Hobbes' Problem

+REEHV૷3UREOHP 3HWHU6WRQH The Good Society, Volume 24, Number 1, 2015, pp. 1-14 (Article) 3XEOLVKHGE\3HQQ6WDWH8QLYHUVLW\3UHVV For additional information about this article http://muse.jhu.edu/journals/gso/summary/v024/24.1.stone.html Access provided by Trinity College Dublin (21 Sep 2015 11:59 GMT) Hobbes’ Problem pete r stone Abstract his paper considers the contribution positive political theory has made to the resolution of “Hobbes’ Problem”—the problem of describing the nature of artiicial persons, such as states. his problem plays a central role in both empirical and normative work in political science. Several areas of positive political theory, most notably social choice theory, have shed light on Hobbes’ Problem, and the results have been generally negative. hat is to say, they have suggested that the conditions necessary to generate an artiicial collective agent out of the individual agents comprising it may be impossible to satisfy, at least if this is done in a nontrivial way. While eforts to establish the possibility of a collective agent continue—notably in recent work by Christian List and Philip Pettit—the primary accomplishment of positive political theory has been to raise the question of whether Hobbes’ Problem can be solved at all. Keywords: Hobbes, positive political theory, social choice theory, Kenneth Arrow, William Riker In this paper, I will discuss the relationship between positive political theory and one of the central problems in modern political thought—what THE GOOD SOCIETY , vol. 24, no. 1, 2015 Copyright © 2015 he Pennsylvania State University, University Park, PA 2 | THE GOOD SOCIETY | vol. 24, no. 1 I will call “Hobbes’ Problem.” Hobbes lays out this problem in the opening passage to the Leviathan: NATURE (the Art whereby God hath made and governes the world) is by the Art of man, as in many other things, so in this also imitated, that it can make an Artiicial Animal. For seeing life is but a motion of Limbs, the beginning whereof is in some principall part within; why may we not say, that all Automata (Engines that move themselves by springs and wheeles as doth a watch) have an artiiciall life?. . . . Art goes yet further, imitating that Rationall and most excellent worke of Nature, Man. For by Art is created that great LEVIATHAN called a COMMON-WEALTH, or STATE (in latine CIVITAS), which is but an Artiiciall Man; though of greater stature and strength than the Naturall, for whose protection and Hobbes took it for defence it was intended. granted that states were, in some important sense, artificial persons . . . The past seven decades of work in positive political theory have called into question this assumption . . . “To describe the Nature of this Artiiciall man” is the primary task Hobbes sets for himself in Leviathan. In this paper, I will ofer a progress report on this project, with a special eye to the contribution made by positive political theory.1 he irst thing to note about Hobbes’ Problem is that it bears on both empirical and normative work in political science. Put another way, it is a problem with an explanatory and an evaluative dimension. Hobbes wanted to know what made states possible and how states work. But he also wanted to know what obligations people have to the state, and why. he answer to both questions, he thought, could be discovered by thinking of the state as an artiicial man. One can generalize from the problem as Hobbes deined it to the more general problem of collective persons composed of individual natural persons. Any such generalization will be relevant to both empirical and normative work. If there are such things as collective persons in the world, then understanding how the world works is going to involve understanding their behavior. And if such collective persons exist, then their moral status will surely be important to any normative theory of politics. Vox populi, vox dei has long been the slogan for the strong democrat. We need not go that far to believe that if a collective person really has a voice, we should care what it has to say, and that the collective agent’s voice should count, morally speaking, as more than just a set of individual voices. And so sorting out Hobbes’ stone | Hobbes’ Problem | 3 Problem—the problem of collective personhood—is critically important for the entire ield of political science. Of course, all of this assumes that collective persons exist. Hobbes took it for granted that states were, in some important sense, artiicial persons—that they were close enough, in terms of their behavior or their moral status or both, to natural persons that it made sense to classify them together. he past seven decades of work in positive political theory has called this assumption into question, a topic I will address in my conclusion. his result may be the most philosophically important one generated by this line of research. If collective persons exist, then they must somehow be similar to natural persons. For Hobbes, the critical property shared by all persons was rational