What's not wrong with ideal theory

Ralph Wedgwood has a new blog where, in his first post, he argues "Against Ideal Theory." For those of you who may not know much about the distinction between "ideal theory" and "nonideal theory", the distinction, very roughly, is this: ideal theory deals with what it is to be fully just, or virtuous, or rational, or whatever (depending on one's area of inquiry), whereas nonideal theory deals with what justice, or virtue, or whatever requires in some domain short of that ideal. Now, the distinction is actually much more contentious than this (see e.g. my discussion in section 1 of this paper, where I argue that the distinction has been misdrawn in the Rawlsian literature). But, let us set this aside.

Although I agree with Wedgwood that ideal theory has often seriously led philosophers astray -- for about 40 years, Rawlsian philosophers often argued about what "justice requires" in our world, this despite the fact that Rawls' arguments for his principles of (domestic and international) justice abstract away from the nonideal, and no bridging argument was ever given for extending his arguments to nonideal theory -- I think Wedgwood's critique of ideal theory is misguided, and promises to mislead philosophers in the opposite direction: the direction of thinking that they can say sensible things about the nonideal world without first defining an appropriate ideal. This is a popular new move in political philosophy, and it is one that I believe is deeply mistaken.

Wedgwood's critique begins with a "methodological problem." He writes:

An ideal F is one of the F's that are, in the relevant respect, the best possible F's. So, e.g., an ideally virtuous agent is one of the best (i.e. most virtuous) possible agents; to think in an ideally rational way is to think in the best (i.e. most rational) possible way; an ideally just society is one of the best (i.e. most just) possible societies.

So, to determine what an ideal F would be like, one has to answer two quite different questions:

• What F’s are (in the relevant sense) possible?
• What is it for one F to be (in the relevant respect) better than another?

Surely it is clear that these two questions belong to entirely different categories.

• The first question is a modal question, about what is possible. Normally, the relevant sense of ‘possibility’ will make this an empirical question.
• The second question is a normative or evaluative question, about what is better and what isworse in the relevant respect. Normally, the relevant sense of ‘better’ will make this a purelyphilosophical non-empirical question.

Clearly, it is dangerous to try to answer these two very different questions simultaneously. Unarticulated assumptions are liable to get smuggled in, without receiving adequate scrutiny.

It seems to be for this reason that it is almost never explained exactly what the relevant notion of ‘possibility’ is, and hardly ever argued that the ideal is indeed “possible” in that sense. (One honourable exception is Plato’s discussion of the sense in which the ideal state is “possible” in Republic 471c-473e.)

Instead, philosophers all too often shift back and forth: when they want to suggest that their “ideal” has some practical relevance, they strengthen the sense of ‘possibility’, in order to suggest that the ideal might be practically within reach; when they want to defend their ideal against objections, they weaken the sense of ‘possibility’, in order to suggest that the worlds where the ideal is realized are remote from our non-ideal actual world.

Wedgwood's argument here, as far as I can tell, is essentially this:

1. It is really important not to confuse two different things.
2. People who do ideal theory tend to confuse those things.
3. Therefore, ideal theory has methodological problems.

This is fallacious. I entirely agree that it is important not to confuse the things that Wedgwood mentions, and that philosophers who work in ideal theory often do confuse those things -- but none of this shows that ideal theory is methodologically flawed. It shows, at most, that many people have done it badly!

Wedgwood then writes of certain "theoretical mistakes" he sees in ideal theory:

For evaluative and normative theorizing, what is most important is to articulate a plausible conception of what it is for one item in the relevant category to be better than another.

I think this is just wrong. I don't think "the most important thing" in normative theorizing is to know "what is better than what." That is an important thing to know, but to say it is the most important thing -- without argument -- is simply an assertion. Here, instead, is what I want to say: There are many important things in normative theorizing. We should want to know what is better than what. But that is not all. We have every reason to want to know what would be best. To ignore ideal theory -- without argument for why "what is best" is not something worth knowing -- is to arbitrarily set aside an important question as irrelevant.

