|Faust complained about having two souls in his breast, but I harbor a whole crowd of them and they quarrel. It is like being in a republic.
Otto von Bismarck1
Individual human beings – their choices, agency, rights and reason – are the basic units of analysis in most modern philosophy and social science. From the rational agent model to the moral individualism, the individual has become something of a mascot for the modern Anglo-American academy. Debate over the nature and proper understanding of the individual and her relationship to political community has been a dominant force in political philosophy over the last quarter century. The first three decades of post-war political theory primarily focused on providing an analytic account of liberal, non-utilitarian Western political discourse, culminating in the seminal egalitarian and libertarian theses of John Rawls and Robert Nozick, respectively.2 The liberal thesis, characterized by Allen Buchanan as the view that “the state is to enforce the basic individual civil and political rights”, came under vigorous attack from a number of communitarian critiques during the 1980’s.3 Several thinkers found fault with liberal theory along various contrasting avenues. Michael Sandel targeted the flawed and internally inconsistent metaphysical conception of the self that he claimed is fundamental to the liberal, and particularly Rawlsian, conception of justice.4 Alisdair MacIntyre alleged that liberalism lacks a coherent moral underpinning, while Michael Walzer took issue with what he saw as the reductive, unitary liberal conception of justice.5 In response, liberals sought firmer and more broadly appealing foundations for their core arguments as well as means of synthesizing into liberal political theory important ideas suggested by the communitarian critics.6 As issues of multiculturalism and the appropriate treatment of minorities have played an increasingly prominent role in modern political discourse, liberals and their critics have worked to develop accounts of notions like “group rights” and “ethno-linguistic minority protections” consistent with their broader political vision.7
One current running through these debates, usually taking the form of an implicit assumption, is the primary status of individual rather than collective agency. Liberals, communitarians and their synthetic progeny often disagree about the nature of the good for the individual, whether collective good is a sensible notion, the importance of the community to individual identity and well-being and many other matters. However, they seem to implicitly agree that individuals, but not groups, can sensibly be thought of as unitary agents or selves.8 Beginning with Rawlsian concept of individual choice behind a veil of ignorance as the basis of just political order, it is clear that most liberal theory assumes (even if it need not) this view, which I will refer to as the individualism of agency. Perhaps more surprisingly, however, communitarian MacIntyre also implicitly affirms the individualism of agency when he writes that “the possession and exercise” of virtue “tends to enable us to achieve…goods.” Note that he never refers to groups of individuals as possessing or exercising collective traits or being enabled to take actions or achieve goods.9 Nor do the defenders of liberalism generally see communitarianism as an attack on the individualism of agency. For example, in offering what he sees as a communitarian defense of liberalism, Buchanan writes that liberal rights “allow individuals to partake of the… good of community…by giving individuals freedom to unite with like-minded others to create new communities.”10
Despite the prevalence of this maintained hypothesis, it clashes with prominent theoretical arguments and an emerging body of evidence from science which I develop below. Important decisions that individuals make are sensitive to small changes in the framing of the choice. Different parts of individuals are often in conflict with each other. Individual lives and decisions may therefore be fraught with the same paradoxes that undermine group agency. It is far from clear that individual agency is on qualitatively better footing than collective agency.
While the individualism of agency is commonly assumed in liberal, as well as non-liberal, political theory, I do not want to claim that this is necessary or fundamental to liberalism. While I suspect my critique of the individualism of agency may have broader implications for liberal theory, exploring the nature of these is beyond my ambitions here. Instead this paper has two principle goals. The first is to draw on a wide range of arguments and evidence to add force to previous philosophical critiques of individual agency and thereby show the weakness of distinctions commonly assumed to exist between the plausibility of individuals and groups as agents. The second is to note the potential destructive implications of this thesis for a rather narrow liberal view, which I call rights individualism: individuals, but categorically not groups, may be reasonably thought to have fundamental, ethically irreducible and inviolable political claims to rights. Rights in this very strong sense seem closely tied to agency: it would seems strange if the fundamental units of political currency were be protections of choice if choice were not of fundamental import. However, there may be other possible justifications for rights individualism that do not rely on the individualism of agency. My purpose here is simply to argue that the agency-based argument, more plausibly integral to rights individualism than to liberalism more broadly, is invalid.
