Amartya Sen’s nonideal theory

ABSTRACT Amartya Sen argues that Rawls’s theory is not only unnecessary in the pursuit of justice, but it may even be an impediment to justice in so far as it has discouraged more useful work. Against what he considers the dominance of transcendental theory, Sen calls for a more realistic and practical ‘comparative’ theory of justice. Sen’s negative point has been widely discussed, but here I develop a reconstruction of Sen’s positive theory (a combination of Adam Smith’s Impartial Spectator, Social Choice Theory, and the Capabilities Approach) in order to evaluate it on its own terms. I find that the theory is technocratic, despite Sen’s insistence to the contrary.

rather, it is a subset of a larger theory of justice that both proceeds from different assumptions and has different goals than the ideal subset of the theory (Simmons 2010). In referring to Sen's nonideal theory, I am continuing a trend which uses the term 'nonideal theory' in a broad sense to refer to theories or subsets of theories which lack idealized assumptions and/or take the elimination of injustice as their motivation. I do this in spite of the fact that Sen denies that he is engaged in nonideal theory, since he uses 'non-ideal theory' to refer only to those theories which lack Rawls's famous 'strictcompliance' assumption, and he considers his own challenge to Rawls to be bigger (90). In fact, the absence of the strict compliance assumption alone can't characterize nonideal theory. According to Rawls, the parties to the original position consider alternative principles of justice under the assumption that society will comply strictly with any principle chosen. Many have attacked this, arguing that such an assumption taints the reasoning process that leads to the principles, and thus taints the principles themselves. But the process that Rawls describes cannot abandon the strict compliance assumption, since assuming that society will comply with the principles is the way that the parties make comparative judgments. If everyone complied with egoistic principles of justice, cooperation would be impossible. This is what makes egoistic principles inferior to the principles of fairness that the parties ultimately choose. So the strict compliance assumption creates the conditions that make the reasoning inside the Rawlsian original position possible. In other words, we cannot abandon this assumption without abandoning the entire hypothetical social contract. So we cannot make sense of Sen's claim that he is doing more than others who merely question strict compliance since to question the strict compliance assumption is to question the hypothetical social contract itself.
Though many have defended Rawlsian ideal theory on the basis that the assumptions Rawls makes are a necessary part of correcting injustice there has not yet been any thorough assessment of how the alternatives that Sen offers actually function to make the world better. This assessment must take place if we are to assess broader claims about the need for a practical or realistic approach to political philosophy.
The paper is broken up into five sections: first I will give a broad overview of Sen's negative argument against Rawls and point out some of its shortcomings, though these shortcomings do not affect the positive proposals. The next three sections will break down Sen's alternatives to Rawls into three major parts, Social Choice theory, the Capabilities Approach, and the Impartial Spectator (borrowed from Adam Smith). I will discuss the way in which these three parts make up a proposal for the nonideal theory that amounts to a formal methodology for finding a compromise between actual preferences. The last section will argue that this proposal is practical only in the sense that this is a proposal that a technocrat might use to impose reform on a particular society. Sen's proposed alternatives to Rawls are technocratic because they are procedures that purport to rise above substantive disputes about justice between citizens.
Dworkin, Rawls, and Nozick, the second, 'realization-focused comparison' he identifies with Smith,Condorcet,Bentham,Wollstonecraft,Marx,Mill,and Kenneth Arrow (4). These are themselves hybrid theory forms combining transcendentalism with institutionalism or arrangement-focused and realization-focused with comparison or social choice type theory (6). According to Sen, transcendental theories concentrate on describing a perfectly just society, whereas comparative theories concentrate on comparing societies that have existed or could feasibly emerge in order to find a feasible option that is 'more just' rather than perfectly just. Institutional theories focus only on finding just institutions whereas realization-focused theories focus on the actual behaviour of people as well as institutions. Social contract theories, according to Sen, focus on a transcendental identification of ideal institutions. Some, like Rawls and Kant, also transcendentally investigate requirements of behaviour. In this respect, they are 'arrangement focused' as opposed to 'realization-focused' because they assume compliance with their ideals. So, according to Sen, Rawls's theory of justice is transcendental, institutional, and arrangement-focused, all of which Sen finds problematic.