Second, I do not think that we can specify what is better than what without at least some ideal in the background. To say that it would be better for people of different races to have equal rights than for one race to have more than others is to say that it is more ideal. But, what is it to say that something is more ideal? It is to say that it is closer to some ideal. Thus, I say (along with Rawls), the idea what we can do "nonideal theory" without ideal theory is nonsense. Any attempt to do nonideal theory inevitably -- if only tacitly -- appeals to ideals.

Wedgwood then adds:

Ideal theorists may assume that their description of the ideal gives us an easy way of articulating such a conception of what it is for one item x to be better than another item y: x is better than y if and only if xis closer to the ideal than y.

However, simply describing an ideal does not determine a unique metric of closeness to the ideal. This point has recently come to be appreciated within formal epistemology: even if perfect probabilistic coherence is part of being ideally rational, there are many different ways of measuring how closely an incoherent set of beliefs approximates to being perfectly coherent. Further substantive discussion is required to see which is the appropriate metric. The same point surely applies to ideal theories of individual virtue or of social justice: just describing an ideally just society will not in itself give us a way of measuring how closely a less-than-perfectly just society approximates to ideal justice.

In this way, a focus on ideal theory has obscured one of the crucial tasks of evaluative and normative theories – viz. to articulate an account of degrees of justice, or degrees of virtue or of rationality (i.e. an account of what it is for x to be more just, or more virtuous or rational, than y).

But now notice: this commits the same error as before. When Wedgwood says that, "Ideal theorists may assume that...", he is saying ideal theorists can theorize badly about nonideal conditions. This is certainly correct, and ideal theorists are guilty of it. But, just because something has been done wrongly, that does not mean that it should not be done, or that it is methodologically flawed. Wedgwood is right than describing an ideal alone does not determine a unique metric of closeness, and that "a focus on ideal theory has obscured one of the crucial tasks of evaluative normative theories...to articulate an account of degrees of justice...". But, here again, it is fallacious to reason from this to the conclusion that ideal theory should not be done. A better way to go would be to do ideal theory correctly, and then correctly extend one's normative reasoning to the nonideal.

Finally, Wedgwood gives several cases of how, "[A] focus on ideal theory can also lead to mistaken evaluations of particular states of affairs. Here there are mistakes both on the right (the mistakes of conservatives), and on the left (the mistakes of revolutionaries)." I totally agree that the mistakes he then lays out have been made -- but still, the fact that people have made mistakes doing something does not establish that that thing should not be done. As I have explained above, we have reasons to do both things -- ideal and nonideal theory -- correctly. Thus, I submit, Wedgwood's proposed solution is not a solution. Wedgwood writes, "The fundamental normative and evaluative questions are not about what is ideal, but about what is better and what is worse." No, the fundamental normative and evaluative questions are two:

1. What is the ideal?
2. What do the normative foundations of that ideal require in the nonideal?

And this is just to say that the ideal theorists were right all along. 

Comments

[Nick]

So, I agree that Wedgewood's post is kind of incomplete, and that it doesn't add much to the debate. However:

Ideal Theory (1): a theory that generates a conceptual analysis for a social concept (justice, goodness, etc) by abstracting away from real-world details.

Ideal Theory (2): Theorizing about what would be best, or ideal.

I'm worried that you oscillate, here, since (1) is plainly not (2). This means that positive arguments which lead to (2) do not lead to (1), and (1) really is the Rawlsian project.

So you say: "We have every reason to want to know what would be best. To ignore ideal theory -- on poorly justified grounds -- is to arbitrarily set aside an important question as irrelevant." But this does not really lead to the Rawlsian project, which requires a particular kind of theorizing about what is best.

Also, I'm guessing you don't want to say that it is "nonsense" to say that we can do non-ideal theory without ideal theory: it can't be that every time in world history someone has reasoned about betterness rankings they have drawn on an ideal theory in either the sense of (1) or (2). Unless you mean "drawn on" in the sense of "I can interpret them as having drawn on", in which case the claim is trivial.

[Marcus Arvan]

Nick: Thanks for your comment. I think you're simply wrong about the Rawlsian project. Rawls is *not* doing conceptual analysis of "justice" (and he never says anything to suggest as much). Rather, he is entirely clear that he is providing a *substantive* argument in reflective equilibrium for what is in fact best.