agency. A person is a being that (normally) decides how to act on the basis of reasons. And so if collective persons exist, one must be able to attribute to them actions of this sort. And they must be rational actions, at least in the typical case; if one believes that an agent normally acts in ways contrary to the reasons available to her, one must question whether those reasons really are available to the agent—or, indeed, whether the agent is really an agent at all.2 Hobbes’ Problem thus becomes the problem of specifying how a rational collective agent works. his means identifying actions on the basis of reasons that can be attributed to that agent. For Hobbes, those actions all stemmed from the sovereign of the state. he sovereign issued commands, and the subjects obeyed. he voice of the sovereign simply was the voice of the state, and the reasons of the sovereign were the reasons of this artiicial person. he only problem let to solve, or so Hobbes thought, was explaining how the sovereign got to be the sovereign, given the natural equality of human beings. Hobbes tried to accomplish this in terms of the authorization granted the sovereign by his subjects. his is all well and good if the sovereign is a single individual. But Hobbes did not require this to be the case. he sovereign of a state could be a group; it could even be the entire population of the state (popular sovereignty). Hobbes preferred one-person rule on pragmatic grounds, and so throughout the Leviathan he usually takes this form of sovereignty for granted (Hobbes 1968, ch. 19). But if the sovereign is comprised of a group of individuals, then one must explain how these individuals, each with her own reasons for action, somehow generate a single set of actions based upon a single set of reasons. Without this, there is no collective agent. Hobbes’ preference for one-person rule caused him to discount the problem of deriving collective agency from a group of people rather than a single person. For purposes of contemporary political science, however, this problem is critical, and not simply because autocracy is out of fashion 4 | THE GOOD SOCIETY | vol. 24, no. 1 these days. If the voice of the collective agent can simply be the voice of an individual person, then Hobbes’ Problem is solved only in a very trivial sense. Very little is gained from either an explanatory or an evaluative perspective by describing a dictatorship as a collective agent. To explain the behavior of a state ruled by a single person, one need simply appeal to the reasons of that person—no need to attribute any kind of reasons to the collective itself. And whatever normative force the decisions of the one-person sovereign have, they have because the subjects have authorized that sovereign, not due to the decisions themselves or how they were made. Again, nothing seems to be gained by describing those decisions as the decisions of a collective agent. An artiicial person ruled absolutely by a single natural person is thus barely a person at all, and so it is hard not to think that Hobbes dodged the really signiicant part of his problem.3 he really interesting part of Hobbes’ Problem, then, is how to attribute actions based upon reasons to the group in a way that does not reduce to the actions based upon reasons of a single individual. (Otherwise, the actions are properly attributed to the individual, not the group.) And so somehow the actions of the group must give rise to a collective set of reasons that form the basis for the decisions undertaken by the group. his is the problem that social choice theory has attempted to solve. Social choice theory’s eforts to solve Hobbes’ Problem begin with Kenneth Arrow’s Social Choice and Individual Values.4 Arrow sought to specify how a collective preference ordering might be deined as a function of a set of individual preference orderings. An agent’s preference ordering was simply a ranking of all of the alternatives available—for example, option x is superior to option y, which is ranked equally with option z. A ranking of this nature is a minimal requirement for rational choice; if an agent cannot express a preference for the option chosen over the options rejected, then there’s really no sense in which she is acting on the basis of reasons. Arrow set down a number of minimal conditions that any function must satisfy before it would make sense to treat the result as a collective preference ordering. hese conditions include the weak pareto condition (if every member of the group prefers x to y, then so does the group), the independence of irrelevant alternatives condition (the collective ranking of x and y can only be determined by the individual rankings of x and y, and nothing else), and the nondictatorship condition (the collective preferences are not reducible to the preferences of any one individual— that is, the group is not a collective agent in the degenerate sense of Hobbes’ one-person sovereign).5 In what has come to be known as Arrow’s stone | Hobbes’ Problem | 5 Impossibility heorem, Arrow