Against transcendental theories, he argues that they are both impractical and redundant. They are impractical in the sense that there can be no impartial agreement on a single transcendental theory. Throughout his discussion, it is somewhat unclear whether Sen holds that it is impossible for there to be an objectively best transcendental theory of justice or whether he holds that it is merely impossible for there to be some kind of collective agreement as to the best transcendental theory of justice. 2 It seems most likely that he holds both, though his arguments seem to focus more on showing the latter. He discusses a hypothetical example of three children arguing over a flute.
Each child has what Sen judges to be impartial and non-arbitrary reasons why he or she should get the flute. One knows how to play it so can make better use of the flute, another is poor and has no other toys so needs the flute more, and the third made the flute herself. Sen sums up the point of the example: "The general point here is that it is not easy to brush aside as foundationless any of the claims based, respectively, on the pursuit of human fulfillment, or removal of poverty, or entitlement to enjoy the products of one's own labour. The different resolutions all have serious arguments in support of them, and we may not be able to identify, without some arbitrariness, any of the alternative arguments as being the one that must invariably prevail. There are some problems with this argument. One problem is that the fact that each child has non-arbitrary reasons for his or her preferred outcome doesn't mean the children couldn't come to an agreement. Nor does it mean that an outside observer would not be able to decide the argument without any arbitrariness. Just because there are non-arbitrary reasons on each side doesn't mean that there can be no further nonarbitrary reason(s) that would be decisive.
Sen argues that transcendental theories are not necessary to guide our action. He uses two analogies, one that we do not need to know what the tallest mountain is in order to judge whether one mountain is taller than another and two that we do not need to know what the best painting in the world is to judge whether one painting is better than another. Again, the argument goes by quickly, and the analogies are questionable. In the case of the mountains, John Simmons argues that knowing the tallest mountain is, in fact, necessary, since we don't know whether finding taller mountains is leading us eventually to the more just situation. It is possible that the path to justice is not straight up, and things need to get worse before they can get better (Simmons 2010).
David Schmidtz, in his review of Sen's book, argues that Sen should have said that there is no tallest mountain; rather, the metaphor for justice should be a featureless plain with pockets of injustice (2011). But this new metaphor isn't much better at accounting for the possibility that things must get worse before they get better. Both mountain and plain metaphors suppose that situations can be ranked along a scale of justice (either we rank the justice of a situation according to the height of the mountain or we rank the injustice of a situation according to the depth of the valley). They both presume that ranking situations according to how just they are is all that is necessary to improve our situation with respect to justice; a fact that is questionable if we allow for the possibility that improvements in the long term requires some short-term sacrifice.
As for the paintings, Sen's point is arguably false. Of course, we wouldn't want to say that a person needs to have an opinion on the best painting in existence in order to have an opinion on which of two given paintings is better than another. But, it seems likely that identifying the painting that is the best can help to identify which of the given paintings is better. Unlike the mountains, the paintings cannot be easily ranked against each other. In this case we can imagine a dispute about two paintings in which person A argues that painting 1 is superior because it is more colourful than painting 2, and person B counters the argument by pointing to the best painting (or a better painting) and showing either that it is not colourful at all, or it isn't better simply because of its colour. In this case, it is necessary to have principles for what makes one painting better than another (even if these are merely implicit) in order to advance the argument. Identifying the best painting helps to identify the implicit reasons that make it the best (Swift 2008;Gilabert 2012).