So, Rawls is not vacillating between (1) and (2), and neither am I. I have little interest in conceptual analysis -- as (following Millikan) I think this is one of the big red herrrings of contemporary analytic philosophy. We do not share clear concepts with clear satisfaction-conditions. What we do share is the ability to communicate using our own "unicepts", and we can use these unicepts to make substantive arguments to one another. Although Rawls might not have thought about it this, I do -- and, in any case, Rawls is not engaging in conceptual analysis.

[Nick]

I worried that "conceptual analysis" would gum this discussion up, and I should have heeded that worry. I am 100% with you on conceptual analysis, and I didn't intend the contrast to hinge on it. So forget I said it.

The relevant contrast here is between the projects of figuring out what is best and figuring out what agents would choose in a world markedly dissimilar from our own. Those are not the same projects, and my claim is that you (largely) argue for the former, while the debate ought to be focused on the latter. (I do not think this is *your* mistake by the way, the equivocation is common in the literature I've read, though Mills is good at distinguishing senses of 'ideal').

[Marcus Arvan]

Hi Nick: Thanks! I agree that I am not making the mistake. ;) They are indeed two different projects. However, I do not think the latter project (viz. what agents would choose in a world markedly dissimilar to our own) is *necessarily* mistaken as a guide to the former.

For instance, although I think Rawls mucked things up in execution, the question of what a world -- very different than ours -- where everyone was treated as free and equal individuals (a model of "what would be fair") *can* be correct as a model of an ideal. The crucial thing is not to let the ideal model lead you astray in the *nonideal*. And I think can be done.

[Pierre]

There are reasons not to believe that Rawls focuses on “the best” when he does ideal theory. The main such reason is that he provides an analysis of the burdens of judgment *within ideal theory* (Political Liberalism, around p.55sq if I’m not wrong). So there is, under the assumptions of ideal theory, disagreement about what the “best” is, and this disagreement is rooted in stylized facts about our conduct, the role of our experience, and the difficulties of arriving at a correct judgment when confronted with conflicting evidence. What is ruled out in (this aspect of) ideal theory is, basically, *stupidity* --- the “ideal” disagreement is reasonable (no one will use fallacious arguments, be it intentionally/malicious or out of sheer imbecility).

Now I’m not sure that Sen, whom you mention, is *really* opposed to ideal theory per se. I believe that (a) if his criticisms are read as against ideal theory (rather than what he calls “transcendental” theory), then they are both mistaken as regards Rawls’s project; (b) his critique could be read as against a kind of philosophical practice that’s not, properly speaking, “ideal theory”, or, otherwise stated, that (what he labels) “transcendental theory” does *not* amount to ideal theory, at least not of the Rawlsian kind. There is more in support of this than the mere change of wording: for short, I believe Rawls’s insistence on the burdens of judgment allows us to devise a kind of “comparative ideal theory”. (A more detailed argumentation is required, of course: that’s why I’m preparing a draft on this issue. I’d be happy to discuss it with you as soon as the draft is presentable.)

[Nick]

Hi Marcus: so, I understand Ideal Theory as a two-stage process: (1) idealize in a *descriptive* sense: generate a model world that is not ours. (2) generate a conception (not an analysis) of some important concept/unicept by direct reference to that same idealized world. Stage (2) generates the 'ideal' in the normative sense.

Your second paragraph seems to indicate that you are collapsing this into a single stage, where we ask questions about a world that is stipulated to be ideal in the normative sense. But the key sentence is (sorry!) kind of ungrammatical, so can you clarify what you meant to say?

[Marcus Arvan]

Hi Pierre: Thanks for your comment!

Personally, I think Rawls made a fateful -- and deeply mistaken -- move in his later political philosophy (viz. Political Liberalism); a mistake that gives rise to all kinds of issues, including the "sliding back and forth" problem between the ideal and nonideal that Wedgwood is on about in his post.

Rawls' error, in my view, was in holding that justice is *not* a comprehensive conception of the good. I think he should have simply gone Kantian, claimed that the original position embodies the notions of universal law, respect for humanity, and a kingdom of ends -- and left it at that.