I critique two types of arguments for the individualism of agency. The first type of argument is philosophical; for concreteness I address a case put forward by Chandran Kukathas in his essay “Are There Any Cultural Rights?”, which begins with a defense of rights individualism through the individualism of agency. Kukathas’s broad case is composed of three related arguments covering the relationship between groups and the political system, temporal problems of groups and challenges arising from divisions within groups. The second class of arguments is related to the first, but is phrased in the language of rational choice theory in social science. Again for specificity, I focus on Arrow’s classic Impossibility Theorem demonstrating the general incoherence of group decisions, though I briefly discuss other formal critiques of group agency.11 I use a variety of theory and evidence to demonstrate the ways in which these arguments against group agency can be brought to bear on individual agency and therefore demonstrating the qualitative symmetry between groups and individuals that contradicts the individualism of agency. While a critique of Kukathas and an analysis of Arrow’s arguments in relation to the individual organize the paper, my target is the doctrines of individualism of agency and rights individualism, rather than the (possibly evolving) views of particular authors.
Following this introduction I first discuss Kukathas’s arguments in somewhat greater detail. I then provide a non-technical overview of Arrow’s Impossibility Theorem and the challenge it poses to group agency. I invoke a classic argument by Kenneth May to apply Arrow’s result to internal divisions within an individual. Recent evidence from several disciplines and further theoretical arguments make such an application more plausible than at the time of May’s writing.12 I use this evidence, along with a variety of philosophical and formal arguments, to rebut each of Kukathas’s points and provide a number of examples of the symmetries between the problems of individual and group agency. I conclude by indulging in a bit of speculation about possible broader implications of my argument for political theory.
Wrong for Rights
The central pillar on which Kukathas rests his case is the inherent dependence of groups on the political and social system. “Groups or cultural communities,” Kukathas writes, “do not exist prior to or independently of legal and political institutions but are themselves given shape by those institutions.”13 Even if groups happen to be constituted chronologically earlier than the political systems that support them, this is a historical accident and “does not confer on a community the right to continued existence” because “[such l]egal rights can themselves be important determinants [in the creation, structure, or substances of cultural groups].” Granting fundamental political rights to groups which are themselves constituted on the basis of the political system would lead to circular and indeterminate political reasoning at best and invite a paralyzed and arbitrary political philosophy. Kukathas summarizes, “It is not acceptable to evaluate or choose political institutions or to establish legal rights on the basis of the claims or interests of cultural communities because those very institutions or rights will profoundly affect the kinds of cultural communities individuals decide to perpetuate or form.” 14
Supporting this point, Kukathas argues that groups lack the important quality of temporal constancy and reliability that we require of agents if they are to form the foundation of a political philosophy. “Groups are constantly forming and dissolving,” Kukathas argues, “There is no more reason to see [their] particular interests as fixed than there is to see particular political arrangements as immutable.”15 Groups, being inherently hard to pin down or even characterize, do not seem sound agents to constitute a polity.
The divisions that inherently exist within groups offer Kukathas’s last reason to avoid collective rights. “Within cultural communities,” he argues, “there may be important differences of interest.” Particularly troublesome are divisions that exist between the masses within a group and the group’s elite.16 Granting privileges and rights to the group as a whole may empower the elite to harm the interests of many members. “Minorities within a cultural community which might over time have formed quite different coalitions with other interests may find that their interests are to a significant degree subject to control by the larger rights-bearing community.”17 Granting rights only to cultural groups when important alliances and organizations may form between minorities that are part of various rights-bearing groups seems arbitrary and illiberal. “From a liberal point of view, the divided nature of cultural communities strengthens the case for not thinking in terms of cultural rights.”18
In summary, Kukathas views groups as essentially like electoral majorities: ephemeral, comprising complex internal structure, inconsistent over time and not inherently worthy of political protection. 19 This analogy is apt, given that the field of social choice theory in economics and political science has extensively addressed the problems of group agency that arises from the inherent flaws of voting systems. The seminal result in this field, and the one I focus on primarily below, is Arrow’s Impossibility Theorem. However, it is important to note that I focus on Arrow’s result not because it is the definitive or even most important such formal critique. In fact, the Gibbard-Satterthwaite Theorem, which I mention briefly below, shows that under weaker conditions if voters are strategic, the limits on group agency are severe.20 More broadly, game theory has taught us that strategic interactions among individuals do not usually result in outcomes that can plausibly be thought of as in the interest of the group.21 Many of the critiques of individual agency I will put forward below depend on these more general forms of internal conflict. Nonetheless I will extensively discuss Arrow’s result here because of its prominence as a critique of group agency, because of its elegance and because it offers a particularly clear structure for analyzing the symmetries between group and individual agency.
