The Tyranny of the Ideal

by Gerald Gaus

  1.2 Sen’s Partial Normalization Theory

  1.2.1 A Social Choice among Multiple Spectators. Although many philosophers have been dismissive, if not downright contemptuous, of Amartya Sen’s analysis in The Idea of Justice, it is one of the most important advances in the post-Rawlsian public reason project. It seeks to reconcile the idea of a definite theory of justice with the recognition that there is no single, normalized perspective from which to reason about justice. This latter point is one of the lessons to be derived from Sen’s parable of three children and the flute.12 Three children are quarrelling about who is to get a flute. If we consider only claims based on who can best use the flute, it goes to Anne, who alone can play it; if we consider only claims of need, it goes to Bob, who is so impoverished that he has no other toys;13 if we consider only claims to desert and self-ownership, it goes to Carla, who made the flute. We can construct a choice situation in which all will agree with any of the three distributions if we filter out information about the other relevant distributional criteria—that is, if we normalize the evaluative perspective (to a common evaluative standard) such that only one criterion matters. But all three qualify as reasonable impartial principles of justice, and there is no uniquely correct way to weigh them.14 “At the heart of the particular problem of a unique impartial resolution of the perfectly just society is the possible sustainability of plural and competing reasons of justice, all of which have claims to impartiality and which nevertheless differ from—and rival—each other.”15 Unless we invoke a highly controversial normalization procedure (for example, simply excluding information relating to desert), rational and impartial free and equal persons will rank the alternatives differently, disagreeing on the optimal element.

  Even after we have done our best to identify impartial spectators (i.e., eligible partially normalized perspectives) who base their judgments on considerations acceptable to all, and who are free from bias and parochialism, Sen holds that we are left with multiple impartial spectators.16 We have multiple partially normalized perspectives. Even if we suppose that each impartial spectator could give a complete justice-based ranking of feasible alternatives,17 their rankings will only partially overlap. “There will, of course, be considerable divergence between different impartial views … [and] this would yield an incomplete social ranking, based on congruently ranked pairs, and this incomplete ranking could be seen as being shared by all.”18 Figure 4-1 provides a drastically simplified, stylized example of three impartial spectators ranking five alternatives. Here the ordering shared by our three impartial spectators (or, in the terms of our model, partially normalized perspectives) is a≻b,19 b≻e (so, by transitivity, a≻e). Notice that this is to apply a Paretian rule over impartial spectators: if for all spectators a≻ b, then according to the full (comprehensive) theory of justice, a is more just than b. Notice that while each impartial spectator can rank c and d, we might say that “reasonable conceptions of justice” disagree (for Spectators Alf and Betty, c≻d, but for Charlie, d≻c). On Sen’s view the (full) theory of justice cannot order c and d (or the pairs [c, e], [d, e]) as there is reasonable disagreement about their relative justice.

  Figure 4-1. Orderings of partially normalized impartial spectators

  Much of Sen’s most innovative work has been on rational choice from incomplete orderings.20 He thus points a way out of Rawls’s conundrum. Even though we have multiple, inconsistent, partially normalized perspectives, if we allow our theory of justice to be incomplete, we can acknowledge that different perspectives will have sometimes conflicting unequivocal pairwise judgments of justice, and yet we still can generate rational choice based on consensus about justice in a wide range of situations. We can achieve definiteness (though not completeness of definite judgments) with only partial normalization.

  1.2.2 Severely Constrained Evaluative Diversity. If, employing Sen’s idea of impartial spectators, we end up with a small group of eligible normalized perspectives, it seems plausible that the Pareto rule will generate an interestingly large class of pairwise comparisons, so that our (full) theory of justice, while incomplete, will not be empty. But as we increase the set of possible normalized perspectives so that the number of impartial spectators becomes large, the Pareto rule is apt to be of little help. As soon as a single spectator deems b≻a, the full theory cannot conclude that according to (overall, or complete) justice a≻ b. Is this a problem? The reply by a proponent of Sen’s view appears straightforward: plausible normalized perspectives about justice can disagree about a lot, but so too must they agree about a lot. Any normalization, a defender of Sen might say, partial though it might be, necessarily restricts the domain of possible orderings of the worlds to be evaluated. Worlds characterized by severe human rights violations and famine are not admissibly preferable to those that score very high on protection of rights and have sufficient food for all, on any plausible perspective on justice.21

  While all this is true as far as it goes, it leads us right back to the core of Rawls’s worry concerning the diversity of reasonable views about the right or just: how deep and wide are these differences? Rawls insisted that the partially normalized views that resulted were all versions of egalitarian liberalism,22 but there is no sustained argument for this conclusion; if there are many political values that can be weighed in many ways, we have prima facie grounds for thinking that the set of eligible perspectives on justice may be rather larger than either Rawls or Sen intimated. And if we are confronted with a large set of diverse eligible perspectives on justice, then Sen’s Paretian rule is apt, after all, to be radically incomplete. It thus seems that Sen’s approach works well for the sort of limited diversity of eligible perspectives that Rawls sometimes seemed to have in mind but falters if we cannot make considerable progress in identifying a rather small set of appropriately partially normalized perspectives.

