As I shall use the term “group agency” in this paper, a group agent exists when the condition of individual agency is realized within a group of human beings, with the result that the group agent can function, and also be treated as, an individual agent in its own right.

The question I want to address in the paper is whether it is correct to regard group agency in this sense as a social phenomenon.

It might be wondered how group agency could fail to be a social phenomenon since it is, by definition, a case of agency involving many distinct human beings. But I do not mean “social” in the biologists’ sense, on which many organisms are involved, for it is obviously, and indeed, trivially true that group agency is social in that sense. I mean “social” in the sense that is of interest to philosophers working in the fields of action theory, social ontology, and moral and political philosophy, on which social relations arise among reflective rational agents who mutually recognize one another as such, and as a result, also recognize their social relations as such. To view group agency as a social phenomenon in this sense is to view it as a special case of collective agency—as the agency of many agents, each of whom remains an individual agent in its own right, as they exercise their agency together in such a way as to be the agency of one.Footnote 1

Even with these clarifications, it might still be wondered how group agency in the sense at issue could fail to be a social phenomenon—for it might be wondered how the human constituents of a group agent could fail to be individual agents in their own rights, and correlatively, how the intentional activities through which group agency is achieved could fail to be social activities on the part of those individual agents of human size. I will address these wonders in due course. But first some preliminaries.

It bears mention that when I ask whether it is correct to regard group agency as a social phenomenon, I am not raising an empirical question. I am calling for philosophical work that is part conceptual analysis, part metaphysical argument and part normative recommendation. I take it for granted that such philosophical work should rightfully inform any empirical claims that we make about this or that empirical case, as well as any normative stance that we take regarding group agency.

It also bears mention that my arguments about group agency in this paper require only a possibility claim, to the effect that it lies within human practical capability to forge group agents who satisfy the condition of individual agency. It is compatible with this possibility claim that no such group agent has ever yet come into existence—though I do happen to think that some of the empirical cases that philosophers have discussed may qualify.

My guiding question, concerning whether it is correct to view group agency as a social phenomenon, is open to different modal interpretations. Thus we may ask, more specifically: Must group agency be a social phenomenon? Is it so much as possible that group agency be a social phenomenon?

My answers to these modal questions draw on a reductionist account of agency, according to which an agent is nothing but, or nothing more, than a body of intentional episodes standing in the right sorts of relations, so as to fall together under a commitment to rational unity, with the understanding that this commitment to rational unity is constitutive of individual rationality.Footnote 2 On the reductionist view, group agency arises when the right sorts of relations hold among intentional episodes that are located in different human lives, so that the condition of individual rationality is realized within a group of human beings. When this is so, the intentional episodes that constitute the life and point of view of the group agent stand in intra-personal relations even though they are located in different human lives.

Reductionism invites us to reformulate the question whether group agency either must be, or can be, a social phenomenon in terms of the question whether the intra-personal relations that reach across the human constituents of a group agent either must be, or can be, inter-personal relations as well. The answer to the question, so reformulated, turns on whether the human constituents of a group agent either must be, or can be, the site of separate commitments to achieving overall rational unity through the whole of their separate human lives, so as to constitute individual agents of human size who co-exist alongside the group agent.

Reductionism entails that this is not necessarily so, and furthermore, that it is not typically so. The reason why is that reductionism affords the possibility of rational fragmentation within a human life. If this way of putting the point makes it sound unduly negative, let me try to put it more positively. According to reductionism, the sort of rational unity that is characteristic of individual agency is always an achievement, and it can be achieved within different boundaries. It can be achieved within a whole human life, so as to constitute an agent of human size; and it can be achieved within a group of human lives so as to constitute a group agent; and it can be achieved within parts of a human life so as to constitute multiple agents within it. When such unity is achieved across human lives so as to constitute a group agent, this typically comes at the cost of achieving rational unity within the lives of the group agent’s human constituents, and as a result those human lives are typically sites of rational fragmentation in the following sense: some of the intentional episodes in those human lives figure in the life of the group agent, leaving the remainder to constitute the life of an agent who is somewhat smaller than human size (though of course, only so long as there is a commitment to achieving rational unity within that remainder).Footnote 3