It is worth noting here that I don't think Sen's criticism of transcendental institutional assessment focused theories apply to Rawls. I believe that Rawls's ideal theory is neither institutional nor assessment focused, but more importantly Rawls's theory is not transcendental, if we take transcendental to mean not comparative. Sen's characterization of transcendental theories and comparative theories doesn't make much room for what Rawls was up to in his theory of justice as fairness. The problem is that Sen assumes that comparative assessment is the goal of transcendental theories. So he sees those that favour a transcendental approach as being mistaken in their belief in the sufficiency and/or the necessity of a transcendental approach for a comparative assessment. But Rawls clearly doesn't believe that we need his theory of justice as fairness to make comparative assessments. Rawls uses the process of reflective equilibrium in developing his theory; this process presupposes that we have some considered ideas about justice that the theory needs to reflect in order to be a good theory. The impetus behind Rawls's theory of justice as fairness is at least partly to provide our pretheoretical assessments of social realizations (comparative or not) with secure foundationsto show that they are not merely arbitrary preferences, but in fact, required by justice. Even if one reads Rawls differently, it can't be denied that at least some transcendental theories of justice are attempts to justify our comparative judgments, rather than attempts to make new comparative judgments. Another, related purpose of a transcendental theory would be to attempt to resolve inconsistencies in the comparative judgments that we begin with. None of these types of theories are captured in Sen's taxonomy, which reflects his aforementioned implicit scepticism of objective theories of justice. 4 But even if Sen's criticisms of transcendental theories fail, he may still be offering a valuable alternative to Rawls's theory. Sen advocates for comparative, realizationfocused theories as opposed to transcendental institutional assessment-focused theories. As an example of comparative, realization-focused theory Sen offers three positive theories, social choice theory as pioneered by Kenneth Arrow and further developed by Sen himself, his own capabilities theory, 5 and Adam Smith's device of the Impartial Spectator. Sen doesn't say much about how these three theories fit together, so in what follows I'll attempt to reconstruct Sen's positive proposal for nonideal theory.

Social choice theory
Since social choice theory is generally advanced using mathematical formulations and requires a great many idealizations, it is counter-intuitive to understand it as a type of nonideal theory. But here is Sen's gloss on it "As an evaluative discipline, social choice theory is deeply concerned with the rational basis of social judgments and public decisions in choosing between social alternatives. The outcomes of the social choice procedure take the form of ranking different states of affair from a 'social point of view', in the light of the assessments of the people involved" (95).
Take, for example, Sen's liberal paradox (1970aSen's liberal paradox ( , 1970b. Sen defines a collective choice rule as a functional relationship that specifies a social preference relation R for any set of n individual orderings. R must generate a choice function such that for every set of alternatives there is at least one alternative that is at least as good as all the other alternatives; this is a 'social decision function'. (In other words, the relation R must be acyclitic but need not be transitive or even quasi-transitive, Sen 1970b) Sen shows that there is no social decision function compatible with three conditions: 4 To pick just one example of theory as a way to justify prior judgments about justice, we can look to John Roemer, who Sen is certainly familiar with and even cites approvingly. (Roemer 1982) compares two different theories of exploitation and argues for his favoured theory on the basis that it best captures true cases of exploitation. (Roemer 2002) argues that egalitarians should not use the veil of ignorance theories as these theories don't yield egalitarian requirements. 5 Martha Nussbaum has further developed Sen's capabilities theory, but largely in ways that Sen disavows, so I will not be discussing Nussbaum's work on capabilities.
• Condition U (unrestricted domain): every logically possible set of individual orderings is in the domain of the collective choice rule. • Condition P (Pareto principle-in a weak form) if every individual prefers some alternative x to another alternative y; then, society must prefer x to y. • Condition L (liberalism): for each individual, there is at least one pair of alternatives such that if this individual prefers x to y then society should prefer x to y. In fact, the impossibility holds even for what Sen calls condition L* (weak liberalism) which is the condition that there are at least two individuals that are decisive with respect to at least one social preference. If proven for condition L* it holds for condition L too, so I concentrate on just L*.
The proof: Consider the pairs of alternatives (x, y) and (z, w). If they are the same pair, condition L* does not hold. Thus, they must have at most one alternative in common. Assume they have an alternative in common, say, x = z, and assume person 1 prefers x to y and person 2 prefers w to z, and everyone else, including them, prefers y to w. By condition L*, x must be preferred to y and w must be preferred to x aka z. By condition P, y must be preferred to w. But this means that every alternative is worse than another alternative. Thus, there is no way to rank the alternatives, and thus no social decision function. Now assume they are four distinct elements, and say person 1 prefers x to y and person 2 prefers z to w and everyone, including them, prefers w to x and y to z. By condition L*, x must be preferred to y and z to w. By condition P, w must be preferred to x and y must be preferred to z. But again there is no best alternative, each is worse than another.