In his classic 1951 dissertation Social Choice and Individual Values, Kenneth Arrow took formal aim at group agency. He demonstrated mathematically that group decision-making is an inherently troubled process. In particular, he showed that it is impossible to devise a means of aggregating individual preferences into group decisions which are both consistent and satisfy basic axioms of social choice. By “consistent”, Arrow meant that these choices obey the most basic principles of rational choice: the group must have some preference between any two possible courses of action, X and Y (prefer X, prefer Y, or be indifferent) and such preferences must obey transitivity (if the group prefers X to Y and Y to Z, it must prefer X to Z).22 Arrow’s other axioms are equally intuitive and are fundamental to any meaningful definition of group decisions:
Non-dictatorship: No individual should act as a dictator, unilaterally determining the group’s preferences.
Independence of Irrelevant Alternatives: In deciding whether the group should prefer X or Y, only individuals’ rankings of X against Y should matter.
Unanimity: If all individuals prefer X to Y then the group should.
Universal Domain: The method of social choice must be defined (must give some answer) for any arrangement of individual preferences.
Additionally, Arrow’s theorem requires two basic assumptions about the environment being considered, which are easily satisfied in most situations of social choice:
Ordinality: Preferences are known in determining social preference only as rankings of alternatives, not as cardinal numbers or degrees of preference. Non-ordinal social choice systems would require knowing the degree of individuals’ preferences, which they would always try to exaggerate. This would also make comparisons of “degree of preference” across individuals necessary, thereby potentially valuing some individuals more than others in voting. Each of these is sufficiently unpalatable or impractical, Arrow argued, so as to rule it out.
Richness of Setting: The social choice rule continues to work if there are more than two alternatives and more than two individuals.
Thus, Arrow showed that, in most reasonable settings, the concept of collective choice cannot be well-defined.23 As Arrow put it in his classic statement:
[T]he only methods of passing from individual tastes to social preferences which will be satisfactory (in the sense defined above) and which will be defined for a wide range of sets of individual orderings are either imposed or dictatorial.24
Any non-pathological method of group decision-making will be susceptible to incoherence or will only be applicable to a limited set of social choices. Even if we are willing to restrict the domain of situations (arrangements of preferences) in which the group may make a decision, this is simply a hack. Presented with other situations (if individuals change their preferences) the group will be unable to choose or will choose in an inconsistent or pathological manner. Thus granting cultural groups the right to self-determination, or anything else, seems deeply misguided. Because Arrow’s critique fundamentally undermines group agency, it also undermines the capacity for groups for groups to bear rights, at least of one who views rights as closely tied to agency.
While Arrow’s theorem may seem the ultimate watertight critique of group agency, a closer examination of the result, first suggested by Kenneth May in his 1953 essay “Intransitivity, Utility and the Aggregation of Preference Patterns”, shows that it may pose equally serious challenges to individual agency. These problems have roots in the conception of the individual that Arrow and classical economic theory more generally, assumes. This “rational agent” suffers none of the problems of inconsistency that beset groups. She can easily rank all possible social outcomes in expressing her preferences and these rankings constitute the basis of consistent choices. What is so surprising about Arrow’s theorem, in fact, is that the problem of making a group decision can cause such chaos when the individuals that compose the group are perfectly rational.
Such an idealization of individuals, however, is poorly instantiated by humans. An extensive psychological literature inspired by the path-breaking work of Amos Tversky and Daniel Kahneman has catalogued hundreds of situations in which individuals act in ways that clearly violate the most basic principles of rational choice.25 In situations where many conflicting and qualitatively different attributes are important in making a decision, when uncertainty plays a role or when the consequences of a choice can be described relative to many starting or “reference” points, these framing effects are particularly pronounced.26 Thus, individuals often exhibit the same sorts of inconsistencies that plague group decisions. Their status as rights-bearing agents may, therefore, be subject to the same criticism.
