I. What Would Be Ideal

Let’s say realism studies the human condition as it is, while idealism studies the human condition as it could be. No one objects to studying the human condition as it is, or as it could be. What divides scholars is not whether to theorize about ideals, but how to do it competently.

Utopian idealism equates “what could be” with what is logically possible, or more narrowly, what is metaphysically possible. Still more specifically, utopian idealists focus on what we could do if we tried — tried hard. ^{Footnote 1} That is what makes utopian idealism utopian.

Realistic idealism parts ways with utopian thought at this point. Realistic idealism works in a different space; let’s call it the realm of what is politically possible. A realistic idealist says it matters not only what could be if we try, but what will be if we try. Simplifying considerably, a utopian asks what is possible; a realist asks what is predictable. ^{Footnote 2}

Utopians concede that what will happen if we try x bears on whether we should try x, in practice, but say that this has no bearing on whether we should call x ideal. This is why David Estlund (arguably our most prominent utopian) dwells on Professor Procrastinate. Procrastinate is weak-willed. He tries in his predictably unimpressive way, but does not try hard enough. Thus, what he will do predictably falls short of what he could do. He knows he should promise to finish a particular task, but also knows he would not keep his promise. We infer what Procrastinate should be, and ideally would be, from what he could be, not from what he is. Given who Procrastinate is, we ask, “is someone like that even allowed to make promises?” But we ask that question without casting any doubt on what Procrastinate ideally would do.

So far, realists agree. Estlund is talking about parametric contexts. Procrastinate being unable to will his own compliance with morality has no bearing on whether he ought to comply.

II. What Would Be Ideal in a Strategic World

Then Estlund extends his point to strategic contexts. In a Carens Market, everyone is taxed in such a way that everyone ends up with equal disposable income after taxes. Yet, despite this, we imagine everyone working hard to maximize gross income. Everyone working hard is within the realm of what (conceivably) could be, but outside the realm of what (predictably) would be. Accordingly, as all sides agree, the Carens Market is a utopian ideal but not a realistic ideal. ^{Footnote 3}

All sides accept that the behavior that the Carens Market predictably would induce is not ideal. Yet, Estlund stresses, the supposition “that we shouldn’t institute the Carens Market because people won’t comply with it, doesn’t refute the theory” that people should comply. ^{Footnote 4}

This is the point well illustrated by Professor Procrastinate, but it does not apply when the problem is strategic. My strategic problem, as a political animal trying to cope with the all-too-predictable logic of the Carens Market, is not my faux-inability to command my own will, but rather this perfectly real fact: commanding the wills of my fellow citizens is nowhere to be found in my option set. That I do not choose for everyone is the political fact of life. It is nothing like weakness of will.

Suppose I imagine that pawn to E4 is the ideal move, but the idea of so moving brings on a panic attack. Estlund’s point: whether I can bring myself to move my pawn has no bearing on whether pawn to E4 is the ideal move. True.

My point: Although my inability to move myself has nothing to do with whether pawn to E4 is the ideal move, my inability to move my partner has everything to do with whether pawn to E4 is ideal. Suppose I say pawn to E4 is my ideal move. You note that Black would checkmate me in three moves. Suppose I say, that’s relevant to whether I should move my pawn to E4 in practice, but it is not relevant to whether pawn to E4 is an ideal. Ideally, strategic contexts would not be strategic contexts. In that utopian sense, pawn to E4, and the Carens Market, are ideal. ^{Footnote 5}

To a realist, however, imagining what would be ideal in a parametric world is no substitute for being able to see what is ideal in a strategic world. ^{Footnote 6} Imagining a world so unlike ours that what one wishes were ideal actually would be ideal is no substitute for seeing what is ideal in worlds like ours. We have no warrant for setting aside features of reality that embarrass the vision we long to find believable.

III. Ideals and Contingency

The consensus seems to be that x can be ideal even if not achievable, but cannot be ideal if x is not worth wanting.

Suppose I see lasagna as the perfect dish to serve my guests tonight. Lasagna seems ideal for tonight’s dinner, but that changes when ransacking my kitchen confirms that a key ingredient is missing. Lasagna would have been perfect — ideal in the sense that I could not have done better (even leaving aside considerations of feasibility). But when my reality check reveals that lasagna is not feasible, I switch to Plan B. ^{Footnote 7}

By contrast, imagine learning instead that my guest of honor is allergic to tomatoes. This second reality check tells me not that lasagna is infeasible but that lasagna is a bad idea.

My guest being allergic to tomatoes limits what I should cook, but not by limiting what I can cook. ^{Footnote 8}

“Reality checks” bring us down to earth. One kind reveals the limits of what is feasible. Another reveals the limits of what is desirable.

Similarly, when we ask whether we are looking at an ideal picnic spot, answers won’t depend on whether ravines stand between us and an otherwise ideal spot. Ravines affect whether getting there is feasible, not whether being there is desirable. But if we are talking about a reality check of the second kind (say, learning that a picnic spot would be a terrible place to eat), that tells us that the spot is less than ideal. The point is not that we can’t get there, but that even if we could, we would not want to.

