In A Theory of Justice, John Rawls argues that it is possible to describe an appropriate initial situation from which to form reliable judgements about questions of justice.Footnote 1 This initial situation, which Rawls calls the ‘original position’, models the practical implications of fundamental and shared intuitions regarding justice, in particular intuitions concerning the role that considerations regarding fairness, impartiality, and the avoidance of arbitrariness should play in grounding judgements of justice. Once these intuitions are modelled faithfully, Rawls argues, the structure of the resulting choice problem dictates that rational persons occupying the original position should apply a maximin rule of choice in choosing principles of justice to regulate their joint social relations. As a result, the structure of the choice problem narrows the set of acceptable choices sufficiently to require the choice of a unique pair of principles. According to this argument (the ‘maximin’ argument), then, the constitution of an account of justice may accurately be represented as a problem of rational choice under uncertainty. The maximin argument plays a central role in Rawls's justification of his account of justice in A Theory of Justice, and Rawls continued to emphasize the centrality of this argument to the justification of his theory throughout his discussions of the original position procedure, from the 1970s through his last published works.Footnote 2
It is a striking irony that while A Theory of Justice is widely regarded as the most influential work of political philosophy of the last century, the maximin argument – which constitutes Rawls's most salient justification for the acceptance of his two principles of justice – has widely been dismissed as unpersuasive. This argument is often described, even by sympathetic interpreters, as ‘perhaps the single worst argument in Theory’ (Wolf Reference Wolf, Davion and Wolf2000: 103–104). Indeed, as Rex Martin suggests, the credibility of Rawls's entire theoretical project has been undermined by his reliance upon the apparently unpersuasive logic of the maximin argument (Martin Reference Martin1985: 102).
The current unsympathetic view of the maximin argument is the product of a number of apparently devastating critiques published during the decade after the initial publication of A Theory of Justice. Rawls's argument, it was argued, relies upon an imprecise account of the satisfactory minimum to be secured under the maximin rule. In the absence of specific information regarding the minimum share of goods that the choosers may secure for themselves, Brian Barry argues, it is impossible to determine ‘whether . . . maximizing the minimum is a good idea or not’ (Barry Reference Barry1973: 98; see Buchanan Reference Buchanan, Blocker and Smith1980). Moreover, critics claimed, the maximin rule is not the only or even the most attractive decision rule that is available to rational persons facing choice under extreme uncertainty. In particular, John Harsanyi argues that the appropriate decision rule for choice in the original position is the Bayesian expected-utility maximization rule (Harsanyi Reference Harsanyi and Harsanyi1976). The objections thus criticized both Rawls's analysis of the choice under uncertainty problem facing choosers in the original position and his choice of a decision rule to regulate that choice. As will be discussed below, these objections have strongly influenced the interpretations of Rawls offered by Ken Binmore, Daniel Hausman, Michael McPherson, John Roemer, Robert Taylor and Philippe van Parijs, among others (see Binmore Reference Binmore1994: 146; Hausman and McPherson Reference Hausman and McPherson1996: 155; Roemer Reference Roemer1998: 172–176; van Parijs Reference van Parijs and Freeman2003; Taylor Reference Taylor2011: 194). These objections have been particularly influential both because they were offered by leading rational choice theorists and because they appealed to assumptions widely shared among such theorists, such as the view that one decision rule – the Bayesian criterion – is uniquely appropriate to govern rational choice under uncertainty. While Rawls could respond persuasively that the particular nature of choice in the original position – involving both extreme uncertainty and a choice problem involving the selection of principles rather than goods – differed from the conditions that characterize standard problems in rational choice, criticism of the maximin argument on grounds persuasive to rational choice theorists seriously undermined its credibility.
These objections were formulated more than thirty years ago and have been more or less taken for granted since that time. I will argue, however, that the maximin argument is more robust than has generally been recognized and that this argument performs a number of important functions in clarifying the nature and implications of Rawls's argument for justice as fairness. First, Rawls's employment of the maximin rule as a central element of his argument for the two principles underlines the fact that justice as fairness is not an allocative theory designed to regulate the division of a collection of goods among definite individuals.Footnote 3 Rather than providing support for an account of justice that focuses on (i) securing an optimal distribution of expected utility or (ii) maximizing the share of the least advantaged, the maximin argument ‘forc[es] us to consider what our fundamental interests really are’ (Rawls Reference Rawls and Kelly2001: 99). Second, the maximin rule guides reflections in the original position regarding the choice of (i) standards to determine the acceptability of basic social institutions, not (ii) guaranteed shares of objects or goods. Rawls's maximin justification of the principles of justice therefore develops, not an account of the value of goods or social surplus to be realized through application of the principles, but rather an account of the reasons that require the adoption of standards to regulate basic social institutions. Third, a conception of justice capable of satisfying the maximin rule must establish conditions that – taken as a whole – provide satisfactory protections for each person's fundamental interests. This focus on the realization of social conditions that secure protections for fundamental interests establishes a clear contrast between justice as fairness and consequentialist theories of justice that aim primarily to compensate persons for deficits in well-being. Finally, the maximin argument illustrates the potential of Rawls's ethical constructivism to identify structural features of moral questions that are not obvious from a merely discursive presentation of those questions. Once it is accepted that the original position constitutes the uniquely appropriate standpoint from which to form judgements of justice, Rawls argues, a careful consideration of the structure of the particular conditions of choice in the original position identifies the maximin rule as the uniquely appropriate decision rule to govern judgements formed from that standpoint.