Nonetheless, one might argue that whereas groups are inherently given to inconsistency and incoherence, individuals have the capacity for rational choice, which they can and should exercise, even if they sometimes fail to do so. The view that the inconsistency we observe in individuals is, in Tversky and Kahneman’s words, “normatively distasteful” and essentially a form of confusion on the part of actors is one widely held within the psychological community.27 Inconsistent decisions are described as “biases” or as the consequences of “heuristics”. Incoherence in decision-making is often considered intimately related to errors in positive judgments.28 In the realm of rights, then, it would seem that individuals may fail to exercise rights meaningfully, but, unlike groups, there is nothing preventing their doing so. It still appears reasonable that individuals, and not groups, should hold rights.
This limited “judgmental psychology” view of inconsistency tells only part of the story, however. Individuals do not deviate from being rational agents merely through small inconsistencies or mistakes. More than merely being inconsistent, individuals do not seem nearly as unified or “whole” as the rational agent model posits. In fact, recent research in economics, neuroscience and philosophy paints a picture of the individual which more closely resembles the troubled collective in Arrow’s problem than it does the rational individuals that compose it. And as May pointed out, individuals with multiple incomparable desires, or who are divided into several selves, may just as easily as groups suffer from Arrow’s conundrum. In fact, economists have made important strides in explaining a variety of empirical phenomena by considering the conflict, cooperation and game playing of various “selves” within an individual.29 Psychology, dating back to Plato, has considered theories of the individual as divided into many relatively independent and sometimes conflicting parts.30 A number of philosophers have investigated metaphysical and some ethical issues relating to the problem of multiple selves and internal conflict.31 The idea that many areas of the brain act in an environment of relative independence to one another in evaluating choices is an important finding of modern cognitive neuroscience.32 Internal conflict in non-human animals, leading to indecision and intransitive preferences, has been observed by biologists and animal psychologists since the 1940’s.33
Unifying several of these trends, Adi Livnat and Nicholas Pippenger have shown that this indecision in mice can be parsimonious explained as the result of a computationally optimal division of labor within the brain. 34 Furthermore, they prove that in such a computationally optimal arrangement, the internal agents (selves) can be selfish, engaged in explicit (game theoretic) conflict and only have ordinal preferences available for higher level aggregation. 35 This work not only provides strong evidence confirming the multiple selves hypothesis; it also places this hypothesis on extremely firm theoretical foundations, leading us to believe that, from an evolutionary and computational perspective, we should expect multiple, conflicting selves.
Livnat and Pippenger’s work also suggests a somewhat different interpretation of the notion of multiple selves. We need not see potentially conflicting preferences as separate agents or selves within an individual. Instead, as May originally argued, they might represent independent, distinct criteria for evaluating actions that are not commensurable with one another. For example, suppose that an individual cares about various things in life: she values her health, her relationships with others, her material well-being, etc. And suppose that she can (at least in theory), like a rational agent, rank every conceivable outcome according to how well it satisfies these desires (as I will refer to them henceforth). 36 If the individual considers several relatively independent (or independently computed) attributes there is no guarantee that these various desires can be brought into some common currency of “utility” that allows the expression of a consistent preference.
If we are to take this idea seriously, we must substantially alter our normative and positive expectations for individuals. Even if each of the selves that make up an individual is rational and has coherent preferences, a person attempting to make decisions may fall prey to the same problems of social choice that Arrow details in the context of group decision-making. That is, we should not expect an individual to be able to aggregate the preferences of multiple selves fundamentally better than groups are able to aggregate the preferences of the individuals that they comprise.37
It is important, however, to remember that Arrow’s theorem is not an immediate or universal result. It follows only if one requires certain axioms of choice, as discussed above. These axioms may or may not be fully compelling in the context of social choice; this has been a long topic of debate in the economic and philosophic communities.38 What matters for the current discussion is whether these axioms are qualitatively and fundamentally more applicable to a context of social than one of individual choice. In order to answer this question, we must consider each axiom listed above in turn. 39
Dictatorship in the context of individual decision-making translates into the lexical priority of one desire over others or, under the multiple selves interpretation, the dominance of one self’s preferences in all contexts over the desires of other selves. When conceptualized in this second manner, dictatorship does not initially seem unreasonable. One of the most common framings of multiple selves, emphasized in various forms by thinkers from Plato and Freud to modern behavioral economists, is the division between passions and reason. Some psychologists and economists who consider individuals divided among passion and reason not only believe that it would be best if reason could act as a dictator; many in fact advocate policies designed to facilitate this dominance. 40 While useful in some contexts, this resolution is far from satisfactory, even in this narrow view of internal division. Individuals with brain damage in “passionate” regions of the brain can be far more debilitated than those suffering legions in the most “rational” regions, suggesting that simple dominance of reason over passion is unlikely a template for a flourishing life.41 On a more casual level, I suspect most people would find a life in which reason always triumphed over passion barren and boring.