A rule of thumb regarding the two reality checks: In general, what it would take to get there is a question of feasibility. What it would be like to be there is a question of desirability. So, if someone warns us that the Carens Market would not work in a strategic world, are they warning us against thinking the Carens Market is feasible, or against thinking the Carens Market is desirable?

IV. Idealization

Idealizations simplify, setting aside details for tractability’s sake. Realistically, every theory idealizes. Every map sets aside details so as to help anticipated users focus on what brought them to the map. Every theory is an idealization. Every idealization is a risk and a trade-off. So, idealization per se cannot be a mistake, yet not every trade-off is well handled. ^{Footnote 9}

Well-handled idealizations simplify by setting aside variables that make no difference to the question at hand. Suppose we aim to determine water’s boiling point. So we say, let’s classify questions about altitude as distracting details and set them aside. That particular idealization would, of course, contingently turn out to be a mistake. As matter of fact, altitude is no mere distraction when it comes to determining water’s boiling point. Boiling point turns on atmospheric pressure. Atmospheric pressure turns on altitude. Whether altitude or anything else is a mere detail can be an empirical matter — a matter for discovery, not stipulation.

For Rawls, to assume bargainers choose for a closed society “is a considerable abstraction, justified only because it enables us to focus on certain main questions free from distracting details.” ^{Footnote 10} What would settle whether this is a mistake? If x is a mere distraction, nothing changes when we set it aside. If everything changes, x was not a mere distraction. If we must set aside factor x in order to have circumstances in which y would be ideal, that explains both when y would be ideal, and when it would not be. ^{Footnote 11}

Rawls says, “Until the ideal is identified, at least in outline — and that is all we should expect — nonideal theory lacks an objective.” ^{Footnote 12} To be clear, the pitfall with doing ideal theory first arises if we treat ideal theory as a task we can finish and put behind us before moving to practical applications. The truth: at stage one, we are merely imagining hypotheses; we are not yet testing them. At stage two, we are testing ideals, not taking for granted that the only remaining question is how to apply them. Suppose an idea turns out to be predictably incompetent as a response to a real problem, but all we ask is that it be an ideal response to a more “perfect” problem. In that case, we are insulating ourselves from the kind of feedback that tells theorists when their idea is not good enough. ^{Footnote 13}

V. The Logic of the System

One enduring feature of the human condition is that we are, after all, political animals. (1) We are decision makers. (2) We are decision makers who want and need to live together. (3) As decision makers, we respond to circumstances. (4) As social beings, we respond to the circumstance that we live among decision makers — other political animals who treat our choices as part of their circumstances and respond accordingly. If our theorizing is not about that, then we are not theorizing about politics. ^{Footnote 14}

To be a political animal is to be faced with the fact that “mutual cooperation” is a possible outcome, but not a possible choice. Political animals can pray for mutual cooperation. They can work toward it. What political animals cannot do is simply choose it.

It is fine to set aside details to reveal an underlying logic predictably operating across worlds. But if we set aside the fact that incentive structures affect behavior in a law-like, robustly predictable way, then we aren’t setting aside details to reveal an underlying logic. We’re setting aside the logic. ^{Footnote 15}

There is a literature on whether Rawls was warranted in assuming that ideal bargainers would fully comply with principles of justice. But consider how much greater a stretch it is to assume that ideal bargainers not only take their own compliance but the compliance of others as given. Rawls says, “An important feature of a conception of justice is that it should generate its own support,” ^{Footnote 16} but if we take the compliance of others as given, we are not checking to see whether a conception generates its own support. Instead, we are imagining how beautiful it would be to not need to check — to not have a political problem. ^{Footnote 17} To say “ideally we would not have compliance problems” is like saying “ideally we would not need to drive defensively.” It is a remark about a world whose problems, and therefore whose solutions—whose ideals—are not like ours.

An institutional structure is an incentive structure, so to call an institutional structure ideal is to say the incentive structure it instantiates is ideal. To choose an incentive structure is to choose a compliance problem. To set aside our chosen compliance problem, as a detail best ignored, is to set aside the nature of what we are choosing as a detail best ignored. We can conceptually distinguish a basic structure from the compliance problem that goes with it, but if we imagine they can be picked separately, we misunderstand the nature of a basic structure. There is only one thing to pick: to pick the structure is to pick the problem. To have picked a bad problem is to have picked a bad structure.

So, again, our issue is not whether to theorize about ideals, but how to do it well. Are we trying to identify an ideal metaphysical possibility, or an ideal logic?