Thus, there is no social decision function that can satisfy all three conditions. Sen claims that this demonstrates that 'in a very basic sense liberal values conflict with the Pareto principle ' (1970a, 157). This indicates that Sen thinks that liberalism is ultimately not pareto-efficient and pareto-efficiency is ultimately not liberal.
However, Sen's condition L* is too strong to capture anything like what we mean by liberalism. Condition L* specifies that at least some people must be able to do whatever they like at least sometimes. Though condition L* is weak in the sense that it doesn't require that everyone in a society have the right to decide something for themselves, it is on the other hand very strong in that it imposes no conditions on the choice that the two individuals are allowed to make. Even the staunchest libertarian would specify that one person's liberty must not infringe on the rights of others.
Sen's favourite example of a situation in which the liberal paradox arises involves two individuals, the novel Lady Chatterly's Lover, and three options: (x), that person 1 reads it; (y) that person 2 reads it; and (z) that no one reads it. 6 Person 1 prefers, in decreasing order, z, x, y. But person 2 prefers x, y, z. (Person 1 is a prude, whereas person 2 is a libertine who especially delights in forcing prudes to read erotic novels.) Liberalism dictates that each person should be able to decide for himself whether he will read the novel or not; thus, z is better than x, and y is better than z. But y cannot be the best option because both person 1 and person 2 prefer x to y, thus Pareto dictates that society should prefer x to y.
But obviously, liberalism is inconsistent with allowing society to force a prudish individual to read Lady Chatterly's Lover. In the given example, neither person values liberalism very much, since they each want to constrict each other's personal liberty. Sen acknowledges that the conflict only arises for an outside observer who does not want the Pareto principle to be rejected even when individual preferences are illiberal. He says, 'This position is slightly schizophrenic, but a great many people are schizophrenic in this sense ' (1970b, 86). On one hand, he has a point; people do have desires and preferences that conflict with their other desires and preferences. On the other hand, these particular schizophrenic preferences only arise in particular situations, i.e. when we ask one individual about what another individual should do. If we restrict the possible set of social choices to laws or policies that would apply to everyone the paradox disappears. Not because we couldn't construct a similar paradox with a smaller domain, but because if we restricted the possible choices in this way we would have to get rid of condition L* as well. That is, if we restricted the possible set of social choices to scenarios in which people mind their own business (no forcing anybody to read anything, to each his own, etc.) then condition L* would no longer be plausible or desirable. Condition L* would then require that at least two people get to decide some social policy, but if you don't count purely personal choices, there is no reason to believe that any individual's opinion should matter.
The real issue is not that Sen has mischaracterized liberalism. The problem is that in this construction of the paradox liberalism is essentially contentlessit is about letting one individual's choice be decisive without regard to the content of the choice. But as we saw with the Lady Chatterly's Lover case there isn't a problem necessarily with making liberal principles consistent with Pareto, there is a problem with reconciling the different preferences that individuals have into a social ranking without invoking a more robust value system. The liberal 'paradox' is essentially the paradox of a value-free social choice ranking. Sen specifies the paradox using very weak versions of the liberal and Pareto conditions, on the assumption that if we cannot reconcile weak forms of these two conditions then we cannot reconcile strong forms. He repeatedly invokes the idea that the weaker the conditions are the more disturbing the paradox should be to the reader. But in fact, the opposite is true, given the way that Sen has specified the liberal condition. On his specification the condition is weak quantitatively, in other words, it is weak with respect to the number of things that can be decided by the number of people. But a quantitatively weak liberal condition implies a qualitatively strong liberalism, i.e. a liberalism that had no restriction or conditions on the decision an individual should be able to make. The paradox then arises because there are no values or principles to govern the aggregation of the individual's preferences. It is not surprising that given at least two individuals who disagree on at least three possible choices that there should be a problem ranking the options according to a standard that is entirely neutral between the two. But this is not the situation that we find ourselves in when we are in the real world; in a real-life version of the Lady Chatterly's Lover dispute the third party would have to take a stand on sexual morality.