demonstrated that no function yielding a collective preference ordering could simultaneously satisfy all of these conditions. he only kind of function satisfying both the weak pareto and independence of irrelevant alternatives conditions is a dictatorial one—the degenerate Hobbesian case of collective agency. Arrow, it should be noted, assumed that the collective agent would face a decision involving three or more alternatives. When only two alternatives are involved, there is a function capable of generating a collective preference ordering that satisies the nondictatorship and weak pareto conditions, and several other attractive conditions as well—simple majority rule.6 his is a reasonable enough assumption to make. Most real-life decisions involve three or more alternatives; if the agent only considers two of them, it is because of some earlier process that reduces the size of that set. And this earlier process must surely be The only function considered part of the rational decision-making prothat can generate a cess of the agent. minimally rational Social choice theorists have generated an extremely set of collective rich literature in response to Arrow’s heorem.7 his literature has taken Arrow’s treatment of Hobbes’ preferences is Problem in three diferent directions. First, social a function that choice theorists have relaxed the rationality conassigns a veto on dition assumed by Arrow. Obviously, a collective group decisions to agent must be able, in some minimal sense, to act each member of on the basis of reasons, if the collective is to be some subsection of regarded as an agent at all. But perhaps this does not require everything Arrow assumed it would. Arrow’s the group . . . collective agent, like the individual agents that comprise it, has complete and consistent preferences over all the options it faces. his is suicient to produce rational choice, but not necessary. Might it be possible for a function of individual preferences to generate not a collective preference ordering but at least a set of preferences adequate for action on the basis of reasons? In fact, it is possible, but at a rather high price. he only function that can generate a minimally rational set of collective preferences is a function that assigns a veto on group decisions to each member of some subsection of the group (which could be the entire group; Austen-Smith and Banks 1999, heorem 2.4). In many respects, this is the exact opposite of the degenerate case of the collective agent considered by Hobbes, but it poses problems no less severe for the idea of collective agency. he preferences of the collective person now look, not like the 6 | THE GOOD SOCIETY | vol. 24, no. 1 preferences of a single natural person but like the preferences of each and every member of the veto bloc.8 If the idea of a collective agent is to make any contribution to political science, whether empirical or normative, the collective agent must be distinct enough from each agent comprising it to justify the additional theoretical complexity. It is questionable whether the collective actor generated by the liberum veto qualiies. Second, social choice theorists have introduced the spatial model of voting, which restricts the possible conigurations of preferences that the individual members comprising the collective agent might hold. Arrow assumed that his preference aggregation function would have a universal domain—that it should be able to generate a collective preference out of any conceivably set of individual preferences. But perhaps this assumption is too demanding. Perhaps if the members of a group are too dissimilar in their desires, it would not make sense to speak of the group as an agent. One obvious way to restrict preference conigurations is to assume that preferences can be represented in a space of one or more dimensions. Individuals have “ideal points” in such a space and prefer options closer to their most preferred points to options further away. If the space has only one dimension—a line—then this structure is indeed enough to generate meaningful collective choices. If the preferences of a collective can be arrayed in a natural way along a single dimension (as with the let-right ideological dimension in politics), then a straightforward restriction on preferences produces the median voter theorem.9 In this case, the ideal options of the voters can be arrayed along the single dimension of interest, and the collective’s most preferred option coincides with the median of these ideal points. he restriction to one dimension, however, is incredibly limiting in the real world; even in politics, voters care about more than the point on a let-right axis on which a candidate stands (assuming such an axis could even be constructed). And in a multidimensional issue space, the conditions required to produce even a minimally rational collective preference become extremely demanding. Speciically, group member preferences must be distributed in a very precise way if an otherwise sensible preference aggregation