p9

Defenders of this radical proposal, like defenders of the moderate proposal—and like defenders of the idea that group agents should be practically restricted—must treat group agents as seriously deficient in comparison with the individual agents who make them up. 7 Escape Route 3: Group Agents Are a Fiction The first escape route suggests that group agents should not be allowed to act on the basis of votes that would give rise to a discursive dilemma, only on a more demanding basis, implying in effect that group agents should be practically restricted in a way that would make them infeasible. The second escape route, which comes in two versions, suggests a parallel result. Arguing that group agents should not be allowed to treat on–off votes properly as evidence (either as non-derivative evidence or as evidence at all), they imply that group agents should be epistemically restricted in a way that would also make them infeasible. Both of these escape routes suggest that group agents provide no real challenge for the view that the requirement of coherence is merely a byproduct of the requirement of reasons-sensitivity. They suggest that the group agents that might be formed on the template described in Section 4 cannot be a source of challenge for that view. The third escape route we should mention is even more radical than the other two. It argues that the group agents that might be formed on the Section 4 model do not count as agents at all, or not at least as agents in more than a fictional sense. This escape route is potentially more interesting than the other two. Those escape routes start from a thought that is close to the view that reasons-sensitivity is the only appropriate requirement on agents. The thought behind these routes is that group agents should not be allowed to act on on–off votes, or to treat on–off votes asREASONS AND RATIONALITY 225 evidence, or to think that responding to on–off votes articulates the demands of reasons. And the motivation for that thought is presumably that if such votes are taken as articulating those demands, then that will mean that tracking reasons may fail to ensure coherence. Thus those escape routes may seem to presuppose the very thesis they are designed to preserve. No such difficulty would attend the third, for it argues on quite independent lines against the very reality of group agents. The argument against the reality of such agents may be that groups cannot be agents at all or that they cannot be agents in a real as distinct from a fictional sense. The argument that they cannot be agents at all, which abounds in some economics textbooks, and in some legal discussions, routinely cites the observation that group agents come into existence in virtue of contractual relations among their members, as if that in itself meant that they could not also count as agents. This approach views commercial companies—and by extension other forms of group agent—as sites of contracts akin to the market, which are distinguished only by the fact that the nexus of contracts between parties is much more pervasive than in markets more generally. One commentator formulates the core idea as follows: “the Nexus of Contracts Theory…treats the company as little more than a collective noun for the web of contracts that link the various participants, which include shareholders, management, employees and creditors. The function of company law is thus conceived of as the facilitation of the parties’ bargains” (Grantham 1998: 579). The assumption in this approach seems to be that since group agents are made out of the same material as markets—individuals in contractual relations with one another—they have no more claim to be agents than markets themselves have. But while markets are certainly not agents—while it is mere metaphor to speak of what the markets think or expect—corporate bodies certainly are (List and Pettit 2011). Unlike markets, they fit the functional characterization of agency given earlier. They have discernible purposes and representations such that in general they reliably pursue those purposes in accordance with reliably formed representations. They may occasionally fail, as we all fail, but when they do so they are generally ready to recognize the failure and even, if appropriate, to make amends. To say that they are not agents on the grounds that they are composed out of individuals in contractual relationships with one another would be as unconvincing as saying that you and I are not agents because we are composed out of cells in biological relationships with one another. What makes a system an agent is not the stuff out of which it is composed but the purposive-representational function that it is capable of discharging. Group agents can function in a distinctively agential way, embracing goals, forming views about how best to realize those goals, and pursuing the goals on the basis of those views. And moreover they can make commitments to other agents, individuals or corporate, by speaking in a self-committal way about what they hold or what they will do. It might be said that there are some clearly non-agential systems, such as the heating system in a building, that nonetheless fit the template for agency: they act226 LARA BUCHAK AND PHILIP PETTIT reliably for a goal on the basis of a reliable representation, as the heating system maintains the temperature of a building above a certain level on the basis of registering any fall below the threshold. Perhaps group agents are like that, it may be objected; perhaps they fit the template but