The question whether group agency can be a social phenomenon is much harder to answer than the question whether it must be. My working view is that reductionism casts doubt on whether group agency can be a social phenomenon, except in the very special condition that Rousseau identified in connection with his conception of the general will. What he envisaged was a political agent—in the form of a community governed by a social contract—that exists for the sole purpose of preserving the individual liberty and well being of the citizens who comprise it. It was Rousseau’s ideal that it should be possible for each citizen to be wholly identified with the common will of the political agent that it helps to constitute, and I take this to be just another way of saying that it should be possible for citizens to forge a common will without incurring the sort of rational fragmentation that I have just said is the typical concommitant of group agency. As interesting as Rousseau’s ideal is for the purposes of political theory (and real politics), it is not a good or convincing model for group agency in general. If it were, then we should have to suppose that, in general, the only aims for the sake of which group agents would ever exist would be the aims of agents of human size within them. But this simply is not a good or convincing way to characterize the aims of many of the candidates for group agency that philosophers have discussed, such as corporations, armies, orchestras, philosophy departments, sports teams, etc.

I think most philosophers who argue for the possibility (or reality) of group agency believe that human beings are, necessarily, individual agents—or at any rate, that they are necessarily sites of individual agency. We shall see that this mistaken belief does not always sit well with the accounts of individual agency to which they themselves appeal in the course of their discussions of group agency. What is more, it inclines them to the mistaken view that group agency must be a social phenomenon, which then leads them to overlook the various reasons why group agency is not necessarily, or even typically, a social phenomenon.

Here is the plan of the paper:

In Section 1 I sketch the reductionist account of agency and why it entails the possibility of group agency.

In Section 2 I clarify why reductionism entails that group agency is not necessarily a social phenomenon, and how this implication is bound up with the possibility of rational fragmentation within human lives. Along the way, I highlight some of the differences between the genuinely individual deliberations and activities of a group agent, and certain social phenomena with which they might be confused.

Once these differences are fully in view, we will be in a better position to understand why it would be a mistake to take Rousseau’s general will as a paradigm for the will of a group agent. I explain why this is so in Sect. 3—and correlatively, why it would be a mistake to suppose that group agency would typically be a social phenomenon.

In Section 4 I put my arguments in this paper in relation to the work of some fellow travellers, namely: Christine Korsgaard, Michael Bratman, and co-authors Philip Pettit and Christian List.

1 An argument for the possibility of group agency via reductionism

The reductionist account of agency that I defend here takes it for granted that some aspects of agency, such as reasons for action, and rational requirements on thought and action, are irreducibly normative. To say that they are irreducibly normative is to say, among other things, that they cannot be fully captured in functionalist terms, for a functionalist account would reduce them to causal dispositions—specifically, dispositions to act in the light of reasons, and to reason in accord with the requirements of rationality. Such a functionalist reduction would fail to capture the fact that both our reasons for action, and the requirements of rationality that govern our deliberations, introduce normative ideals of which we may fall short, even as we are nevertheless committed to living up to them. Thus, I might have reasons on which to act without necessarily being caused to act upon them, and I might embrace the requirements of rationality without necessarily being caused to deliberate in perfect accord with them. When I possess such reasons, and embrace such requirements, I am in states of mind that cannot be specified except in irreducibly normative terms—they are normative commitments on my part. The fact that I may fail to live up to them is what distinguishes them from mere causal dispositions. But they are nevertheless real commitments on my part, so long as I think I ought to live up to them, and so long as I take my failures to live up to them as occasions for self-criticism. Note that it is the very contents of my own commitments that supply the normative standards for such self-criticism, and this gives the positive sense in which commitments are irreducibly normative, alongside the negative point that they cannot be reduced to causal dispositions.Footnote 4

The argument for reductionism exploits a conceptual tie between the concept of an individual agent and certain normative requirements that define what it is for an individual agent to be ideally rational. These normative requirements include the requirements of consistency (an agent ought to resolve any conflicts among its beliefs), closure (an agent ought to accept the implications of its attitudes), and transitivity (an agent ought to achieve a transitive ordering of its options and values). An agent is able to apprehend, as well as respond to, these normative requirements only from a first person point of view. And from this first person point of view, an agent can raise the deliberative question, what would it be best for me to do, all things considered? The things-to-be-considered are all of the agent’s commitments, by which I mean, the agent’s beliefs about what is the case, along with its evaluative attitudes about what is good (and what is better than what), where these commitments are not construed in functionalist terms as causal dispositions, but as attitudes that hold an irreducibly normative significance. Thus, what an all-things-considered judgment registers is the joint normative significance of all of an agent’s commitments taken together, with regard to the question, what would it be best for me to do? When an agent arrives at and acts upon such judgments, it achieves overall rational unity. The most general rational requirement to which an individual agent is subject is a requirement to achieve precisely such unity, and it strives to do this by striving to meet other more specific requirements such as consistency, closure, transitivity, and the like.