In Sen's more recent work, he has emphasized that his motivation when he was writing his impossibility result was to question Pareto efficiency (2009,311). This connects to his claim that social choice theory need not aim at complete orderingsthat partial rankings are instead the useful result. This means that in the case of Lady Chatterly's Lover, we can make overlapping partial rankings of preferences as a basis for further discussion, so we don't need to have the best choice (Sen 1997). So Sen doesn't think liberalism is itself paradoxical, or that this impossibility result is the end of the story. Sen explains the usefulness of impossibility results in social choice theory in his Nobel Prize lecture: "When a set of axioms regarding social choice can all be simultaneously satisfied, there may be several possible procedures that work, among which we have to choose. In order to choose between the different possibilities through the use of discriminating axioms, we have to introduce further axioms, until only one possible procedure remains … The real issue is not, therefore, the ubiquity of impossibility (it will always lie close to the axiomatic derivation of any specific social choice rule), but the reach and reasonableness of the axioms to be used. We have to get on with the basic task of obtaining workable rules that satisfy reasonable requirements" (1999a, 353-4).
So though Sen admits that the impossibility results of the social choice theory say more about the axiomatic method than they do the content of the axioms, he nevertheless holds that we can use this axiomatic method to identify rules to live by. If Sen's liberal paradox is representative of social choice theory, then this is doubtful. Because we didn't need the formal axiomatic method in order to find out that it is impossible to satisfy both the prude and the vengeful libertine. It is true that the paradox forces us to call the axioms into question, and if we had thought that these were reasonable requirements the paradox tells us that they are unworkable and need to be revised. But an impossibility result can never tell us which axiom or axioms need revision and how they can be revised. Further axioms can only provide guidance if there is some fixed point, some axiom or axioms that we hold to be more reasonable than the rest. Social choice theory alone cannot advance what Sen calls our 'basic task' since it cannot itself advance any normative claim.
Social choice theory can be considered a practical procedure only for a person with no preferences of their own tasked with figuring out how to satisfy most people. What is troubling about this as a methodology is that Sen emphasizes taking account of actual preferences from the point of view of a person that has no preferences. Thus, he is already supposing that the political theorist is working from this perspective, and that a valuable part of their task should be finding ways to satisfy preferences.

Capabilities
But I am far from having fully specified an example of the type of nonideal theory that Sen is proposing, since I have not yet explained the capabilities approach, the impartial spectator, or how these are meant to function together if they are. Sen's capabilities theory was originally intended as a response to economists who measured development in the dollar amount that an individual lives on per day, as well as a response to Rawls's second principle of justice, which Sen believes is too tied to the metric of income in order to serve as a plausible principle of justice. In both cases, Sen has pointed out that different individuals have different capabilities to use what they own, so money is a very bad measure of quality of life. So far as it goes, this has proven a useful way to think about development economics as well as the currency in which we measure equality. Sen has also suggested that the capability perspective can be used in multiple ways, such as assessing poverty, guiding health policy, or even assessing the inequality of overall advantages of different persons.
But it is unclear what role capabilities are meant to play when coupled with social choice theory and the impartial spectator. We might initially see social choice theory and capabilities as mutually exclusive theory forms, as social choice theory is a formalized mechanism for aggregating preferences or deducing incompatible types of preferences while capabilities are a way to understand how development or welfare should be measured. And aggregating or comparing preference sets is a separate task from measuring development. Each of these also seems separate from the device of the impartial spectator which is a way to think impartially, by imagining what a disinterested spectator would say.
Mozaffar Qizilbash observes that Sen has both a 'thin' and a 'thick' view of capabilities (2007). Sen's 'thin' view is merely that capabilities are normatively significant. Sen's refusal to identify a definitive list of capabilities can be understood in terms of his 'thick' view, which is his commitment to reasoned debate and public acceptability. Qizilbash notes that this would explain Sen's statements that disagreements that arise in the application of the capability approach should be seen as social choice exercises (Sen 1999a(Sen , 1999b. Qizilbash identifies two main areas in which social choice exercises could be used in the application of the capabilities approach, namely in identifying which capabilities are important in a given context as well as how to weigh various capabilities against each other. Qizilbash uses the Human Development Index as an example of an application of the capability approach. 7 The HDI is used by the United Nation to rank countries according to human development. It includes indices of life expectancy, education, and income. So, in the case of the HDI, which capabilities are measured as well as how they are combined with each other to make the human development ranking can both be seen as social choice problems. Sen interprets the use of social choice in these cases as a stand for democracy against technocracy (1999b, 79). But in terms of the theory's practicality, it is only practical for a technocrat. I will return to this point later when the whole of Sen's proposal is in view.