function is to yield a core, that is, a set of options that cannot be beaten by any option outside that set.10 If preferences are not distributed in this way, then there is no core, or equivalently, the core is constituted by the entire space of options.11 he spatial model thus provides little comfort to anyone trying to use social choice theory to specify collective agency. hird, social choice theorists have considered the problem of incentive compatibility. his problem is admittedly crucial for any theory of collective agency constructed using social choice theory, although the other problems stone | Hobbes’ Problem | 7 facing any such theory remain. Suppose that the problems of social choice theory could be solved—that one could identify a collective set of reasons suitable for dictating the actions of a collective person and that one could do this using the reasons available to the natural persons that comprise it. It will be the actions of those natural persons, not simply their reasons for action, which determine how the collective person will act. And those natural persons might not ind it in their interest to reveal what they really know or think. his kind of strategic misrepresentation happens when, for example, a voter votes for a less-preferred candidate that has a higher chance of winning. he GibbardSatterthwaite theorem demonstrates that this problem is unavoidable; any nondictatorial function used to generate collective preferences is prone to strategic manipulation of this kind, provided a few other conditions are met.12 Despite this impossibility result, there is guarded reason for optimism in this area. he Gibbard-Satterthwaite theorem presupposes that all The fact that voters vote sincerely (naively, one might say), and the incentive shows that a single sophisticated voter might prove compatibility poses capable of taking advantage of the situation. But if all voters are sophisticated—that is, prepared to vote no insuperable in such a manner as to guard against exploitation of problem does this kind—then the problem will not arise.13 If there not make the exists an outcome that can be achieved by all votsolution to Hobbes’ ers voting sincerely, then there will be a way for the Problem any voters to ensure that same outcome via sophisticated 14 voting. In other words, if there is a set of reasons more apparent. that could form the basis for a collective agent’s Riker regarded behavior, then there is a way to ensure that the indithis problem as viduals comprising the collective make this set of insoluble. reasons operative. Of course, the fact that incentive compatibility poses no insuperable problem does not make the solution to Hobbes’ Problem any more apparent. here still remains the problem of generating a set of collective preferences on the basis of which a collective person could act. A number of social choice theorists, notably William Riker in his book Liberalism against Populism, have regarded this problem as insoluble and therefore denied the possibility of collective agency. For Riker, this was reason to deny the possibility of one form of democracy, which he called populism. Democracy, said Riker, cannot reveal the will of the people, for there is no such will to be found.15 here is much to be said for this position—it avoids all the complications raised above—but it is easily misinterpreted.16 It does not prove, for example, that democracy is a bad decision-making procedure. Oligarchies 8 | THE GOOD SOCIETY | vol. 24, no. 1 face the same obstacles to the establishment of a collective agent, whereas dictatorships (as noted before) are collective agents only in the degenerate sense. But Riker’s argument does pose a challenge to anyone wishing to speak of the “people’s will” in the wake of Arrow et al. here are two distinct approaches in contemporary political theory that respond to this challenge. Each approach raises afresh the question of what counts as a rational agent and whether any group could ever count as such. Consider the irst approach. Social choice theory presupposes an extremely thin theory of rationality. An agent has preferences over outcomes as well as beliefs17 about how actions relate to outcomes. he agent then carries out the action that yields the best expected outcome given those preferences and beliefs. But this understanding of rationality says nothing about the quality of those preferences and beliefs. As Jon Elster puts it, “If an agent has a compulsive desire to kill another person, and believes that the best way (or a way) of killing that person is to stick a pin through a doll representing him, then he acts rationally if he sticks a pin through the doll.”18 A broader theory of rationality ofers an alternative way of modeling collective rationality. Citizens usually do not want the state simply to do what they want because they want it; they want the state to do the right thing. And so one can conceive of collective agents as vehicles for identifying the right collective action to take. his is an epistemic understanding of