In order to develop my argument, I will first sketch Rawls's account of choice under uncertainty, emphasizing the structure of the choice problem that faces Rawls's hypothetical choosers. I will then examine the objection, developed by Brian Barry and Allen Buchanan, to Rawls's claim that a rational chooser in the original position will feel no need to secure advantages above a guaranteed satisfactory minimum. The objections presented by Barry and Buchanan that are discussed in this section constitute the most influential examples of the argument that the maximin argument relies upon an imprecise account of the satisfactory minimum to be secured under the maximin rule. I will reject the assumption – central to this objection – that the satisfactory minimum that persons employing the maximin rule seek to achieve is a minimum level of primary goods. Next, I will examine John Harsanyi's criticisms of the maximin choice criterion – the most influential example of the objection that the maximin rule is not the only or even the most attractive decision rule that is available to rational persons facing choice under extreme uncertainty, Here, I will argue that Rawls provides persuasive arguments for rejecting the view that the Bayesian criterion could be an acceptable criterion to guide choices of fundamental importance under conditions of complete uncertainty. Finally, I will argue that the application of the maximin rule under the specialized social contract conditions of the original position performs an important function in focusing deliberations on identifying the central concerns that an acceptable conception of justice must address. Both this final argument and the discussion of Harsanyi's argument respond to the objection that the maximin rule is not the most attractive decision rule that is available to rational persons facing choice under extreme uncertainty
1. THE MAXIMIN ARGUMENT
If we accept the claim that a just distribution of social goods should not be determined by ‘factors [that are] arbitrary from a moral point of view’ (63), Rawls argues, then we will agree that deliberations regarding the nature and requirements of justice should be conducted from the standpoint of an original position in which information is not available regarding a society's (i) endowments of resources, (ii) current level of wealth and economic development and (iii) distribution of wealth, income and power. If, in addition, we accept the claim that persons should not be able to ‘tailor principles [of justice] to the circumstances of [their] own case’ (16), Rawls argues, then we will agree that information regarding one's own situated interests should not be available. Rawls represents these informational constraints through the imposition of a veil of ignorance that deprives deliberators of information regarding (i) their society's wealth and level of development and (ii) their situated interests – in particular, information regarding both social endowments, which include inherited wealth, social position and class advantages; and natural endowments, which include talents, abilities, intelligence, ambition, and other physical and psychological traits. In depriving the choosers of information regarding their endowments, the original position represents concretely the intuition that an appeal to the mere possession of an inherited trait or status does not provide a satisfactory justification for a claim to goods: justice should not, for example, reward people simply for being intelligent or talented, although it may reward the employment of those qualities (87–89).
In order to prevent all appeals to arbitrary factors in deliberations about justice in the original position, Rawls argues, the veil of ignorance must deprive persons not merely of knowledge regarding their specific endowments, but also of knowledge regarding the nature of the society they will occupy when they emerge from behind the veil. Otherwise, arbitrary particular facts regarding a society's level of development and wealth at a particular point in history might determine the parties’ general judgements regarding the requirements of justice; and rational choosers pursuing their own interests would be likely to base their choice of principles on probabilistic assessments of the likely effects of the principles chosen on relative advantage.
Rawls argues that it is rational for persons facing such a constrained choice problem under uncertainty to employ a maximin rule of choice – a rule that instructs the chooser to select that option that secures the most ‘satisfactory minimum’ state of affairs (133–135). While the maximin rule is not appropriate for all, or even most, choices under uncertainty (133), Rawls argues that it is the appropriate rule to regulate judgements behind the veil of ignorance because of three features of that choice position.Footnote 4 First, choosers in the original position know nothing about the probabilities of possible outcomes. Second, Rawls asserts that rational choosers in such a position of uncertainty will be more concerned to secure a satisfactory minimum than to secure the possibility of receiving greater advantages. If potential losses and gains are both unlimited, Rawls asserts, it is rational to be more concerned to avoid the worst possible outcomes than to insist upon preserving the possibility of the greatest possible gains. In particular, if the choosers are able to ensure that the minimum guarantee is an attractive one, they may feel little need to ensure the possibility of securing significantly greater advantages (135). Third, the choosers in such a position will insist upon ruling out completely certain unacceptable outcomes. If, for example, slavery is a real possibility – as it must be for persons behind a veil of ignorance – and if a person can eliminate that possibility simply by choosing a principle forbidding slavery; then, Rawls argues, any rational person would insist upon the choice of that principle.