Now, of course, I think he's wrong about *all* of those things. (I'm writing a book on this now!). I don't think the original position properly embodies any of Kant's notions, and I'm developing a new moral and political theory that I believe fixes this problem (as well as problems with Kant's theory).

But, anyway, that's where I think Rawls went wrong. He should have just been a Kantian, but (A) he interprets Kant wrong, (B) Kant's theory is messed up as it is anyway, and (C) his (Rawls') later theory problematically blurs the distinction between ideal and nonideal theory (in addition to many other problems as well).

[Marcus Arvan]

Nick: Thanks for your reply. I believe that a *proper* approach to ideal theory does not separate the descriptive from the normative.

The descriptive element (viz. a world modeling free and equal individuals choosing), if done properly, *embodies* a normative ideal.

This is what Rawls was doing with the original position. He was claiming that the original position w/the strict compliance assumption (DESCRIPTIVE) models the *normative* ideal of perfect free and equal treatment of citizens.

[Pierre]

Marcus: I’ll be happy to read it :-)

Still, for the time being, I believe comprehensive liberalism is mistaken, and that Rawls’s point in “JaF: Political not Metaphysical” is overall correct. I’m slightly more inclined to Quong’s anti-perfectionist approach (although I’m uncomfortable with many things he says, especially with respect to the kind of justification owed to unreasonable citizens, and also as regards the way “autonomy” is defined). Of course I remain open to new arguments: if what you propose is to extend justice to a comprehensive-yet-not-sectarian conception of the good (e.g. one that allows me and my girlfriend freely to determine whether it’s wrong of us to cheat on each other --- which we agree would be wrong, yet admit that it is not our business, even less the state’s, to command anyone so to believe), I may come closer to this view. (The issue is awfully puzzling anyway.)

[Marcus Arvan]

Hi Pierre: Thanks! :)

The new theory of morality and justice I am developing is indeed comprehensive-yet-not-sectarian in the kind of way you suggest you like.

[r]

"But, what is it to say that something is more ideal? It is to say that it is closer to some ideal."

It seems to me at least coherent that goodness can increase without bound (and that many consequentialists should say this). But if that's so, then it's open that there is a goodness ordering on states of affairs, but yet no maximally good ideal that those states are approaching as they get better.

[Marcus Arvan]

r: yes, but even there there *is* an ideal (a conception of what is best)--namely, maximizing utility. You give me a conception of what is "better", and I will show that it involves reference to an ideal.

[Phil H]

I didn't really agree with Wedgewood's post, but I agree with yours even less! Specifically this part:

"I do not think that we can specify what is better than what without at least some ideal in the background. To say that it would be better for people of different races to have equal rights than for one race to have more than others is to say that it is more ideal. But, what is it to say that something is more ideal? It is to say that it is closer to some ideal."

There are lots of concepts that just don't work like this. I can very comprehensibly say that pencil A is shorter than pencil B, but that has nothing to do with proximity to an ideal. There is no ideal of shortness (zero length? that's not short, that's non-existent!). Beauty may well be another one: we can say that one thing is more beautiful than another without necessarily believing in or referencing any ideal of beauty.

There are a few different ways it could work. I think shortness is probably an inherently comparative concept. With beauty, we may be using a proxy (strength of affective response?), and the proxy may not have any ideal. (Think of the best orgasm you ever had, times 1000 - if I showed you a painting that gave you that response, you might think it's ideal. But then I make it a little bit better. There's a scale here on which comparisons can be made, but no endpoint, no "ideal").

With something like fairness, it seems like we can define an ideal (everybody the same), but as soon as you allow a slightly complex concept of fairness, you can see that there may be competing conceptions, e.g. fairness of opportunity vs. outcome, to use a tired old example. In that case it seems sensible to resist saying that absolute equality in any one axis represents an ideal. (I suppose you could call it a "local ideal" or something.)

Even if you can define an ideal, there are many cases in which closeness to the ideal may not be an interesting concept. Complex machinery is an example: you can get 99 out of 100 parts right, and your machine is still no better than one with only 10 parts right. You need 100% to get it working. That might be true of the virtues, too: a person who has the virtues of charisma, honesty and courage (to take a random set) might turn out to be a dictator because she lacks humility and compassion. Getting closer to the ideal of virtue by giving her more charisma is only going to make things worse.

