged by a suitably defined "reliable moral actor" or coalition thereof. This is a "local" theory of justice, intended to address a specific problem: Buchanan and Keohane think there are circumstances where it would seem to be a good idea to have the "good guys" wage war to prevent the "bad guys" from doing bad things, and they argue that these circumstances would need to be circumscribed through a specific institutional process. But it would also seem that if the rule proposed by Buchanan and Keohane were to become entrenched, it would prevent developments in the international system towards more "ideal" states (e.g., a Kantian cosmopolitan federation), since the rule seems ripe for abuse by militarily hegemonic powers (like the USA). So from the point of view of an "ideal" theory of justice, one would want to say something like "avoid entrenching a rule that is easily exploitable by military hegemons." This would only be evidence against the local theory of justice, not proof against it; given that the global theory is uncertain, no such proof would be given. But presumably a good local theory of justice would be properly informed by a global theory of justice insofar as it tries to avoid a merely "local" maximum of justice (which could not be abandoned for a more global maximum without injustice), or situations where local improvements in justice in one area (e.g., the rules of war) may lead to injustices in some other area (e.g., institutions concerning the distribution of wealth or income). (Note also that there might not be any "universal continuity" of justice: maybe there is no path towards a globally just society that would not involve injustice). To be sure we do not know much about the global maxima of justice (the states of maximal justice), if there are any such, so that for the most part we are reduced to arguing about local improvements in justice; but to the extent that we can constrain arguments about local justice and injustice byviews about what would be globally just, that would seem to be a good thing no? Anyway, the thought seemed more interesting at the time, but here it is. (I had some other thoughts about the "ought implies can" principle as well, but I will leave these to some other post).