still fail in an intuitive manner to count as agents. The suggestion is absurd, for two reasons. One, the group agent can commit itself in words, which the heating system cannot (Pettit 2014b). And two, there are indefinitely many ways in which a given goal or representation may be realized in a group, whereas the system realizes its goal and representation on the sole basis of a simple mechanical construction; thus, the language of goal and representation is extravagant in the latter case, but not in the former. Turning now to a second form of this third escape route, group agents may be said to be agents but not agents proper: not agents in anything more than a fictional sense. The best interpretation of this line on group agents, which goes back to Thomas Hobbes (1994 [1651]: ch. 16), holds that a group agent comes into existence by fiction insofar as a pre-existing agent or agential system is recruited to an extra role over and beyond the role of enacting and speaking for their or its own attitudes (see Pettit 2008: ch. 5, 2014a; Skinner 2009). In Hobbes’s picture, a group agent may come into existence in either of two ways. It may exist in virtue of a particular individual being authorized or accepted by the members of the group as speaking for them: this authorization will involve a readiness on the part of members to act as required in order for the commitments made by that individual in the name of the group to be fulfilled. Alternatively, a group agent may exist in virtue of the fact that an independent, more or less mechanical agential system like a majoritarian committee is authorized or accepted in the same way by the members of the group. The idea is that in neither case is the group agent anything other than a fiction. It makes commitments, by courtesy of its spokesperson, and it generally lives up to those commitments, holding by the beliefs avowed and keeping the promises made. But it does so in a way that is wholly parasitic on an independent, pre-existing agent or agency: hence that agent or agency can count as the real agent, with the group counting only as an agent by fiction. The main examples of group agents that operate on the model of Section 4 are not agents by fiction in this sense. They rarely have a single, dictatorial spokesperson who is entitled to speak for all members, and that case may be regarded as irrelevant. And they cannot operate on the basis of recruiting an independently agential committee— a committee that mechanically generates statements of purpose and representation f it to be authorized by a group agent—for reasons with which we are now familiar. A mechanical procedure like majority voting is not a reliable source of statements that the members of a group agent might endorse, being ready to act as the commitments embodied in those statements require. Any sort of committee, majoritarian or otherwise, that uses a bottom-up procedure to generate candidate commitments is liable, as Hobbes failed to see, to produce statements that are inconsistent with one another.REASONS AND RATIONALITY 227 This means that in order for a group agent to come into existence, or at least a group agent that does not operate via a dictator, its members have to construct that agent from scratch, using procedures with a top-down as well as a bottom-up component in order to produce statements of purpose or representation that can mediate the commitments of an agent. This means that the agent constructed will be a distinct agent from its members—there will be no pre-existing agent with which it might be identified—even though it is composed entirely out of those individuals and their relationships. Once established as a group agent, the body in question may serve in a Hobbesian fashion as a spokesperson that the members of a wider group can authorize, thereby constituting themselves as an agent by fiction. But in order for such parasitic forms of group agency to materialize, there have to be group agents that exist and function in a non-parasitic way. Conclusion We do not think that the considerations mustered over the last three sections establish definitively that the case of group agents undermines the case for treating the requirements of reasons-sensitivity as the only real requirements on agents. Nor do we believe that the escape routes described necessarily are the only avenues that those attached to that view might explore. We hold only that the case of group agents provides new data in light of which to consider the argument for and against an error theory about the requirements for coherence. But suppose we decide that the escape routes for the error theory really are closed, maintaining that the group agents cannot be dismissed as relevant and that they do operate in a way that gives coherence requirements a real, guiding role. What in that event should we conclude about the error theory? One potentially attractive line is to hold that the need for group agents to rely on the coherence requirements, using their non-satisfaction to guide its attitudes, teaches us that a corresponding lesson holds with individual agents too and that the error theory is not quite right, after all (at least in the case of conjunction).10 WhenIasanindividual look to the evidence for and against a particular proposition, I have to look to what my eyes and ears tell me, to what I learn by testimony, and to what accumulated experience and theory support. But these channels of evidence may speak to me with many voices, as the votes of members speak with many voices to a group. Just as