politics.19 he principle result underlying the epistemic approach is the Condorcet Jury heorem (CJT). Suppose that a state faces a choice between two alternatives, one of which is superior to the other. Each citizen has probability p of identifying the correct choice. his probability p exceeds ½ (asking a random citizen is better than tossing a coin) and the probabilities associated with each citizen are independent of one another. hen the CJT states that (1) the probability that a majority will be correct exceeds p, and (2) the probability that the majority will be correct approaches 1 as the state’s size increases.20 Even semicompetent natural persons—persons who igure out the right answer just over half the time—can combine together to produce a highly competent collective person, on the epistemic view. he epistemic approach has its own problems, some of which are easier to address than others. For example, the CJT presupposes that individuals sincerely vote for the option they think best. But even in the epistemic environment, individuals may face the incentive to vote strategically.21 As with social choice theory, this problem is easily overstated. Suppose that one could identify the voting rule which optimally aggregates the judgments of individual voters; in that case, the rule in question would be strategy-proof.22 stone | Hobbes’ Problem | 9 hree other objections, however, give reason to doubt that the CJT solves Hobbes’ Problem. First, the CJT assumes a dichotomous choice. his is extremely restrictive, and it is diicult to extend the results to choices involving three or more options.23 Second, many regard the epistemic interpretation of politics as implausible. Politics, these critics suggest, may sometimes involve questions of right and wrong, but more oten it involves bargaining between actors who simply want diferent things. Finally, a skeptic might doubt whether the epistemic approach really solves Hobbes’ Problem. As with the liberum veto, the collective actor generated using the CJT approach could be described as a degenerate case. Once again, it is questionable whether there is anything to be gained by calling such a group a collective actor as opposed to a group of individual actors who share a common understanding of the problems to be solved by politics (even if they disagree regarding the solutions to those problems). List and Pettit he second alternative approach is to make a argue that virtue out of an apparent vice. his is the approach taken by Christian List and Philip Pettit in their book collectives can have Group Agency: he Possibility, Design, and Status of wills of their ownCorporate Agents (2011).24 List and Pettit start with a because there is model of collective decision-making that is broader no sensible way of than that employed by Arrow et al. Arrow gave each deriving collective group member a set of preferences and used them reasons directly in an attempt to derive a group set of preferences. List and Pettit, by contrast, give each individual a set from individual of attitudes. hese attitudes include preferences, for reasons! example, x is better than y, but they also include any other reasons an agent might have that are relevant to decision-making. (he set might include, for example, items like “if the state of the world is s, then x is better than y” and “the state of the world is s.”) In short, each individual has a set of reasons suitable for deciding how to act, a set that could be extremely detailed and complex. List and Pettit then try to derive from these individual sets a set of collective reasons suitable for the collective to decide how to act. Like Arrow, List and Pettit generate an impossibility result. here is no way to derive a coherent and complete set of collective reasons from the reasons held by individuals, if the derivation is to satisfy certain intuitively attractive properties (List and Pettit 2011).25 But List and Pettit draw a surprising result from this fact. heir conclusions can usefully be contrasted with those drawn by Riker in Liberalism against Populism. Riker argued that collectives 10 | THE GOOD SOCIETY | vol. 24, no. 1 cannot have wills of their own, because there is no sensible way of deriving collective preferences directly from individual preferences. List and Pettit argue that collectives can have wills of their own—because there is no sensible way of deriving collective reasons directly from individual reasons!26 Where Riker sees a situation that precludes a solution to Hobbes’ Problem, List and Pettit see the only situation that could produce such a solution. If collective reasons could be derived straightforwardly from individual reasons—through simple majority rule, for example—then there would be little for them to do. One wouldn’t need them to explain the behavior of the collective agent—one could accomplish this purely by reference to individuals and their reasons—and so there would be little reason for calling the collective agent an agent at all. List and Pettit regard such a collective agent as “degenerate” much as the dictator-as-collective-agent case, and so they reject the “thin, rather redundant realism that allows easy translation from talk of group agents into talk of individual agents” (List and Pettit 2011, 76). For List and Pettit, if collective agency is possible, then the reasons of the artiicial person cannot simply be read of the reasons of the natural persons that comprise it. he process determining the group’s reasons must necessarily be more complex than that. List and Pettit pay particular attention to the idea of collective reasoning. hey distinguish between a rational agent, which acts on the basis of reasons (such as preferences and beliefs), and a reasoning agent, which can relect on those reasons and adjust them in light of new information (List and Pettit 2011, 29–31). Whereas the standard model of rational action is static, a model of a reasoning agent must be dynamic in nature.27 List and Pettit imagine a group capable of such reasoning. heir impossibility result indicates that no simple aggregation rule will allow a group to generate a consistent and complete set of reasons for actions. hey imagine a group capable of recognizing inconsistencies when they arise in the process of generating collective reasons and taking action in order to minimize or eliminate them. his use of feedback requires a complex organizational structure, but it is precisely this complexity that for List and Pettit makes any kind of meaningful collective rationality possible. If List and Pettit are correct, then the form of collective rationality generated by such a procedure can be as sophisticated as that demonstrated by any natural person.28 List and Pettit may distinguish between the reasoning agent and the rational agent, but they clearly believe that rationality is a necessary (but not suicient) condition for reasoning. And so their picture of a rational collective agent must not depend on reasoning, even if they hope that rational collective agents will also reason. his means that List and Pettit need a story about just stone | Hobbes’ Problem | 1 1 what collective-level reasons are and how they work. And here List and Pettit are walking a very thin line. On the one hand, they deny that the reasons of a rational collective agent can be straightforwardly derived from the reasons held by the collective’s individual members; if they could be so derived, the collective agent would be degenerate in the Hobbesian sense. On the other hand, they also airm that the reasons of the natural persons comprising a collective provide the only possible source for the reasons of the artiicial person they comprise.29 he relationship they want to establish between natural and artiicial persons must therefore of necessity be a semimysterious one. he less mysterious the relationship becomes, the more the artiicial person fades from view and the natural persons come into focus. But the more mysterious the relationship, the less reason there is to believe it exists in the irst place. Perhaps List and Pettit can navigate between these two dangers, but how exactly this is to be done remains to be speciied convincingly. he quest to describe the nature of a collective agent—an “Artiiciall man,” to use Hobbes’ term—has motivated a signiicant part of the ield of positive political theory. It has informed the development of social choice theory, the Condorcet Jury heorem, and the model If Riker is correct, of group agency developed by List and Pettit. he quest then groups are is motivated by both empirical and normative concerns; indeed, there is a close relationship between the empirisimply not agents in cal and normative concerns here, as with many problems any meaning sense. in political science. But at the same time, positive political theory has not yet produced a deinitive answer to Hobbes’ Problem. Rather, its most surprising and interesting contribution has been to raise the question of whether the problem can be solved at all—or even whether it is a problem that needs to be solved. If Riker is correct, then groups are simply not agents in any meaningful sense. Political science, then, faces the problem of either proving Riker wrong (as List and Pettit have tried to do), or else developing the full implications (both empirical and normative) of Riker’s position for the study of politics. To perform either task would be to take political science to a new level of sophistication. And neither task would even be imaginable today without the help of positive political theory. Peter Stone is Ussher Assistant Professor of Political Science at Trinity College Dublin. He is the author of he Luck of the Draw: he Role of Lotteries in Decision Making (Oxford University Press, 2011) and the editor of Lotteries in Public Life: A Reader (Imprint Academic, 2011). 12 | THE GOOD SOCIETY | vol. 24, no. 1 NOT ES A preliminary version of this paper was presented at the Friday Seminar Series, Political Science Department, Trinity College Dublin. his seminar was quite helpful to me in formulating my ideas. Work on this paper was supported by a grant from the Arts and Social Sciences Benefaction Fund at Trinity College Dublin. 