  1.2.3 Complete Normalization of Social Worlds. But that is not the truly deep worry about Sen’s approach. Sen makes much of the fact that his account is based on a “social choice perspective,” which directly evaluates societies, states of affairs, or, we might say, social worlds, rather than focusing on rules or institutions. Drawing on Indian thought, Sen emphasizes the contrast between niti—which seems a severe, rule-based approach to justice related to deontology, and nyaya, according to which “justice is not just a matter of judging institutions and rules, but of judging societies themselves.”23 Nyaya is a broadly consequentialist idea—the focus of evaluation should be the state of affairs that constitutes a society, not simply its rules and institutions. “Justice is ultimately connected with the way people’s lives go, and not merely with the nature of the institutions surrounding them.”24 Thus, Sen insists, we must focus on evaluating the justice of entire states of affairs. “Even though the possibility of describing any state of affairs ‘in its entirety’ is not credible (we can always add some more detail) the basic idea of a state of affairs can be informationally rich, and take note of all the features that we see as important.”25

  This raises a subtle but fundamental problem: Sen’s account supposes a full normalization in how the eligible perspectives characterize the elements in the domain {X} of alternative social worlds.26 Although Sen would appear to allow differences in evaluative standards (ES) such as desert, need, and so on, and the mapping relation (MP), in the form of different judgments by impartial spectators, the domain of social worlds in {X} and the relevant properties of these social worlds is fixed and fully normalized among the partially normalized perspectives (§II.1). If their exercise is to make any sense, the impartial spectators must order the same domain of social worlds, where the spectators agree on the options to be ordered (what social choice theorists often call “the feasible set”). Consider again figure 4-1. Our impartial spectators must agree on the identity conditions of a and b if we are to make sense of the conclusion that they agree that a≻ b. Now as we saw earlier (§III.2.4) states of affairs are not like artifacts or cities where we can identify the
thing (perhaps via a name) but disagree about almost all its properties. The state of affairs simply is its relevant properties. So in this case to agree on the identity of any given option, a, is to agree on its relevant properties. If the spectators do not agree on its properties, they do not agree on what they are ordering, and so they cannot agree in their orderings. That all agree that a≻ b would not tell us much if they did not agree on what a and b are. If Alf’s “a” is really a′ and his “b” is really b′,27 then he is not really agreeing with Betty’s ranking that a≻ b.

  It might seem that this problem could be overcome by partitioning these differently described social worlds (a and a′) into smaller social worlds about which they do agree. Suppose the impartial spectators come to realize that Spectator Alf has in mind by a′ the set of features {f, g}, while Spectator Betty identifies {f, j} as the relevant features of her a. Alf and Betty do agree on feature {f}, so we might think that they could agree to focus on the “small common world” defined by the intersection of a and a′—characterized by {f}—and rank that world in relation to others. But our spectators will not think this small world defined by {f} is very important, as it is divorced from other relevant features, which, when considered (recall the interdependencies of features [§II.2] may make all the difference when evaluating the justice of the “larger” social world). Given that the evaluation of a social world arises out of the evaluative standards (ES) as applied to the relevant features (WF), Alf’s and Betty’s judgments about a social world will change when they consider it as, say, possessing only relevant feature {f} and as possessing features {f, g}, so an evaluation based only on the shared, common, feature {f} will not be of much interest to either.

  Another alternative—let us call this the common projection criterion—is that Alf and Betty use {f}, as it were, merely to identify a common social world, but Alf continues to see the world as composed of {f, g} and Betty {f, j}. We might say that Alf’s perspective cannot “see” feature {j} while Betty’s cannot see {g} as properties relevant to justice. But they do concur that “world a” is that world which they both see as possessing {f}.28 On this alternative each evaluates the full (large) social world as he or she sees it, but they individuate social worlds by the shared feature they both see, in this case {f}. This clearly avoids the problem of the “small-worlds” approach, as each evaluates the full set of relevant properties as he or she understands them. But recall that in our model social worlds are individuated by their full set of relevant features (§II.1.2). This was not arbitrary, for many worlds with vastly different justice will share many common features. Consider another world (let us assume that it makes sense to call it b), that Alf’s perspective identifies as {f, x} and Betty’s as {f, z}. If we use our common feature as the common identification criterion, both Alf’s a world, {f, g}, and his b world, {f, x}, will be publicly identified as the same social world—that world characterized by feature {f}. And of course the same will be true for Betty’s quite different social worlds, {f, j} and {f, z}. Worlds a and b are the same world as identified by common features, but Alf and Betty each see these as quite distinct, with different justice scores. Alf might well be confused by Betty, who sometimes when evaluating the {f} world gives it one score (that is, when it is a world she sees as {f, j}) and sometimes another score (when she sees it as {f, z}). As Alf sees it, she gives differing justice scores to the same common world. And the same will be true of Betty in relation to Alf’s evaluations. Each will see the other as inconsistent when each is, in fact, entirely consistent in their full world evaluations.29