group agents have to listen to the voices of members in constructing the view they form on any proposition, so individual agents often have to listen to 10 It is worth noting that Kolodny does not claim that the error theory will explain all purported coherence norms (Kolodny 2007a: 253–7), and conjunction may be among those to which it does not apply. Those who take the viewthat the individual error theory does not apply to conjunction may consider the case of group judgment to be support for their view.228 LARA BUCHAK AND PHILIP PETTIT various voices—the different sensory, testimonial, and memory-based channels of evidence—in determining the view that they ought to take on a proposition. And the suggestion would be that in forming the required view, both group agents and individual agents have to rely on coherence in a way that does not merely amount to treating incoherence as additional evidence about what they have reason to believe. They have to silence the evidence available at any moment in support of a proposition when they check and find that it would give them an incoherent profile of views. Andthey then have to consider how to resolve the conflicting bodies of evidence that occasion the incoherent views. This suggestion, which turns on an epistemic analogy between the position of individual and group agents, may be buttressed by consideration of a further analogy on the practical side. This is that just as group agents are going to fail the rationale for establishing them unless they are prepared to form beliefs in the absence of conclusive evidence, so something similar holds of individuals. Those who defend an error theory about the requirement of coherence suggest that in the absence of conclusive evidence for believing p, or for believing not-p, we individual agents ought to suspend judgment. But this may well look like an unworkable constraint. If as individuals we were to form beliefs only in the presence of conclusive evidence—only when the facts spoke to us in anunambiguousway—thenwewouldhavetosuspendbelief on agreat range of matters. As individuals we would be in much the same position as group agents that were required to form beliefs only when a unanimity constraint is satisfied. Like such group agents, we would be unimpeachable in epistemic terms. But, like such groups, we would also be rendered more or less ineffective as centers of agency. Wewould be incapable of making up our minds on so many matters that we would be deprived of the beliefs essential for resolving most of the choices that arise in ordinary life: most of the choices that the pursuit of our goals requires us to make. Nor is that all. We not only need to have a rich set of beliefs in order to be able to pursue the purposes that are important for us as individual and group agents. We also need to be able to speak for such a set of beliefs in establishing relationships with one another and in presenting ourselves to others as agents with whom they can do business: agents they can expect to be willing to stake out positions on various issues and to prove generally reliable in acting in the manner associated with those positions. Suppose we were to shrink from forming opinions in the absence of more or less conclusive evidence. Or suppose we were to form such beliefs without practical guidance from the requirement of coherence. In any such event we would prove to be less than fully conversable, as we might put it; less than fully capable of relating to one another reliably in avowing beliefs or intentions or in promising to perform one or another action. These two analogies are less than perfect, of course, and even if we reject the escape routes described in the last three sections, they do not make an irresistible case for applying the lesson of group agents to individuals. When we try as individuals toREASONS AND RATIONALITY 229 make sense of various evidential inputs in order to determine our view on a proposition, we do not employ a rote procedure of tallying the “votes” of our individual sources of evidence. Rather, we consider how reliable each source is apt to be in the given case, and how the deliverances of one source bear on the deliverances of the other. And, unlike the group case, in which degrees of belief would need to be explicitly expressed in order to be used by the group to determine the group’s opinion on the proposition, we can take degrees of support offered by various evidential sources into account in a largely automatic, subpersonal manner. Furthermore, the second-order evidence that we are mistaken in some judgment is easier to employ, since as individuals we often have a clear view about where mistakes are more apt to occur. By contrast, the group agent has no privileged vantage point from which to adjudicate the views of the various members, or to judge the relative reliability of members on a particular question. As wedonotsaythat the escape routes available to those who hold an error theory about the requirement of group coherence are definitively closed, then, so we do not say that even if they are closed, that means that the error theory should be given up on all sides: not just for groups but for individuals too. All we claim to have established is that there is an interesting tension between the error theory of coherence, on the one side, and the theory of group agency on the other. The two theories are associated with different literatures and different traditions and what we hope to have shown is that it may well be profitable to bring them more closely into contact.