It is important to emphasize that Rawls does not claim that choosers in the original position will seek to achieve the guarantee of a minimum income or bundle of primary goods. Rather, Rawls argues that the ‘satisfactory minimum’ that choosers will attempt to secure constitutes ‘an adequate minimum conception of justice’ – that is, the conception that provides the most satisfactory minimum guarantee of protections of their fundamental interests (153). In particular, Rawls argues, the choosers will choose a theory that: (i) minimizes invasions of fundamental rights, (ii) promotes equal opportunity to develop and exploit their talents and (iii) mitigates the inequalities that continue to exist in a social order that ensures equal opportunity.
2. THE NATURE OF A SATISFACTORY MINIMUM
Brian Barry offers an influential objection to this argument that was later elaborated by Allen Buchanan. Barry and Buchanan object to the claim that a rational chooser in the original position will feel no need to secure advantages above a guaranteed satisfactory minimum. In the absence of information about the precise minimum share of goods that may be secured by the maximin criterion, Barry argues, ‘we cannot say . . . whether in any particular situation maximizing the minimum will be a good idea or not’ (Barry Reference Barry1973: 98). In addition, Buchanan argues, in making such a claim, Rawls relies upon unrealistic assumptions about diminishing marginal value. Both of these objections, however, focus on the size and value of a share of goods to be assured under the maximin rule. As a result, I will argue, both objections reflect a confusion regarding the character of the satisfactory minimum that Rawls argues that rational choosers will seek to secure through their choice of principles. In particular, as I discuss below, both objections falsely assume that the satisfactory minimum sought under the maximin rule is a share of primary goods assured under the difference principle.
2.1 A minimum threshold and diminishing marginal value
Barry argues that choosers in the original position will only view the maximin rule as an attractive criterion if they can be sure that the minimum share of primary goods attainable under the rule will be at least equal in value to the threshold amount above which they care little for any gains. Unless they can be certain that the attainable minimum falls at or above this amount, Barry argues, the choosers cannot be certain whether they prefer to maximize the minimum or to adopt some other principle (Barry Reference Barry1973: 97–98). In particular, if the guaranteed minimum falls below the minimum threshold amount, the choosers will assign positive value to gains above that minimum. Moreover, Buchanan asserts, Rawls's claim that a chooser cares little for any gain above a minimum stipend requires an implausibly strong assumption about the diminishing marginal value of goods (Buchanan Reference Buchanan, Blocker and Smith1980: 27). It may be plausible, Buchanan concedes, to claim that there exists some minimum threshold amount above which gains are of negligible value compared with the disutility of falling below that minimum (27); but it does not follow that parties who are assured of this minimum will care very little for gains above it. Moreover, Buchanan asserts, Rawls cannot consistently assume such a level of diminishing marginal value because Rawls, himself, attacks utilitarians for making similarly extreme assumptions regarding diminishing marginal value.
Defenders of Rawls's theory have responded to these objections by arguing that choosers in the original position would focus on the minimum guarantees secured by the principles because securing such minimum protections will: (i) strengthen the bases of self-respect and secure social stability (Cohen Reference Cohen1989: 736–750); (ii) elicit cooperation from all members of society (Pogge Reference Pogge1989: 264); or (iii) vindicate intuitions reflected in the structure of the original position (see Kymlicka Reference Kymlicka1990: 70).Footnote 5 Such responses, however, fail to identify the most serious shortcoming of the objections offered by Barry and Buchanan.
These objections are unsound because they are based upon the assumption that the satisfactory minimum that the choosers seek to achieve is a minimum level of primary goods (see Buchanan Reference Buchanan, Blocker and Smith1980: 27).Footnote 6 Such a view conflates the idea of a satisfactory minimum with the idea of a guaranteed minimum level of primary goods, because such a view assumes that it is the difference principle that secures the satisfactory minimum sought under the maximin rule by guaranteeing a minimum bundle of primary goods to the least advantaged members of society. This view, however, is contradicted by Rawls's explicit account of the character of the satisfactory minimum sought under the maximin rule.