  The upshot is that the social choice approach to diversity requires normalization of the features of states of affairs. This is no mere formal point. Recall that, referring to the “intractable” struggles in our history of religious conflict, Rawls proclaims that “political liberalism starts by taking to heart the absolute depth of that irreconcilable latent conflict.”30 A fundamental reason why these struggles are so deep and wide is not simply (perhaps in some cases not at all) because various religious perspectives and secular worldviews have radically different understandings of the principles of justice (ES), but because they have fundamentally different understandings of the social world that principles of justice are intended to regulate (WF).31 An Evangelical Christian has a very different understanding of the social world—its real, underlying features—than does, say, a secular Darwinist philosopher. The social world of the Evangelical perspective is one where sin and sanctity are features of states of affairs; they are not merely values or preferences, but basic aspects of the ontology of the world that determine the circumstances of social life. These features simply do not exist in the social world of the secular Darwinist—as he sees it, they are illusions or fantasies. He has no categories that correspond to them; in his understanding of social worlds (WF) that are within the domain to be taken seriously, none has such properties. If he is being generous and “liberal-minded,” he may admit that Evangelicals have “preferences” to continue with their religious practices, and since preferences are indeed real, they at least count, but the underlying social world simply does not have the fantastic entities the Evangelical assumes. There is only one world, and it is the secular world revealed by science. This is the normalized world that all theories of justice, says the secular Darwinist, must presuppose.

  This seems to be Sen’s view. It is revealing that nowhere in The Idea of Justice do we confront a serious discussion of religious perspectives, and the nature of a religious person’s social world—a critical issue in many, perhaps most, societies today. Sen defends freedom of religion as a sort of liberty to participate in one’s ancestral culture,32 but it is clearly supposed that any eligible impartial spectator will have a normalized secular, naturalistic, social ontology. The impartial spectators have different evaluations, but they categorize social worlds in the same way. However, we do not even begin to understand the problem of reasonable pluralism about justice if we translate the Evangelical’s claim of a duty to obey the word of God as a claim to a right to engage in a cultural practice. We see no hint in Sen’s writings of the worry that consumed Rawls’s later work—whether liberal justice can be endorsed from a variety of comprehensive views that interpret the world and the universe in fundamentally different ways. And that is not because of any oversight by Sen. Unless he strictly normalizes what constitutes the social worlds to be evaluated, his Paretian social choice–based solution to the problem of diversity of views about justice cannot possibly work.

  It is important to stress that disagreements about the nature of the social worlds in which we live are neither peripheral nor can they be redescribed as value or preferential disputes (i.e., pushed into the evaluative standards element of a perspective). Some of our deepest and most intractable disputes are not about values or principles of justice, but about the world to which these principles apply. The most obvious instance is the long-standing and persistent struggle concerning abortion rights. Advocates of such rights see the case as decisively about fundamental rights of personal autonomy; opponents of abortion rights are depicted as having little sensitivity to a woman’s claim to control her own body.33 But this by no means follows, and often is simply not the case; opponents of abortion can be deeply devoted to such autonomy, but not in cases where it entails overriding another’s right to life. And, of course, in the abstract, most advocates of abortion rights would also draw back in such situations. The dispute is centrally about the social world to which the principles of autonomy and the right to life apply: the two social worlds do not have the same set of persons, and so even perfect agreement about abstract principles of justice would not resolve the dispute. It is only because so many moral philosophers agree with Sen that there is only a single, fully normalized, secular social world that the dispute has to be misdescribed as one simply about values or abstract principles of justice.

  A similar problem has plagued discussions of the harm principle. Recall a classic objection by Ro
bert Paul Wolff, who notes that according to Mill’s harm principle,

  I am liable to others when I affect their “interests.” Society may interfere only in those areas of my life in which it has, or takes, an interest. Now this distinction between those aspects of my life which affect the interest of others, and those aspects in which they do not take an interest, is extremely tenuous, not to say unreal, and Mill does nothing to strengthen it. Mill takes it as beyond dispute that when Smith hits Jones, or steals his purse, or accuses him in court, or sells him a horse, he is in some way affecting Jones’ interests. But Mill also seems to think it is obvious that when Smith practices the Roman faith, or reads philosophy, or eats meat, or engages in homosexual practices, he is not affecting Jones’ interests. Now suppose that Jones is a devout Calvinist or a principled vegetarian. The very presence in his community of a Catholic or a meat-eater may cause him fully as much pain as a blow in the face or the theft of his purse. Indeed, to a truly devout Christian a physical blow counts for much less than the blasphemy of heretic. After all, a physical blow affects my interests by causing me pain or stopping me from doing something I want to do. If the existence of ungodly persons in my community tortures my soul and destroys my sleep, who is to say that my interests are not affected?34