1. Emphasis in original. homas Hobbes, Leviathan, ed. C. B. Macpherson (New York: Penguin, 1968), 81, 82. 2. Donald Davidson, “On the Very Idea of a Conceptual Scheme,” in Inquiries into Truth and Interpretation (Oxford: Clarendon Press, 1984); Daniel Dennett, he Intentional Stance (Cambridge, MA: MIT Press, 1987). 3. Christian List and Philip Pettit characterize such an artiicial person as a “degenerate” case. “Rather than speaking of a group or corporate agent, we might as well speak of a corporately empowered individual: an individual—in the political case, a sovereign—capable of calling on those pledged to provide support.” Christian List and Philip Pettit, Group Agency: he Possibility, Design, and Status of Corporate Agents (New York: Oxford University Press, 2011), 8. 4. Kenneth J. Arrow, Social Choice and Individual Values, 2nd ed. (New Haven: Yale University Press, 1963). Duncan Black covered much of the same ground as Arrow around the same time, but his eforts had less of an impact. See S. M. Amadae, Rationalizing Capitalist Democracy: he Cold War Origins of Rational Choice Liberalism (Chicago: University of Chicago Press, 2003), 122–28. 5. here are several equivalent formulations of Arrow’s heorem. For the formulation followed here, see David Austen-Smith and Jefrey Banks, Positive Political heory I: Collective Preference (Ann Arbor: University of Michigan Press, 1999), heorem 2.1. 6. Kenneth O. May, “A Set of Independent Necessary and Suicient Conditions for Simple Majority Decision,” Econometrica 20 (1952). he independence of irrelevant alternatives condition is trivially satisied when only two alternatives are involved. 7. For a detailed exploration of this literature, see Austen-Smith and Banks (1999), as well as David Austen-Smith and Jefrey Banks, Positive Political heory II: Strategy & Structure (Ann Arbor: University of Michigan Press, 2005). 8. his type of degenerate collective agent—one that acts in a way that goes unvetoed by members of some group—provides one way to understand one of Rousseau’s most cryptic passages about the general will: There is often a great difference between the will of all . . . and the general will; the general will studies only the common interest while the will of all studies private interest, and is indeed no more than the sum of individual desires. But if we take away from these same wills, the pluses and minuses which cancel each other out, the balance which remains is the general will. See Jean-Jacques Rousseau, he Social Contract, trans. Maurice Cranston (New York: Penguin, 1968), 72–73. 9. Harold Hotelling, “Stability in Competition,” Economic Journal 39 (1929); Anthony Downs, An Economic heory of Democracy (New York: Harper Collins, 1957). 10. Charles Plott, “A Notion of Equilibrium and Its Possibility under Majority Rule,” American Economic Review 57 (1967); Richard McKelvey and Norman Schofeld, “Generalized Symmetry Conditions at a Core Point,” Econometrica 55 (1987). 11. Richard McKelvey, “Intransitivities in Multidimensional Voting Models and Some Implications for Agenda Control,” Journal of Economic heory 12 (1976); Richard stone | Hobbes’ Problem | 1 3 McKelvey, “General Conditions for Global Intransitivities in Formal Voting Models,” Econometrica 47 (1979); and Norman Schoield, “Generic Instability of Majority Rule,” Review of Economic Studies 50 (1983). 12. Allan Gibbard, “Manipulation of Voting Schemes: A General Result,” Econometrica 41 (1973); Mark Allen Satterthwaite, “Strategy Proofness and Arrow’s Conditions: Existence and Correspondence heorems for Voting Procedures and Social Welfare Functions,” Journal of Economic heory 10 (1975). Interestingly, Gibbard also showed that a diferent set of conditions yields one strategy-proof nondictatorial voting rule—lottery voting, sometimes called the random dictator rule. See Allan Gibbard, “Manipulation of Schemes hat Mix Voting with Chance,” Econometrica 45 (1977). his rule has its defenders; see, e.g., Akhil Reed Amar, “Choosing Representatives by Lottery Voting,” Yale Law Journal 93 (1984). But lottery voting hardly seems like a promising rule for constituting a collective agent. Collective agency must involve action on the basis of reasons, whereas random selection means relying upon a process that explicitly excludes reasons from decision-making. See Peter Stone, he Luck of the Draw: he Role of Lotteries in Decision Making (New York: Oxford University Press, 2011). 13. Cf. Gerry Mackie, Democracy Defended (New York: Cambridge University Press, 2004). 14. David Austen-Smith, “Sophisticated Sincerity: Voting over Endogenous Agendas,” American Political Science Review 81 (1987). 15. William H. Riker, Liberalism against Populism: A Confrontation between the heory of Democracy and the heory of Social Choice (Prospect Heights, IL: Waveland Press, 1982). Interestingly, Riker had a strong interest in Hobbes; apparently, he originally intended to write his doctoral dissertation on the sage of Malmesbury. 