2.2 A satisfactory minimum
The view that the satisfactory minimum sought by the choosers in the original position is simply a share of primary goods – in particular, the share of income or primary goods to be allocated to the least advantaged – is so widely shared (see, for example, Arrow Reference Arrow1973: 251; Barry Reference Barry1989: 213–217, 226–241; Binmore Reference Binmore1994: 146; Roemer Reference Roemer1998: 172–176; van Parijs Reference van Parijs and Freeman2003; Taylor Reference Taylor2011: 194) that it is remarkable to consider how completely such a view is contradicted by Rawls's account of the deliberations in the original position. Rather than a share of primary goods, the satisfactory minimum sought by the parties is ‘a satisfactory conception [of justice]’ that is ‘assured by the two principles in lexical order’ (135). A conception of justice is defined as (i) ‘a set of related principles for identifying the relevant considerations which determine . . . a proper balance between competing claims’ (9) and (ii) associated priority rules (37–38, 266). During deliberations in the original position, the parties focus on the task of ranking conceptions of justice by their acceptability.Footnote 7 In order to select the most acceptable conception, the parties assess ‘a definite list of traditional conceptions’ (102) and choose from that list the conception that constitutes the most satisfactory ‘minimum conception’ (153) of justice. The task of the choosers is therefore to assess the acceptability of conceptions of justice. The most acceptable conception must (i) provide the most adequate protections for citizens’ fundamental interests and (ii) establish the right kind of priority between claims grounded in competing fundamental interests. Thus, the parties in the original position focus, not on choosing an allocation of primary goods to be assigned to the least advantaged, but rather on assessing the character of different conceptions of justice – the kinds of interests that they protect, the kinds of balance that they establish between fundamental interests, and the kind of political and social world that would result from the adoption of each conception.
The choosers thus apply the maximin rule to a very specific task – they assess the substance and practical implications of a wide range of proposed conceptions (e.g. justice as fairness, classical utility theory, perfectionism, intuitionism, mixed conceptions), and they aim to choose a conception that is satisfactory in the sense that it provides the most satisfactory protections for the fundamental interests of citizens corresponding to the ‘two coordinate roles’ of the basic structure of society: (i) securing equal basic liberties and (ii) providing background institutions that secure social and economic justice (53; see Rawls Reference Rawls and Kelly2001: 48).Footnote 8 In addition, and crucially, the choosers consider the priority ranking that each proposed conception establishes between conflicting claims grounded in these interests. Rawls emphasizes that a conception of justice is defined not merely by the principles it contains, but by the priority rules that determine the relative weight to be assigned to the requirements of each principle.Footnote 9 The priority that his conception establishes among citizens’ fundamental interests is central to Rawls's argument that choosers in the original position would prefer that conception to utilitarianism.Footnote 10 Thus, a satisfactory minimum conception is satisfactory because of the relation that it establishes between content and structure.Footnote 11
Rawls develops a number of arguments to support his claim that a conception of justice consisting of his two proposed principles and the associated priority rules constitutes a satisfactory minimum conception. Justice as fairness provides the most satisfactory minimum protection of liberty interests, Rawls argues, because it secures their priority over all other fundamental interests. Rawls develops this argument by noting that the acceptance of any conception of justice will generate ‘strains of commitment’ – strains generated by the justified reservations of persons subject to principles of justice regarding the effects of those principles (153–154). The most severe strain on commitment, Rawls suggests, occurs in cases in which some must ‘acquiesce in a loss of freedom . . . for the sake of a greater good enjoyed by others’ (154). A conception of justice that parties to the choice ‘can rely on one another to adhere to’ (157) must therefore provide the most complete protection possible against this kind of injustice; and Rawls's first principle, with its foundational commitment to the inviolability of the person, responds to the strains of commitment more effectively than any other viable alternative principle because it insures completely ‘against the worst eventualities’ involving losses of liberties (154). Thus, justice as fairness will minimize the strains of commitment, Rawls argues, precisely because it guarantees the most acceptable minimum protection of liberty interests.