Finally, you don't seem to consider the possibility of moderate ideals. (Not sure if this is a major objection.)

[Marcus Arvan]

Hi Phil H: Thanks for your comment.

I appreciate your point about shortness, but I don't think *normative* ideals work that way.

"Tall" and "short" are descriptive notions. Now consider judgments about tallness/shortness in the case of a basketball player. One says, of player A, "He's too short." Then, one says, "It would be better if he were taller." Then one sees another, taller player. One says, "It's better that he is taller than player A." Now consider player C. One says, "He is too tall. It would be better if he were a bit shorter."

See where this is going? Assuming the process goes on long enough, there will be convergence upon some normative ideal (viz. "That is the ideal height for a player of his skill set.")

Once one begins to systematically make *coherent* normative comparisons, one will converge toward some ideal. This is just the problem of universals in a normative context. In order to say that something or someone is *better* in a normative context, one must specify, "Better in what RESPECT?" That respect, if it is coherent, will specify some ideal.

On other hand, if the comparisons (better/worse) are *not* informed by any coherent ideal, then there is no rhyme or reason to the comparisons. To riff on Plato's famous ship analogy from The Republic, making comparative judgments without an ideal in the background is like people on a ship judging, on an ad hoc basis, which direction is "better" than which without any (even implicit!) goal. This is incoherent. If the people on the ship had *no* goal whatsoever, they would have no grounds for making "better/worse" comparisons. They would, in Wittgenstein's terminology, just be "gassing."

[Phil H]

Thanks, the point about normativity is well taken. Not so sure about the ship analogy - sounds a bit like what Popper called historicist thinking.

There's still a bit of a chicken and egg issue. The way you order the argument here: "Once one begins to systematically make *coherent* normative comparisons, one will converge toward some ideal" allows for the discovery of an ideal based on comparative judgments which come first. It's not obvious to me that there can't be issues about which we can form meaningful normative opinions without knowledge of an ideal - though I accept your argument that an ideal would ultimately be implied. For example, the length of a human life. At the moment we just think "the longer the better". But it is not obvious to me that we think that because eternal life is the ideal length. Perhaps four centuries is just right, and we'll find that out when medical science allows us to.

Another example might be inequality. One can make arguments that there is too much inequality (or too little) without thinking that absolute equality is the ideal, or without knowing what the ideal is - though again, I'd accept that there must be one.

And the requirement that we be coherent is a fairly heavy one. Particularly when you get into political philosophy, where many different factors are being juggled at once. It seems very possible to me that we have some inconsistent/incoherent objectives in the political sphere, but that wouldn't invalidate argumentation. Imagine that there really is a trade-off between material wealth and equality, as many argue. I think it's naive to suggest that there *must* be an ideal balance point between the two. There may be no ideal, only an eternal clash that human societies just have to muddle through.

[Marcus Arvan]

Phil H: Thanks for your reply. I'm glad you accept the argument that a normative ideal is ultimately implied by coherent "better than" comparisons. However, I don't think a couple of things about your response are quite right.

First, you say that in terms of life span, we think "the longer the better." I don't think many people really think this, at least if they think about it with any care. For instance, it's a longstanding cliche -- particularly in vampire films -- that one can live too long (that eternal life could be wretched). And indeed, I think your life-span example shows precisely what is wrong with judgments about "what is better" that fail to pay attention to whatever ultimate ideal(s) they tacitly presuppose. If we do not think about ultimate ideals carefully (viz. is eternal life ideal?), we may be led to make incorrect judgments in the short-term (viz. is longer always better here and now?). This is yet another reason why I think ideal theory is important. Focusing on what would be *best* (in terms of justice, life-span, etc.) CAN help illuminate what would be better *now*, at least if we extend the ideal to the (nonideal) now properly.

Second, I don't think one can make clear-headed arguments that there is too much inequality (or too little) without knowing the way things should *be* (i.e. the ideal). If you don't know the way things should be, what coherent grounds do you have for thinking they shouldn't be *this* way (viz. actual inequality). By my lights, whenever people like Sen/Wedgwood/etc. say, "Well, it would be better if there were less X [inequality, etc].", they are tacitly appealing to an ideal, whether they like it or not.