Even if one were to accept dictatorship in such a limited context, there are many reasons, however, to believe that intrapersonal divisions are much deeper and broader than this view suggests. Modern neuroscience and neuroeconomics supports the view, held by thinkers including Plato and Freud, that the brain is divided into more than two separate functional units.42 From the perspective of Livnat and Pippenger’s work mentioned above, this is precisely what should be expected: the computationally optimal decision algorithm is unlikely to involve merely one split in processing; at least some processing units into which computation on a complex task is split will themselves find it optimal to make another split.43 Each of these separate systems is integral to the process of making good decisions; otherwise, they would not be part of the computationally optimal solution Livnat and Pippenger (and, presumably at least in approximation, evolution) compute. When each conflicting self provides an important input to optimal decision-making, the axiom of non-dictatorship becomes immediately compelling. If we accept dictatorship of one desire within the individual, why should we be troubled by the imposed or dictatorial social choices that Arrow rules out? If there is some easily calculable overriding moral principle by which individuals should make all decisions and to which all other desires should be subservient, then might we not simply say the same about groups? Perhaps groups, too, should follow some consistent ethical principle, but fail to; if so why should this fact detract from their agency any more than in the case of an individual?
Despite the preceding arguments, it still seems that the unity of the human body offers an important basis for differentiating individual from group choice. Individuals always must make a single, unified choice; groups often may choose not to choose and allow a particular matter to be handled by its members independently. Yet the necessity of unified action does not logically imply the reasonableness of assuming unified decision-making. To see this sharply, consider the case of Siamese twins. 44 The two individuals that share a body clearly may have different preferences. Clearly they may also have to coordinate their decision-making, either because they directly share control over some organ or because any effective action requires them acting in concert. It seems unjust, impractical and implausible that one of the two twins would ever act as a dictator over the other. The fact that they will be paralyzed during any conflict does not imply that such conflicts are impossible nor does it imply that it would be reasonable that one individual’s preferences should dictate the actions of the pair. By analogy, the simple fact that divisions among preference orderings held by different selves or contained in different desires reside within the same brain does not imply that it is reasonable to think of these desires as unified or to think of one as being an appropriate or actual dictator over the others.
A further challenge to the dictatorship axiom (and the notion of internal conflict more broadly) in the context of individual choice is the notion that individuals may possess more effective, non-coercive means than groups for confronting the problem of disunity by exerting effort in order to bring various parts of the self into agreement. It may very well be the case that individuals more often have effective means of achieving such internal concertation than do groups. However, it is certainly not the case that groups entirely lack such mechanisms, nor is it the case that the mechanisms within individuals are uniformly more efficacious than those within groups. At least since the time of the Greeks straight through to Barack Obama’s presidential campaign, political rhetoric has played a crucial and effective role in achieving consensus around directions of social action.45 The faith-based calls of religion, as well as secular ideologies, to self-sacrifice have also proven powerful forces for unifying individuals with seemingly disparate interests around common cause. In fact a major theme of modern social psychology is the study of the various mechanism, such as conformity, obedience and social learning, through which groups achieve unity of, unfortunately often sinister, purpose.46 Recent work in economics has similarly emphasized the tendency of individuals in groups to sacrifice their independent information processing and choice-making capacity in favor of various forms of group-think.47 More broadly herd-like group behavior, mobs, nationalism, group-think and other such mechanisms of concerting collective action have been important themes in literature and popular discourse at least since the enlightenment.48 Thus it would be difficult to argue that groups entirely lack the means individuals have of achieving consensual internal unity through deliberate effort.