Justice as fairness provides the most satisfactory protection of the interest in the fairness in the distribution of social burdens and benefits because, in addition to ensuring equal opportunity and fair compensation, it provides the most satisfactory minimum guarantee regarding the social bases of self-respect – and, Rawls argues, self-respect is an essential primary good required for the successful pursuit of the individual's conception of the good.Footnote 12 General affirmation of the second principle – and thus shared agreement to an arrangement that secures basic liberties and aims to mitigate the influence on life chances of factors that are arbitrary from the moral point of view (see 156) – publicly expresses the respect of each member of a well-ordered society for every other member.Footnote 13 This public expression of respect, Rawls notes, provides support to the self-respect of members of society that would not be provided under even a generous utilitarian conception that accepted the liberty principle if that conception did not aim to mitigate arbitrary influences on life chances.Footnote 14 Thus, in choosing the principles of justice as fairness, the choosers ‘insure their self-respect as it is rational for them to do’ (156).Footnote 15
It is the conception taken as a whole that constitutes the satisfactory minimum: the requirements of the chosen principles ‘are tied together as one conception of justice which applies to the basic structure of society as a whole’ (136; Rawls Reference Rawls and Kelly2001: 99),’ and the fact that this conception secures a ‘satisfactory political and social world’ – not merely a bundle of goods, resources, and protections – ‘is crucial for the argument’ (Rawls Reference Rawls and Kelly2001: 100). The choosers have a rational interest in ensuring that each member of society is guaranteed both the liberty and the effective capacity to pursue his or her conception of the good.Footnote 16 Justice as fairness provides the most satisfactory protection of this interest because of the combined effect of its basic guarantees, which unite the assurances that (i) liberty interests will be assigned priority over all other social claims; (ii) arbitrary factors will not justify inequalities of opportunity or social status; and (iii) society will honour legitimate expectations of fair compensation (273–277). Rawls refers to this combination of guarantees, and not to a minimum guaranteed income or bundle of primary goods, when he argues that a chooser in the original position would ‘care very little, if anything, for what he might gain above the minimum’ guaranteeable level (134).Footnote 17
It could be argued, however, that Barry's and Buchanan's objections can be reformulated in a manner that avoids conflating of the idea of a satisfactory minimum with a guaranteed level of primary goods supplied under the difference principle. According to this view, both the first objection – that the maximin rule will only be attractive if the attainable minimum exceeds some threshold level of value – and the second objection – that the maximin argument is grounded in an implausible assumption about the diminishing marginal value of wealth and other primary goods – could be reformulated to refer to the value of the protections of fundamental interests provided by the two principles rather than to the value of a bundle of primary goods.
A reformulation of the Barry and Buchanan objections, however, would require a precise estimate of the value of the interests protected by the two principles – and no such precise estimate is available, because the two principles do not guarantee the provision of any particular social goods, resources, or level of welfare. The claim that the two principles secure the most satisfactory minimum protection of citizens’ fundamental interests, in fact, involves a characterization of the principles’ adequacy in addressing fundamental interests and not a description of its sufficiency in guaranteeing resources or welfare.
Both the Barry and the Buchanan objections, however, depend upon the idea of a minimum guarantee of a bundle of objects with specifiable and determinate values. Thus, in order to illustrate his claim that the maximin rule is attractive only if a threshold level of value can be achieved, Barry offers an example in which the maximin rule is unattractive because only ten apples can be provided to each person, while the threshold level with respect to apples is twelve (Barry Reference Barry1973: 98). Since the guaranteeable minimum is below the threshold level, Barry argues, the maximin rule would be less attractive than a rule requiring that we give twelve apples to as many people as possible while allowing a few people to receive fewer than ten apples.
The fundamental interests protected by a satisfactory minimum conception of justice, however, are unlike apples or other material goods because those interests are not objects or goods with a definitive form and specifiable value. Nor is it plausible to suggest that some unique threshold exists below which the protection of these interests would become so much less desirable that members of society would necessarily judge that the decision rule requiring the protection of fundamental interests should be replaced by some other rule (e.g. a rule requiring maximizing expected utility). To see why this is true, consider the following example. A stable and homogeneous society initially guarantees a satisfactory system of equal basic liberties. This commitment to the protection of liberties reflects the generally shared judgement among members of the society on due reflection (on the basis of arguments similar to those that Rawls presents) that a just society should ensure the most satisfactory feasible minimum protection of fundamental interests. After the society has been destabilized by violent conflict resulting from religious discord, however, the level of liberty protections that can be sustained is more limited than the original system of protected liberties – people must submit to metal detector tests, invasive searches, curfews, and other restrictions on liberty. The level of protection for liberties, in fact, falls below the threshold of acceptability for many members – perhaps a majority – of the society. Would such circumstances justify abandonment of the decision rule that required selecting principles that protect fundamental interests? The short answer would appear to be no. Due reflection – a careful consideration of all relevant considerations under conditions conducive to sound reflection subject to none of the usual errors in reasoning – justified the adoption of the maximin decision rule. Members of society might be justified in abandoning this decision rule on the basis of considerations relevant to the grounds for the choice of that rule – for example, if Buchanan could justify his claim that the reasoning supporting the argument for the maximin rule relied upon an implausible assumption about diminishing marginal utility. The fact that social conditions prevent the optimal protection of one of the set of fundamental interests whose protection is required by the decision rule, however, is not a consideration relevant to the reasoning that justified the adoption of the decision rule.