Besides using social choice theory to extend and apply the capability approach, Sen has also indicated that the capability approach can be used to extend social choice theory (1999a). Much of Sen's earlier work is dedicated to expanding the informational basis of social choice theory, beyond utility. For example, Sen has done pioneering work on the measurement of poverty, in which he has pointed out that measurements of poverty should be sensitive to the income distribution of the poor, rather than merely counting the number of poor people (1976). An example of Sen's claims is his Transfer Axiom, which states that measurements of poverty should show poverty as increased if there is a transfer of income from a poor person to a rich person, even if the number of poor people has been reduced as a result of this transfer. Sen's work in developing this and related axioms has led to ways to measure how poor people are as well as how many poor people there are. We can understand the HDI as a similar example. Each of these can be seen as ways to quantify capabilities so that they can serve as input in social choice exercises.
Ultimately, whether it is capabilities that extend social choice or vice versa, Sen is offering formal methods that lack any substantive normative content. As Qizilbash notes, the only normative commitment that follows from these theories is that public debate and democratic decision-making should determine public policy. Ingrid Robeyns notes more specifically that the capabilities framework is not in itself a feminist theory; it could be feminist or antifeminist depending on how it is used (2003). Social choice theory and the capabilities approach are at most methods for determining the will of a given group of people.

The impartial spectator
According to Sen, the Impartial Spectator can be seen as an alternative to the social contract. In social contract theory, the people of a society or some group that represents the people of a society come together to decide what rules they are to be subject to. Sen regards social contract theory, and specifically, Rawls's version, as deeply problematic because it cannot take account of outsider views. This is a very puzzling aspect of Sen's thought. Sen believes that the impartial spectator, which invites one to imagine what an impartial outsider would do, is an improvement on the social contract, in which we imagine what we would have agreed to. The dis-analogy is clear from the start, since the impartial spectator is clearly meant to solve ethical dilemmas from the individual's perspective, i.e. if I am wondering what to do, or tempted by what is probably the wrong thing to do, I should imagine what someone watching me would say in order to remain clear headed. The social contract, on the other hand, is meant to justify social policy, on the grounds that we could imagine that people would have agreed to the policy (its authority is derived from consent). Sen's concern for the outsider's perspective on the social contract is puzzling because as a thought experiment the social contract works in so far as the person who is subject to the policy would have agreed to it. If the outsider would not be subject to the policy under evaluation, then their opinion is irrelevant, i.e. the policy need not be justified to them. If the outsider is subject to the policy, then he or she is not truly an outsider; we would have to imagine that the outsider would have agreed to the social contract containing the policy (or if not then the policy is unjustified). Sen's concern for the outsider is largely motivated by a concern for global justice, which Sen believes cannot be accommodated by Rawls's social contract, since the Rawlsian contract doesn't take account of outsiders. This criticism of Rawls's use of social contract theory is especially strange because it seems to rest on conflating the hypothetical contract (which justifies a rule or policy in so far as we can imagine those subject to the policy would have agreed to it) with the actual laws of a given nation, which may affect non citizens in any number of ways. A hypothetical social contract can only justify a policy if anyone affected by it (or someone representing them) would have agreed to it.