VI. What To Think

G. A. Cohen says, “the question for political philosophy is not what to do but what to think, even when what we should think makes no practical difference.” ^{Footnote 18}

Think about what? Political philosophers think about how to form a community, hold it together, and make it worth holding together. ^{Footnote 19} They think about whether our world is just. Our thoughts about justice may not matter, as Cohen says, but the fact remains that justice itself matters, and in a particular way. Justice makes for a society where people thrive. Thus, if I am horrified to learn that my loved ones will grow up in what I call a just society (where farmers who hoard the People’s food are executed, say), then I need to rethink what I call a just society. A prospect of growing up in a just society may guarantee little, but still it ought to be good news, not bad news.

Cohen supposes we can ask whether communism is ideal in theory without asking whether communism is predictably nightmarish in practice. ^{Footnote 20} To be sure, we all understand that justice can be a general rule even though there are exceptions. ^{Footnote 21} But we also understand that if the human condition is the exception to the rule, then there is no general rule. We can imagine feeling a need to temper justice with mercy in a special case, but if humanitarianism precludes what we call justice in normal cases, then we need to rethink.

We can imagine cases in which doing justice is wrong on humanitarian grounds. But we should not need to be imaginative to come up with a case in which doing what we call justice would be right on humanitarian grounds.

VII. Complacency

To be sure, we want to avoid complacent realism. ^{Footnote 22} But the problem with complacent realism isn’t the realism; it’s the complacency. Utopians worry that realism makes concessions to the reality of the human condition that are out of place when the task is to articulate ideals; the time for concessions is during the implementation stage.

Realists, by contrast, see ideals themselves as testable. We need high standards not only when assessing implementations, but also when assessing ideals themselves. High standards involve testing one’s view that x is an ideal structure by asking whether it predictably, not merely possibly, would get an ideal response.

David Estlund supposes, “that a standard won’t be met might count against people’s behavior rather than against the standard.” ^{Footnote 23} The point is valid. Indeed, it takes two to make the kind of defect that Estlund is talking about. The bare fact that people respond badly to a standard does not entail that the standard is faulty. As Estlund correctly notes, we may predict that students will fail our exam without blaming our exam. ^{Footnote 24}

Yet, noncomplacent reflection on a predictably bad outcome begins with the role we know ourselves to be playing in making it happen. That students predictably misread double negations is not a defect in our exam, but littering our exam with double negations is. To avoid complacency, we internalize something like this imperative: don’t judge people according to whether they fit your vision. Judge your vision according to whether it fits people. If your double negations confuse students in a way that is not ideal, then your exam is not ideal and you need to fix it. ^{Footnote 25}

VIII. Ideals For A Political World, Not a Moral World

Why not judge our world by comparing it to a world without injustice? What could go wrong? Consider that a world without sentient beings is a world without injustice. Could that tell us anything about justice? My conjecture is that comparisons that illuminate will be to worlds that solve problems like ours, not worlds that lack problems like ours. How would recognizably human cooperators solve problems like ours? Would they devise contract law? Would they evolve ways of tracking reputations? ^{Footnote 26}

What about comparing our world to a world of angels — beings who, by definition, cannot need solutions to problems like ours? As a model of a world without injustice, a world of angels is superficially more inspiring than a world devoid of sentient life, but just as incapable of helping human agents sort out what to regard as an ideal that is fit for a political world.

I once playfully speculated that justice is not a thing in itself so much as the logical complement of injustice. Progress, as we observe it, moves toward an open future of expanding potential, not toward a peak (a point of convergence where all possibilities for future progress are exhausted). The peak metaphor is a metaphor for a theoretical construct, not a metaphor for anything ever observed. Pits, by comparison, are all too real. So, I proposed, justice is not a natural kind. Justice is less a peak and more a matter of not being in a pit: an absence of slavery, sexism, racism . . . ^{Footnote 27} Not being in a pit is an ideal — a realistic ideal — but it is not a peak.

Liberalism presupposes that on a range of key questions about how to live a meaningful life (including choice of religion), there is no consensus. The absence of consensus is not an imperfection. Being among separate persons who decide for themselves may not be a peak, but neither is it an injustice. Responding to that reality is not a compromise. Politics is our characteristically human survival mechanism. It is a feature, not a bug. ^{Footnote 28}

Regarding religion, we learned from experience, not from theory, that the political ideal is not to determine who has the best destination but simply to manage traffic. Religion may be our best historical example of how moral ideals and political ideals come apart. That is, even if coordinating on some particular utopian vision were the moral ideal, minimizing the need to coordinate would be the corresponding political ideal. ^{Footnote 29} The definitive liberal political ideal can be a vision of not needing to regard people with different destinations as mortal enemies. ^{Footnote 30}

Among people who lack a common destination, a vision of justice worth wanting will be a set of mutual expectations that effectively manages the traffic (the trucking and bartering, the dealing, the cooperating) that is the essence of mutually advantageous cooperative society. Rawls might say that until we identify an ideal destination, traffic management lacks an objective. But that sounds like a testable factual claim. Is it true?

The contributors to this volume have a lot to say about how we would know.