16. For a lengthy discussion of some such misinterpretations, see Mackie (2004). 17. Social choice theorists have typically abstracted from the problem of beliefs. heir results do not depend on diferences of belief among individual agents, only diferences of preference. 18. Jon Elster, Sour Grapes (New York: Cambridge University Press, 1983), 3. 19. Joshua Cohen, “An Epistemic Conception of Democracy,” Ethics 97 (1986). William Riker’s alternative to populism, which he dubs liberalism, is (perhaps ironically) epistemic in nature. Democracies, according to Riker, cannot enact the will of the people, but they can provide citizens with the means of ejecting from oice politicians that violate basic rights. But this presupposes that the purpose of democracy is to ensure that basic rights are respected. his means identifying just which policies really do respect rights the best, which is obviously an epistemic problem. See Peter Stone, “Voting,” in he Encyclopedia of Political hought, edited by Michael T. Gibbons (Malden, MA: Wiley-Blackwell, 2014). Unfortunately, Riker gives no reason for believing that democracies will systematically reject rights-violating politicians; they might just as easily reject rights-protecting politicians. See Jules Coleman and John Ferejohn, “Democracy and Social Choice,” Ethics 97 (1986). 20. For a simple proof, see Krishna K. Ladha, “he Condorcet Jury heorem, Free Speech, and Correlated Votes,” American Journal of Political Science 36 (1992): 632–33. 21. David Austen-Smith and Jefrey Banks, “Information Aggregation, Rationality and the Condorcet Jury heorem,” American Political Science Review 90 (1996); Timothy Feddersen and Wolfgang Pesendorfer, “Convicting the Innocent: he Inferiority of Unanimous Jury Verdicts under Strategic Voting,” American Political Science Review 92 (1998). 14 | THE GOOD SOCIETY | vol. 24, no. 1 22. Ruth Ben-Yashar and Igal Milchtaich, “First and Second Best Voting Rules in Committees,” Social Choice and Welfare 29 (2007). 23. H. Peyton Young, “Condorcet’s heory of Voting,” American Political Science Review 82 (1988). 24. Not coincidentally, Pettit has a strong interest in Hobbes. See Philip Pettit, Made with Words: Hobbes on Language, Mind, and Politics (Princeton, NJ: Princeton University Press, 2009). 25. For the original result, see Christian List and Philip Pettit, “Aggregating Sets of Judgments: An Impossibility Result,” Economics and Philosophy 18 (2002). Speciically, the function used to aggregate individual reasons is supposed to satisfy universal domain (individuals may hold any complete and consistent set of reasons), collective rationality (the collective must wind up with a complete and consistent set of reasons), anonymity (the reasons held by one individual are treated in the same manner as the reasons held by another), and systematicity (the function determines which reasons the group holds using only individual judgments regarding that reason, and it makes this determination in the same way for all possible reasons the group might hold). 26. As noted before, the set of attitudes that provide agents with reasons for action includes preferences. As a result, List and Pettit’s impossibility result (which deals with reasons) implies Arrow’s impossibility result (which deals with preferences) as a corollary. See Franz Dietrich and Christian List, “Arrow’s heorem in Judgment Aggregation,” Social Choice and Welfare 29 (2007); and Elad Dokow and Ron Holzman, “Aggregation of Binary Evaluations,” Journal of Economic heory 145 (2010). Both are cited in List and Pettit (2009, 50 n. 40). 27. his is true in standard economic theory (which seeks equilibria given ixed preferences and endowments in a market), in game theory (which seeks equilibria of a diferent sort), and in social choice theory (which treats preferences as ixed and then aggregates them under diferent rules). 28. At the same time, List and Pettit recognize that “feedback alone is insuicient for group reasoning,” even though it does appear to be necessary (List and Pettit 2011, 64). he feedback must be handled in a particular way, and it is hard to see what forces could lead a group to create feedback processes of this nature. In addition, List and Pettit identify other forms group agency could take in light of their impossibility result, forms that generate rational but not reasoning group agents (ibid., sec. 3.1). Is there any reason to expect reasoning group agents to be more likely to form than nonreasoning ones? Is there any reason why a person might prefer to belong to reasoning group agents rather than nonreasoning ones? List and Pettit do not say. 29. In technical terms, List and Pettit reject proposition-wise supervenience—“the group attitude on each proposition is determined by the individual attitudes on that proposition”—but embrace holistic supervenience—“the set of group attitudes across propositions is determined by the individual sets of attitudes across these propositions” (List and Pettit 2011, 68, 69).