It may be that Rawls does not take proper account of this point, and his idea of a social contract for a closed society is not attentive to the ways in which a nation is never 'closed'. 8 But if this is true, Sen's use of the impartial spectator does not seem more realistic than the social contract it is meant to replace, since both are hypothetical thought experiments. The impartial spectator is certainly is no less 'transcendental' than the Rawlsian contract. Sen prefers Scanlon's version of the social contract, because Scanlon's contract is directed towards principles which cannot be reasonably rejected, rather than principles that all can agree upon (Sen 2009, 197-200;Scanlon 1998). Sen believes this is a broadening of the social contract along the lines explored by Smith with the impartial spectator. But Sen exaggerates the difference between Scanlon and Rawls by referring to Rawls's theory as contractarian, as in involving a contract that is based on mutual advantage in the style of Hobbes, rather than a contract that respects the autonomy of individual persons, and thus values their consent, as in the style of Kant, which Scanlon famously called 'contractualist' (Scanlon 1982). 9 Ultimately though, Sen's use of the impartial spectator should be evaluated in light of his use of social choice theory and capabilities. I have suggested that social choice theory, even when combined with the capabilities approach, is merely a way of determining what values a given group of people holds. With the impartial spectator, Sen might head off one natural objection to his proposals. Sen acknowledges that any 8 In the beginning of A Theory of Justice he sets out this condition 'Let us assume, to fix ideas, that a society is a more or less self-sufficient association of persons … ' (4). He subsequently refers to this assumption as the assumption that a society is 'closed.' 9 Scanlon called his theory and Rawls's theory contractualist, though he pointed out differences between them. Scanlon contrasted both contractualist theories with utilitarian theories, using Harsanyi as an example. In this section of the book, Sen emphasizes the similarities between Harsanyi and Rawls. particular group of people may be limited by parochial values. The impartial spectator device is meant to guard against this, by providing an outside voice which should matter in community decision-making. Thus, it is, as it turns out, an integral part of Sen's larger view, in that it provides the normative content that social choice theory, and the capabilities approach lack on their own. Social choice theory and the capabilities approach are merely methods for determining the views of a group of people; the impartial spectator is a more substantive normative theory, in that it tells us which people's views should matter. Sometimes Sen speaks of the impartial spectator as a hypothetical voice, as when he is offering the impartial spectator as an alternative to Rawls's model of 'closed impartiality' -Sen's pejorative term. But sometimes Sen speaks of the impartial spectator as if it designates real people, as when he says that 'Smithian reasoning … requires consideration of the views of others who are far as well as near' and when he claims that each society should consider the experiences of other societies as relates to their own discussions of things like the status of women or capital punishment (2009,126,71). So, in so far as Sen is offering an alternative to 'transcendental theories', we can see that the impartial spectator is meant to be an improvement because it is a theory that can be modelled in the real world. Rawls's contract can't be modelled in the real world, since it relies on the veil of ignorance in order to yield impartial results. The impartial spectator provides a method to make decisions more impartial whether she is a thought experiment or a real person.
But as an alternative to Rawls's social contract, the impartial spectator is problematic; the social contract device only works if the contractual discussions are narrowed to include all and only those who will be subject to the contract. The Rawlsian social contract, in particular, will not give you a theory of justice as fairness at all unless the parties to the contract know that they are part of the societythe impartiality comes from the fact that anyone making the contract could be subject to it. This means that you can't have both an impartial (in the sense of fair to everyone, with no special treatment to certain people or groups) contract and a contract that avoids parochialism by including outside voices. If this is true then 'open impartiality' will, in fact, be an oxymoron and 'closed impartiality' will be redundant.
But it seems that Sen is not really concerned with opening up the hypothetical social contract, as he is with opening up actual public debate. This explains why he sometimes slips into referring to the impartial spectator as if it represents real voices rather than hypothetical ones. What Sen is truly advocating is including diverse voices in public debate, points of view that we might then analyse and aggregate using the formal method of social choice theory. This analysis can then be fed back to the public, in order to clarify the terms of their debate. This could be an iterative process that would continue until the people come to their decision.

The limits of Sen's nonideal theory
The limitation of this method is merely that it takes no stand on that decision. Sen is confident that the more voices that take part in public debate, the more liberal and respectful of others the debate eventually becomes. 10 But the theory has no built-in commitment to liberalism or respect, beyond the fact that 'outsider' voices should matter. The theory has even less of a commitment to equality. There is nothing in Sen's method that would temper a selfish or privileged voice in public debate. Social choice theory would help us narrow down the range of publicly acceptable policies, but this would include many objectionable policies that merely serve to benefit the selfish people. Using the capabilities approach, privileged people could come to better understand the nature of their fellow's suffering, and they could use this knowledge to compare alternative policies. But that is as far as it goes if they simply don't care about developing the capabilities of others.