Even a reformulated version of Barry's objection fails to ground a relevant objection to the maximin argument because a choice problem involving the selection of principles differs in character from a choice problem involving the selection of a minimum guaranteed supply of goods. In Barry's example involving a minimum supply of apples, apples are determinate goods with identifiable value, and the value of the bundle of apples that can be supplied under the decision rule is a consideration directly relevant to the justification of the decision rule itself. If the value of that bundle falls below a certain threshold, that fact could – as Barry notes – justify the employment of a different decision rule. The decision facing persons in the original position, however, involves the choice of principles of justice. A principle of justice is not an object or good, but rather a standard that can be employed to assess the acceptability of institutions, legislation, and policy. The choice of such a standard reflects an underlying judgement regarding the kinds of reasons that can justify an institution, legislation or policy. The choice of a principle requiring the protection of liberty interests, for example, reflects the judgement that the fact that a policy or piece of legislation promotes or protects liberty constitutes a sufficient reason to support or adopt that policy or legislation. Such a choice is justified, then, not by consideration of the value of the objects supplied to persons under the principle, but rather by consideration of the reasons that justify affirmation of the principle. The value realized or lost as a consequence of adopting the principle may be relevant – for example, if adopting a principle would destroy all social value – but the mere fact that persons would assign a lesser value to a form of implementation of a principle that they still endorse on due reflection would not count as a consideration justifying the abandonment of the decision rule that led to the choice of the principle. Since Barry's argument necessarily assumes that the objects chosen in the decision procedure have a specific form and a specifiable value, his objections cannot readily be reformulated to apply to a choice problem in which the choice focuses on principles rather than goods.
The objections raised by Barry and elaborated by Buchanan nevertheless continue to influence many leading interpretations of Rawls's work. Many leading scholars have been influenced to identify the maximin rule with the difference principle and to view the choice in the original position as concerned solely with the distribution of goods over persons. John Roemer, for example, claims that the task of choosers in the original position is ‘the choice of [a] distribution [of goods] from behind a veil of ignorance’ (Roemer Reference Roemer1998: 177). Ken BinmoreFootnote 18 and Philippe van ParijsFootnote 19 conflate the idea of a satisfactory minimum under the maximin rule with a guaranteed level of primary goods under the difference principle. Daniel Hausman and Michael McPherson claim that the choosers in the original position would design social and political institutions ‘to advance maximally the interests of the worst-off group’ (Hausman and McPherson Reference Hausman and McPherson1996: 155). And Robert Taylor, in his recent book on Rawls, claims that the reasoning of the choosers in the original position can be summarized as follows: (i) the difference principle is the maximin rule among distributive principles; (ii) the maximin rule is rational under three conditions; (iii) these conditions hold in the original position; therefore (iv) the parties in the original position should choose the maximin rule among distributive principles (Taylor Reference Taylor2011: 194). The analyses of Rawls's arguments presented by these leading scholars – all of whom cite Barry's The Liberal Theory of Justice as a leading interpretation of Rawls's arguments – are thus distorted by fundamental confusions regarding the substance and role of the maximin argument deriving from Barry's critique.
3. MAXIMIN AND CHOICE UNDER UNCERTAINTY
Harsanyi's objection focuses on the plausibility of the maximin rule as a decision criterion for rational choice under uncertainty. While the maximin rule was viewed as an optimal decision rule when the issue of rational choice under uncertainty first attracted attention, Harsanyi claims, the limitations of that approach quickly became apparent (Harsanyi Reference Harsanyi and Harsanyi1976: 38–39). Consider, he suggests, a choice between option A, in which a New York City resident interviews only for mediocre jobs in New York, and option B, which involves the opportunity to interview for a superior job in Chicago. The worst outcome that could result from choosing option A is mediocre employment; but the worst outcome that could result from choosing option B is death in a plane crash. A maximin rule would therefore require that a rational person must always select option A. Since the maximin decision rule produces such implausible results, Harsanyi concludes, it cannot provide an appropriate decision criterion for rational choice under uncertainty (Harsanyi Reference Harsanyi and Harsanyi1976: 39–43). Rather, the appropriate decision rule is the Bayesian expected-utility maximization rule.
Note, however, that this objection is not responsive to Rawls's argument that maximin is an appropriate decision rule only under the carefully defined conditions of uncertainty that hold in the original position – extreme conditions of uncertainty in which the choosers possess no information at all about probabilities. Harsanyi's example merely demonstrates that maximin is an inappropriate decision rule for choices under conditions in which choosers have a great deal of information about probabilities, although the precise probability of each outcome is not known. Harsanyi's objection, then, does not address, much less refute, Rawls's arguments that the maximin rule is the appropriate criterion for rational choice under the conditions of extreme uncertainty that exist in the original position.
Harsanyi's objection, however, is motivated by the view that the criterion for rational choice does not and should not vary under differing choice conditions.Footnote 20 According to such a view, an argument demonstrating that maximin is an inappropriate decision rule for any particular category of rational choices necessarily establishes that maximin is an inappropriate decision rule for all categories of rational choices. If this view is accepted, Harsanyi's example counts decisively (i) against Rawls's maximin argument and (ii) in favour of the claim that the Bayesian expected-utility maximization rule is the uniquely appropriate decision rule for choice under uncertainty. This strand of Harsanyi's argument has been influential, particularly among rational choice theorists (see Goldman Reference Goldman, Blocker and Smith1980; Hardin Reference Hardin1988: 135; Barry Reference Barry1989: 215; Hare Reference Hare1989: 107). Harsanyi's argument is, however, vulnerable to a number of serious objections.