Furthermore, it is not the case that individuals are uniformly more effective at exerting effort to peacefully overcome internal conflict than are groups. Literature has long featured characters from, Hamlet to Scrooge, whose internal struggles provide the dominant drama. Casual observations reinforces this theme: most of us know someone whose struggles to achieve internal cohesion when confronting troubles with drug abuse, career choice or romance can, even when heroic efforts are made, outlast and eclipse their inter-personal conflicts. Post-war debates about stimulating savings tended to focus on capital taxation or social security. However, recent evidence suggests policies aimed at resolving internal conflict, such as adjusting default settings on retirement plans or adjusting individuals’ ability to prohibit themselves from withdrawing from an account, rather than large-scale social policies, may be most effective.49 None of this argues that individuals are usually or on average more divided than groups. This question is clearly empirical in nature and substantial effort would be needed even to pose it coherently. But it does demonstrate that it is difficult to draw a qualitative distinction between individuals and groups in their capacity to achieve unification through purposeful effort. After all, given that difficulty internal motives have in effectually expressing their dissent, the importance to a dominant self of cohesion within an individual at any particular time may be less than that for a group. Thus this final objection to the dictatorship axiom is not categorically stronger in its applications to individuals than groups.
The axiom that is perhaps most deceptive is independence of irrelevant alternatives (IIA), which requires that changes to the preferences of selves or desires over possibilities other than X or Y from do not affect the unified individual ranking of X against Y. One potential objection to IIA in the context of individual decision-making is that the various desires or selves may not constitute merely ordinal preferences; rather there may be some extent of cardinality or “degree” to these preferences.50 If such degrees of preference are fixed and independent of the set of alternatives offered, then an aggregation mechanism which takes these degrees of preference into account need not violate IIA. However, this is fundamentally an objection to the notion that only ordinal preferences of desires or selves are available.
There are a few reasons to believe this is the case, many of them hinted at above. If we believe that selves represent separate neural processes that each output incomparable (ordinal) preferences, then ordinality is an immediate consequence. Furthermore, ordinality is even more crucial if selves are strategic, as they would of course have an incentive to inflate their degree of preference to obtain their desired outcome. If we instead take the “desire” interpretation, then ordinality simply requires that the desires not be fungible or directly commensurable. This, as I argued above seems quite reasonable for many desires. For example, it seems quite intuitive that social relationships are not commensurable with money or health. If desires are construed too narrowly, ordinality certainly will not hold: desires to achieve a high mark in mathematics are likely commensurable with desires to do well in English. But if a broad enough perspective is taken, it seems very plausible, as well as computationally efficient, that there will be multiple, incommensurable desires or selves.
Accepting ordinality, the case for IIA becomes much stronger. Because we have excluded any notion of degrees of preference, violation of IIA can be seen as a reversal of aggregated preference on the basis of the addition or subtraction of some item from the choice set. On the face of it, that behavior seems deeply strange and inconsistent with rational choice.51 We might, however, attempt to resurrect the notion of degrees of preference in a purely ordinal context by violating IIA. For example, we might say that if X is globally the best (among all options in the choice set) for self one and Y globally worst, but for all other selves Y is second worst and X is absolute worst we should choose X over Y; however in the case that X and Y are the only two options, then we should simply take majority vote. Such a preference aggregation rule seems reasonable on the face of it, but fundamentally appeals to an intuition about cardinal preferences which cannot apply when preferences are only ordinal. Note that one self or individual could cause individual or society to choose X over Y by adding to the choice set a number of alternatives that are better than either X or Y for all other selves and are worse than X for herself. Even if these alternatives are never chosen globally, their mere addition to the global menu of choices may cause reversal in preference. Such an outcome seems deeply strange, manipulable and open to pervasive framing effects. In fact, as Arrow hypothesized and Gibbard and Satterthwaite proved, there is a close connection between a social choice rule satisfying IIA and that rule being robust to strategic manipulation, in a game theoretic sense.52 So long as the selves within an individual have at least some strategic sophistication, as economic evidence indicates they often do, IIA is a necessity for choice rule.
These problems do not seem much less acute in the case of aggregating the preferences of selves within an individual than they do in aggregating preferences across individuals. While one may reasonably doubt the strategic sophistication of internal motives or selves, it may at the same time be easier for one self to manipulate a choice set or change the framing of a decision for the individual in which she resides than for an individual to shape a large social group’s choices or twist the way in which choices are framed on such a large scale. Thus, despite its subtlety, it does not seem that IIA is significantly more applicable to groups than individuals.
Unanimity seems immediately compelling and does not merit much analysis. Would it be reasonable for an individual to make a choice which every imaginable desire, consideration or self ranked as inferior to another choice?