Sen writes as if the majority of non-philosophical concern with justice takes the form of comparing two or more options. As if we are perennially in the position of making a choice. But in the real world, we are rarely in this position. And when we are, we often don't realize it until later. Sen's theory offers up the promise of clarity, but this is merely clarity about preferences, in the interest of finding compromise. A 'transcendental' theory like Rawls's, by contrast, offers to tell us why injustices are unjust. Sen seems to assume that this understanding is irrelevant, that comparison is all we need to do, and the comparison doesn't require clear ideals. One could argue that this is the case when the time comes to vote (if one is so lucky as to be able to vote). But the majority of political activity has nothing to do with voting, and it seems that clarifying ideals would be helpful in many of those cases. For instance when we recognize an injustice but struggle to imagine an alternative, when we puzzle over how to get others to recognize injustice, or when we attempt to understand someone else's plea to recognize injustice. All of these seem to be situations we could find ourselves in if we engage in public debate, as Sen is advocating. But it is unclear what the substance of this public debate would be, if we took his advice and avoided 'transcendental' theory.
The main limitation of Sen's nonideal theory with respect to Rawls's is that Sen's theory cannot diagnose injustice. Sen may say that diagnosis of an injustice is irrelevant to what we are going to do about it. But diagnosing why something is unjust can itself be doing justice. Sen's project supposes that we can find a method for the nonideal theory which does not presuppose any substantive political thesis. This attempt to substitute a formal method for an evaluative method is the reason that Sen's theory is unsuited for the realization of any particular political goal.
So on Sen's own terms, we cannot use these methods to do anything about injustice unless we are already in a position of power. Sen's theory purports to be more practical than Rawls's because it provides clear instructions for a person who has the power to single-handedly make changes. But no matter how clear the instructions, a theory like this should not be considered practical if we are not in a position to implement those instructions. It is not practical to use social choice to determine compromise between preferences, or to evaluate poverty, or to ask what an impartial spectator would do because (1) we have values already, so we are not blank slates, and (2) we are not in a position to implement any of this knowledge. This is a theory that is practical for those in positions of power and especially when they imposing reforms from the outside. Sen's commitment to public debate is perhaps the most unrealistic aspect of Sen's nonideal theory because it gives us no criteria with which to engage in these public debates. Instead, it idealizes a procedure which is very far from our experience.
If we imagine Sen's recommendations being implemented, we can only imagine them being implemented by some kind of expert in governance, who is understood to be acting from a position outside of that society's substantive political disputes. Because social choice, the capabilities approach, and the impartial spectator are for Sen procedures without substantive normative commitments, a person using them in practice would not be thereby furthering a particular substantive commitment to what justice entails. And furthermore, a person could only use these procedures if they were given the power to do so. Thus, a society like this would certainly be considered technocratic by most. It would be a society ruled by experts, in which the experts are considered neutral and not partisan. Undoubtedly Sen's disavowal of technocracy is based on the idea that his technocrats are not themselves pronouncing or implementing their own independent ideas of justice, and yet this would seem to be an insufficiently narrow idea of what technocracy is. Those who would promote technocracy do so on the grounds that experts are better equipped to rule because they have greater access to the truthbut this truth is always understood to be factual and not based on values. So Sen's idea of a technocrat that might override some democratic decision in favour of a particular idea of justice doesn't make sense on closer inspection. If there was such an antidemocratic ruler they would not, by definition, be a technocrat, because they would not be considered to be promoting any fact-based knowledge. In actuality, technocrats promote themselves as rising above disputes about justice on the basis of their knowledge, which is exactly the position that Sen's proposals are in with respect to Rawls. Rawls's theory, because it is a theory of justice, is rejected by Sen in favour of a cluster of procedures that do not concern themselves with what justice actually is.
Ultimately, it is possible to reject ideal theorizing about justice without thereby promoting technocracy. There is something to the argument that, for example, we should focus on theorizing about injustice before theorizing about justice. But Sen is not critically analysing injustice any more than he is critically analysing justice. He is instead promoting two formal methods for understanding the preferences of a group, social choice theory and the capabilities approach, along with a method to counteract bias, the impartial spectator, that are meant to replace this kind of analysis. By doing so, he is presupposing there is some way of using technical knowledge to rise above partisan disputes about justice within the society, which is fundamentally technocratic.