It is important to note, first, that the claim that expected-utility maximization is the uniquely appropriate decision rule for choice under uncertainty is ‘highly controversial’ (Hausman and McPherson Reference Hausman and McPherson1996: 31) among rational choice theorists (see Ellsberg Reference Ellsberg1961; Allais Reference Allais, Allais and Hagen1979; Sen Reference Sen1985; Levi Reference Levi1986; McClennen Reference McClennen1990). As Daniel Hausman and Michael McPherson note, talk of maximizing expected utility ‘is obscure when probabilities and even the range of possible outcomes are not known’ (Hausman and McPherson Reference Hausman and McPherson1996: 31). Decision theorists such as Isaac Levi argue instead that it would be rational for a person choosing under conditions of uncertainty to follow a rule requiring the choice of only those options that would produce the highest security level when combined with at least one hypothesis that is true if that option is chosen (Levi Reference Levi1984: 132; see Levi Reference Levi1967, Reference Levi1974, Reference Levi1980). According to this view, when expected utility considerations fail to require the choice of any particular option, reliance upon a secondary decision criterion is rational. Indeed, as Erik Angner notes, the information available to the parties in the original position underdetermines rational choice in precisely the manner described in Levi's argument (see Angner Reference Angner2004: 16).Footnote 21 The parties know nothing about (i) the distribution of wealth and other resources or (ii) their own conceptions of the good. They therefore lack sufficient information to assign utilities to outcomes, much less to maximize expected utility, so that reliance upon a secondary decision criterion is rational.
A number of leading defences of the maximin argument, including those presented by Gail Corrado (Reference Corrado, Blocker and Smith1980) and Binod Agarwala (Reference Agarwala1986), have offered similar responses to the objections raised by Harsanyi – they have argued, that is, that the information available to the choosers is inadequate to ground calculations of expected utility, and that under such conditions maximin reasoning is a rational alternative. While these theorists argue that the Bayesian criterion may be supplemented or replaced in conditions in which that criterion cannot be applied or its meaning is obscure, Rawls presents arguments in Sections 27 and 28 of A Theory of Justice that provide a persuasive basis for rejecting altogether the view that the Bayesian criterion could be an acceptable criterion to guide choices of fundamental importance under conditions of complete uncertainty.
First, the Bayesian criterion is objectionable as a criterion to guide significant social choices under conditions of complete uncertainty because choosers under such conditions have no basis for estimating the subjective probabilities necessary in order to calculate expected utility, while assessments of probability that are to serve as the basis for rational decision ‘must have an objective basis’ (149). The necessity of such an objective basis becomes more urgent as the importance of the judgement increases. A decision involving the choice of principles which will determine the extent to which the most fundamental of human interests receive protection would clearly involve the highest degree of urgency. This urgent need cannot be met in conditions of complete uncertainty. Even if it were possible to assign an equal probability to an unknown number of outcomes, the choosers would have no objective basis for that probability assignment. As Samuel Freeman notes, if the choosers lack sufficient information to assign any set of probabilities to outcomes, they also lack sufficient information to assign equal probabilities (see Freeman Reference Freeman and Freeman2003: 15–18). Parties forming judgements regarding fundamental questions of justice under conditions of complete uncertainty would therefore lack a sufficient informational basis to apply the principle of maximizing expected utility.
In addition, an individual who employs an expected utility criterion in conditions of complete uncertainty cannot base her judgements upon a unified system of preferences. While expected utility calculations are usually derived from the perspective of an individual with a single unified system of preferences, a chooser in the original position does not have ‘aims which he counts as his own’ (150). Any calculation of expected utility from this perspective must therefore incorporate the hypothetical utility functions of many individuals. An argument for the employment of the Bayesian criterion, then, would have to assume that reliable interpersonal comparisons of utility can be made – a highly controversial view.Footnote 22 Even putting that problem aside, however, Rawls argues persuasively that expected utility calculations based upon the assumption that one may turn out to be any one of a number of persons are of doubtful validity (150). In particular, it is not clear what scale of value one should employ in evaluating the worth of another person's way of life and system of ends – should one apply one's own scale of values or the scale of values favoured by that person? Conflicting claims of justice arise not merely because people want similar sorts of things but because their conceptions of the good life differ. Even the choosers in the original position are aware that their final ends differ and are subject to no commonly acceptable criterion of value. Thus, no plausible criterion of value is available to provide a basis for attempts to calculate expected utility from a standpoint of complete uncertainty.