The last of Arrow’s axioms is equally immediate in its appeal. Universal domain states that we cannot, a priori, rule out any ranking over the set of actions by various selves. While this appears a stringent condition, it is actually very intuitive: we cannot ahead of time say anything in particular about the alignment of preferences of various selves or desires with respect to one another.53 It may seem likely, for example, that things that make one healthy will also promote good social relationships. Clearly, however, there will be cases when these conflict and cases when they disagree on preferences between certain actions but agree on other trade-offs. The primary requirement of universal domain, it turns out, is that it is not always the case that there is some one-dimension spectrum on which the actions can be placed that causes every preference to have a single peak along this spectrum.54 Individual choices are at least as complex and variegated as are group choices. It therefore is fleetingly unlikely that sub-individual preferences would be single-peaked while individual preferences are not. So long as intrapersonal preference variation is, at least in some cases, as great as interpersonal preference variation is, universal domain must apply to the aggregation of selves’ preferences as much as it is to aggregating individual preferences. Furthermore, in the Gibbard-Satterthwaite result, this can be further weakened to the even less objectionable assumption that for all pairs of alternatives X and Y, we cannot rule out that there is some preference under which X is preferred to Y and another where Y is preferred to X.55 Most importantly, there seems no immediate reason why preferences between individuals should have a richer relationship to one another than various desires within individuals. In short, universal domain seems just as reasonable in the case of individuals as in the case of groups.
If we are now satisfied that Arrow’s axioms are reasonable in the context of aggregating various desires or the preferences of various selves, then Arrow’s theorem implies that there is no coherent way to take these various things that they value and join them into an overall preference. Since its birth, modern formal economics has simply maintained this counter-intuitive assumption that various desires are fungible with one another. When considering a complex problem where many different, independent desires are important, it is often hard to find a means of trading these various considerations off. Only economists find it a natural concept that years of life are commensurable, through some utility function, with dollars and cents. Arrow’s theorem demonstrates decisively that this unstated assumption of fungiblity in economics, long criticized by anthropologists and sociologists, is far from innocuous.56
Nor does the mere fact that an individual makes some decision imply that this action represents a choice in the sense of embodying a coherent preference. If an individual makes intransitive choices, there is no sense in which these “choices” reflect anything about what the individual would choose or prefer among the options involved. These choices no more reflect meaningful individual agency than the oft-derided actions of an alleged group representative can be thought to report the preferences of a group. If we permit intransitivity in individual choice, then Arrow’s critique of collective agency wilts. Even when an individual is simply incapable of making a choice, we will usually observe this as a choice in favor of the status quo. In fact, I provide much evidence below that this sort of tendency to choose default options is strongly prevalent in individual choice, just as Arrow argued it was in social choice. 57 While some bounds on an individual’s preferences can be recovered by decomposing that individual into her conflicting desires or selves, such a procedure could as easily be applied to problematic group choice.58 Therefore, we cannot use Arrow’s theorem to argue against group agency while defending that of individuals.
The Livnat and Pippenger computational argument for multiple selves suggests an important parallel between multiple selves within an individual and individuals within a group. Many, if not most, groups are constituted to advance some measure of the interests or welfare of their constituent members. Democratic procedures within groups can then often be viewed as a delegation to members of parts of the computation of the group’s welfare. Each individual is likely to be expert in understanding their own welfare. Since each individual’s welfare composes part of the group’s welfare, allowing individuals to draw conclusions about their own good and then utilizing democracy to aggregate these judgments is a sensible way to optimize group well-being. In fact, Buchanan offers a similar “communitarian” justification for liberal rights.59 Of course, Arrow’s theorem teaches us that such separately computed preferences may not lead to any coherent social judgment, but it does not imply that such delegation is a bad heuristic for maximizing group welfare. In the same way, an individual may delegate the computation of health, social, material and other costs and benefits of decisions to separate internal systems, as Livnat and Pippenger argue is computationally optimal, and then seek to aggregate these independent preferences. The fact that group welfare is computed by a messy process of aggregating potentially irreconcilable individual preferences should not undermine the agency of groups any more than the same problems undermine individual agency.
Of course, as discussed above, Arrow’s theorem was only one of several formal critiques of group agency. Nonetheless all of these critiques have the structure that the interactions between interests of various agents (or desires) should not be expect to result in a sensible aggregate decision process. Thus the crucial point is that once one accepts a divided self or a self with many competing desires (almost) any formal critique of group agency is likely to also undermine individual agency. I now turn back to the arguments made by Kukathas that I laid out above and show that the considerations that undermine formal justifications of the individualism of agency run like an erosive river through the terrain on which Kukathas’s case is built.