Many of you have probably seen Simmons’ article just out in PPA on ideal and non-ideal theory. Simmons defends Rawls’ account of the ideal/non-ideal theory distinction and his paper is a must read. That said, I have been ruminating over a slightly different take on the debate over the nature of the ideal/non-ideal theory distinction and so thought I’d throw an idea out there.
Drawing on John Rawls’ A Theory of Justice many have suggested that the ideal/non-ideal theory distinction is akin to the full/partial compliance distinction. In creating his ideal theory, Rawls assumes that people will comply (almost) perfectly with the requirements of justice. He then uses his original position argument to conclude that his first principle of justice should have priority over his second. Next, Rawls weakens his ideal theory assumptions, adding the constraint that people may not abide by the requirements of justice. He concludes that we should only embrace his general conception of justice in non-ideal theory.
Unfortunately, the canonical examples of ideal and non-ideal theories cannot be fully characterized as full and partial compliance theories respectively. As Simmons and others note, even Rawls says ideal theory requires more than perfect compliance. In creating his ideal theory he assumes, for instance, that the circumstances do not prevent justice from being secured. Furthermore, others have more recently provided ideal and non-ideal theories that are not full and partial compliance theories (respectively). The main thing that distinguishes Allen Buchanan’s and Michael Blake’s non-ideal theories from their ideal theories, for instance, is that their non-ideal theories assume that there will be states and consider what we should do given that we are confined to a statist system. Similarly, the main thing that distinguishes Ronald Dworkin’s non-ideal theory from his ideal theory is that he assumes that people only have different talents and disabilities in his ideal theory. Blake’s, Buchanan’s, and Dworkin’s ideal theories do not require perfect compliance. Assuming that there is something to the ideal/non-ideal theory distinction and these authors are not just using the terms in completely different ways, the ideal/non-ideal theory distinction cannot just be the full/partial compliance distinction.
Reflecting on the many ways people seem to use the terms, one might despair at the thought of trying to unify such disparate ideal and non-ideal theories. In the draft of his book manuscript Michael Blake suggests, for instance, that the ideal/non-ideal theory distinction is not that useful because it can mean many different things. He implores others to be careful to explain just what assumptions they are making in advancing any theory. Perhaps this is part of what drives Simmons and others to argue for one or another of these ways of thinking about the ideal/non-ideal theory distinction.
But, even if one or another of these authors is right and one of these conceptions is generally better than the others, a neutral conception of the ideal/non-ideal theory divide may be desirable. For, I do not see why it is illegitimate to do many different kinds of non-ideal theory and perhaps even ideal theory. In any case, I wonder if it would be possible to cash out a neutral conception of this distinction starting from one way of thinking about moral theories.
One way of thinking about moral theories is that fully developed theories provide a value function that ranges over one or more independent parameters. The parameter(s) can take on a range of values subject to constraints provided by the scope of the theory. The function tells us what is optimal or satisfactory when it specifies an acceptable value or range of values for its parameter(s). Often this requires maximizing or minimizing the value of some parameter(s) in the theory. Utilitarian theories, for instance, specify a maximizing function on utility. Similarly, in his early writing, Rawls’ first principle of justice (part of his ideal theory) suggests maximizing each person’s basic liberties subject to the constraint that doing so be compatible with everyone else having equal liberty. Sometimes a theory only specifies thresholds over or under which the values of its parameter(s) must lie. Martha Nussbaum’s capability theory, for instance, only specifies thresholds over or under which the values of its parameter(s) must lie. Nussbaum holds, for instance, that each person must be able to function fully in each of ten basic capabilities specifying that each capability is a parameter whose value must pass the full functioning threshold. Many simple deontological theories (or constraints) are also like this.
Adopting this way of thinking about theories, is it possible that non-ideal theories just include additional constraints or variables that, relative to an ideal, brings them closer to the real world in some sense (let us call such constraints or variables r-restrictions). Any theory not explicitly r-restricted relative to an ideal is an ideal theory. So, for instance, Blake and Buchanan add the r-restriction that we are, in the short term, confined to a world with states in in creating a non-ideal theory relative to the ideal theory that lacks that restriction. A theory that assumes that we are, in the short term, confined to a world with states is closer to the real world than one that does not make this assumption.
To put the proposed ideal/non-ideal distinction a different way, non-ideal theories are theories of the second best relative to some way of specifying what the first best scenario looks like. A non-ideal theory must just be closer to reality in some sense than an ideal theory. So, for instance, Rawls adds the r-restriction that people will not comply perfectly with principles of justice in making his theory non-ideal. A theory that assumes people will not comply perfectly with principles of justice is (presently) closer to reality than one that assumes they will comply perfectly.
This characterization of the ideal/non-ideal theory distinction is different from Christopher Farrelly’s categorization of theories of justice as more or less ideal in terms of their fact-sensitivity if by “fact sensitivity” he means to indicate only empirical facts. For, on the above account, non-ideal theories might be closer to the real world in terms of normative facts.
Furthermore, on this account, a theory might only include non-empirical principles. Hence, this conception of the ideal/non-ideal theory distinction does not hang on debate about the fact-sensitivity of normative theories.
If this account of the ideal/non-ideal theory distinction is right, the key to clarifying the distinction is figuring out what being closer to the real world than another theory requires. One thing we might say is that, on balance, non-ideal theory must include more (significant) truths and fewer (significant) falsehoods than ideal theory. Though, then, we would have to give an account of how to weigh and balance the significance of different truths and falsehoods. Alternately, we might appeal to possible world semantics.
The proposed way of drawing the ideal/non-ideal theory distinction is different than, but related to, the idea of first/second best theory distinction in economics. There is a single well specified ideal in economics (i.e. the perfectly competitive market). There is no single well defined ideal in moral and political philosophy. It is just essential to non-ideal theory, in the suggested sense, that it is r-restricted relative to an ideal theory.
On this account, there is no such thing as ideal theory or non-ideal theory simpliciter. Rather, there are many ideal and non-ideal theories (and almost any theory can be non-ideal relative to a host of other theories). This allows us to account for many of the different ideal and non-ideal theories in the literature. When Rawls weakens his ideal theory assumptions, for instance, he adds the r-restriction that people may not abide by the requirements of justice. This is an r-restriction because it brings us closer to the real world. It eliminates a false (perfect compliance) idealization or abstraction that Rawls assumes in doing ideal theory. Buchanan’s non-ideal theory is r-restricted because it assumes states exist. A world where states exist is closer to the real world than one without states. Dworkin’s non-ideal theory would, presumably, be r-restricted by taking into account other important features of human beings besides their disabilities and talents.
Although it is impossible to rank all theories on a single scale, this way of drawing the ideal/non-ideal theory distinction also helps us account for the fact that theories can be more or less ideal. Some authors distinguish between extremely, moderately, and mildly ideal theories, but on this way of drawing the distinction, theories fall on a full spectrum from completely ideal to very non-ideal.
I suppose the biggest objection to this way of drawing the distinction might be that theories need not be action-guiding on this account. Though I’m working on some arguments against the idea that (ideal or non-ideal) theories must be action-guiding, one could append that condition to the account…