This problem is particularly acute for an argument in favour of the Bayesian criterion of choice, Rawls notes, because such an argument requires a unified account of the chooser's expectations. Such an argument must therefore, it would seem, assume that everyone thinks of themselves as having the same utility function. The argument for the Bayesian criterion is thus based upon a conception of the person as having no determinate independent character of will. If we judge that it is more plausible to view persons as having higher order interests and preferred ends, then we must reject the argument for the Bayesian criterion (151–152).
In the absence of further argument, then, the balance of reasons would appear to weigh against the conclusion that the Bayesian criterion is the appropriate criterion to guide choice under the extreme uncertainty of the original position. Rather, it seems more plausible to argue with Rawls that rational choosers under conditions of extreme uncertainty who (i) are more concerned to secure a satisfactory minimum than to secure the possibility of receiving greater advantages and (ii) insist upon ruling out completely certain unacceptable outcomes (iii) should therefore be guided in their choices by the maximin rule.
4. CONCLUSION
In interpreting Rawls's arguments for the principles of justice, it is essential to keep in mind the limited role of the maximin rule in Rawls's formal argument. In developing this argument, Rawls represents – in the form of his decision procedure – ‘weak and widely shared’ considered judgements regarding the qualities that should characterize fair conditions for resolving questions of justice. The character of the model generated through this constructivist process, Rawls argues, reveals important qualities of the process of judgement in questions of justice. The decision problem of moral deliberators faced with such questions, Rawls concludes, has the formal structure of a problem of choice under uncertainty, and the specific character of this particular choice problem justifies the judgement that a particular decision rule – the maximin rule – is uniquely appropriate to regulate the judgements of the choosers. The maximin rule recommends the choice of the most satisfactory minimum alternative and – since the task of the choosers is to select a set of principles of justice that constitute an acceptable conception of justice – the maximin rule in this case recommends the selection of those principles that, taken together, constitute the most satisfactory minimum conception of justice.
This sketch of the choice process in the original position describes the full extent of the operation of the maximin rule in Rawls's argument. The maximin rule neither justifies nor describes the operations of the difference principle, nor does it apply directly to any substantive questions of justice or morality. It certainly does not require that a just society must maximize the well-being or share of goods of the least advantaged. Rather, the maximin rule is simply a rule for choice that is appropriate, Rawls argues, to guide reflections regarding the choice of principles of justice under the particular conditions that characterize the original position.
While the role of the maximin rule in Rawls's argument is thus more limited than has often been appreciated, the rule – as the choosers apply it under the special circumstances of the original position – nevertheless performs a number of important functions in Rawls's argument. In particular, Rawls argues, the process of judgement involved in applying the rule under the specialized social contract conditions of the original position is well-designed to focus the choosers on identifying and protecting their real interests. In addition, the choosers’ complete lack of information regarding the social position that any one of them may occupy after the veil is lifted justifies the choosers’ rational judgement that satisfactory protections of fundamental interests must be provided to all members of society. Application of the maximin rule under conditions of uncertainty, Rawls argues, ensures that even mutually disinterested rational choosers must select principles that require that social institutions are to be constituted in a manner than abstains ‘from the exploitation of the contingencies of nature and social circumstance’ (156). Thus, application of the maximin rule leads to the adoption of principles that can be seen as ‘a fair undertaking between the citizens as free and equal with respect to these inevitable contingencies’ (Rawls Reference Rawls and Kelly2001: 124).
If it is assumed that (i) Rawls's theory is designed simply to translate the requirements of maximizing rationality into moral principles; and (ii) the central guarantee of Rawls's theory is the difference principle, then the theory will appear to focus primarily on the redistribution of income or resources according to a rigid formula. Once these confusions have been corrected, the theory can be brought into sharper focus. In protecting basic liberties, guaranteeing equal opportunity and a fair distribution of social goods, and establishing the proper priority relation between these interests, Rawls argues, the theory ensures the realization of a ‘satisfactory political and social world’ (Rawls Reference Rawls and Kelly2001: 100) – a world in which basic liberties are respected, all persons are guaranteed the opportunity to succeed to the extent permitted by their talents, and the society's public conception of justice affirms the value of each person as a moral being and not merely as a productive factor in the market. The fact that the satisfactory minimum constitutes a satisfactory form of social life, and not merely a bundle of goods, Rawls asserts, ‘is crucial for the argument’ (Rawls Reference Rawls and Kelly2001: 100). It is this satisfactory form of social life to which Rawls refers when he suggests that it will not be worthwhile to jeopardize the minimum guarantee for the possibility of further gains.
The principal critiques that I have assessed and rejected have played an important role in shifting attention away from some of Rawls's central arguments in A Theory of Justice and have obscured the function and significance of the maximin argument in Rawls's justification of the two principles of justice. Addressing the confusions in the interpretation of Rawls's maximin argument, then, should make possible a more balanced appraisal of Rawls's overall argument.