Finally, it seems to me entirely wrong to say that (1) we have some inconsistent/incoherent objectives in the political sphere, but (2) this wouldn't invalidate argumentation. No. If *anything* invalidates argumentation, incoherence and inconsistency do.

This, again, is why I think we need to do all this stuff -- ideal and nonideal theory -- carefully. In my work, for instance, I show that free and equal people in a nonideal original position would agree to a fair *process* for adjudicating otherwise incompatible values (costs, tradeoffs, freedom, equality), which just shows that there *isn't* any inconsistency or incoherence after all. Rather, freedom and equality entail a certain (fair) way of responding to those things.

In short, I think that what *appear* to be inconsistent/incoherent objectives are really just a failure of theorists to think about them in a consistent or coherent way. And that is a failure on the part of the theorist -- failures that I believe can be rectified by doing ideal and nonideal theorizing correctly, rather than incorrectly.

[Pierre]

Phil H: In my view you are confusing “ideal” and “single best”. A theory can be ideal in that it appeals to a definite set of ideal assumptions, yet remain indeterminate with respect to the “single best” ideal (what Sen dubs “transcendental”). That’s why I believe Sen’s (and others’) criticism of “transcendental” theorizing is best understood as a criticism of “single best ideal” theorizing, rather than ideal theory generally. I also believe that it is possible to do “ideal comparative” theorizing: ideal in that ideal conditions are satisfied (e.g. no one is acting unreasonably), comparative in that it allows for a plurality of comparative rankings, depending on our personal experience and other aspects of our personal life. Thus it is unnecessary to assume that an ideal theory will always be of the “single best” kind.

[Phil H]

Thanks again, Marcus. I think your reply shows me more clearly where we disagree.

I might put it in terms of difficulty and complexity. Let me draw an analogy with maths. The axioms of Euclid's geometry imply, without any shadow of a doubt, that the square on the hypotenuse is equal to the square on the other two sides. But it is possible to do Euclidean geometry properly in ignorance of that fact. Similarly, I would suggest that it is possible to have consistent, meaningful philosophical beliefs (about things like equality, for example) without knowing what the implications of those beliefs are. I absolutely agree that it is the job of philosophy to elucidate the implications and work out what our axioms/principles/arguments imply. But it's putting the cart before the horse to suggest that you *always* have to work out the ultimate conclusion before you can do any good work.

So, for something like equality, it seems to me like there's a lot of work to be done on elucidating the different types of inequality that exist and whether they are problematic. (Torn from the headlines: Do the "teen affluenza" kid and the guy not sent to jail because he's rich represent a shift from economic inequality to legal inequality - this seems like a reasonable question for a philosopher to explore, without necessarily needing to define what ideal levels of equality are.)

And on the inconsistency thing, I think I'm just much less optimistic than you about whether, psychologically, all our different drives do hang together in a coherent way.

But thanks, this has been useful, and you've corrected an error or two of mine :)

[Phil H]

Hi, Pierre. I think you're right that I am conflating "ideal" and "single best". I'm not sure I understand what your more general version of ideal means. If it's a social ideal, then certainly it will allow for some individual variation, but I think Marcus is arguing that there will definitely be one social system/characteristic which is the ideal. Representative democracy, for example, as the one single system which is the ideal for distributing political power. Now, there can be different instantiations of representative democracy, and within a representative democracy individuals will have different experiences. But I understand Marcus to be saying that you have to know that representative democracy is the best system before you can do any useful philosophical work on politics. And I don't think that's true.

[Marcus Arvan]

Hi Phil: Thanks for your reply. I don't disagree with your claim that we *can* engage in "better than" judgments without completely teasing out all of their implications -- or that it is completely impossible to do nonideal theorizing without getting everything in ideal theory squared away (though, for what it is worth, I do think that is a mistake!).

My claim was simply that Wedgwood has not made a case that ideal theory (in the sense I describe) is inherently *flawed*. I think it plainly is not. We can then have a debate over how to do ideal theory properly, and over how much of it we need to do before moving onto the ideal. My point is simply that these are debates to *have*, not debates to shuttle off into the night out of a misguided belief that ideal theory is inherently flawed!