All of these requirements of rationality, general and specific, are internalist in two related senses, both of which reflect the fact that rational activities must always proceed from a point of view. First, rationality does not require me to take into account matters of fact and value concerning which I do not have any commitments already, for such matters are external to my point of view; second, rationality does not require me to take into account the commitments of others, except insofar as my own commitments dictate that I should do so, for otherwise those other-regarding considerations would remain external to my point of view. The second aspect of internalism highlights that while it might be a moral failing on my part to disregard others’ commitments, it need not be a rational failing. It also highlights a deep point about the first personal nature of commitments: it is literally impossible to deliberate from anyone else’s commitments because, in the very act of taking any consideration as a basis for one’s own deliberations, one thereby undertakes a commitment of one’s own. As a result, internalism brings with it an unavoidable individualism, which returns us to the conceptual point from which the reductionist account of agency proceeds, namely, that the requirements of rationality define what it is for an individual agent to be fully or ideally rational.

Taken just by itself, this conceptual point does not directly entail any further metaphysical claim about the condition of agent identity, either in favor of or against reductionism. It merely registers that wherever there is an individual agent, there is something that recognizes, and is committed to meeting, the normative requirements that define individual rationality. The argument for reductionism turns this conceptual point around, and insists that wherever there is a commitment to meeting the normative requirements that define individual rationality there is an individual agent. It should already be clear how an argument for the possibility of group agency should proceed from a successful argument for reductionism along these lines: such an argument should aim to establish that a group of human beings may be the site of a commitment to meeting the normative requirements of individual rationality within the group.

The crucial step in the argument for reductionism insists that the normative commitment that makes for individual agency is never a metaphysical given but always a product of effort and will.

To say that the existence of an individual agent is not a metaphysical given is to say, in particular, that it does not suffice for the existence of an agent that a human being exists as a biologically given thing. Each human being begins life as a wanton in something like the sense that Harry Frankfurt spelled out in his influential account of freedom of the will.Footnote 5 According to him, wantons are simply moved by their desires of the moment, and they do not care what their desires are, and nor do they care which desires move them. What does it take for a human being to leave its initial condition of wantonness, so that it becomes a site of rational agency? It will be helpful to compare and contrast my answer to this question with Frankfurt’s account of freedom of the will.

Frankfurt proposes to model freedom of the will on freedom of action in the sense that Locke and Hume made central. According to them, one acts freely so long as one’s actions are in accord with one’s desires. On Frankfurt’s account, one wills freely so long as the following conditions hold: one has a second order desire to have (or not to have) a certain first order desire; one’s first order desire is in accord with one’s second order desires; in addition, one has a second order “volition”, which is a second order desire that the desired first order desire be effective, in the sense that it issue in action; and finally, one acts in accord with both one’s first order desire and one’s second order volition. If a human being were to meet all of these conditions for freedom of the will as Frankfurt models it, it would certainly not be a wanton—there is a clear sense in which it would care about what its desires are. However, such a human being would not necessarily qualify as a site of rational agency in the sense that I have been elaborating here, which I am claiming involves irreducibly normative commitments from which deliberation and action proceed—for the life of an agent with free will in Frankfurt’s sense would still be a mere play of causal dispositions, albeit interestingly complex ones.

Admittedly, there are some parallels between commitments in the sense I have in mind and the higher order volitions that Frankfurt posits, in the light of which it may seem tempting to try to reduce the former to the latter. The main ground on which I distinguish commitments from dispositions is that an agent may possess commitments without necessarily acting, or even being disposed to act, in accord with them. I claim that the agent can still be said to possess its commitments, so long as it is prepared to criticize itself for its failures to live up to them. Prima facie, it would seem that something like this can also happen with second order volitions—that is, it appears that an agent might have such a volition even if its actions and first order desires were not in accord with it. Yet in spite of this similarity between commitments and second order volitions, it would be misguided to think that there is any straightforward way to reduce the one to the other. Moreover, I think any temptation to attempt such a reduction ought to be resisted. To see why, it will help to consider the following real case. (I do not mean to suggest that a mere example can substitute for philosophical arguments—they will follow.) As it happens, I am always altogether lacking in any first order desire, construed as a mere disposition, to attend political rallies. Yet I have sometimes had second order desires to possess such first order desires; and on occasion I have even had a second order volition to have such first order desires be effective, and carry me all the way to action. Since this has never happened, my second order volition has never been satisfied (and so I have lacked freedom of the will in Frankfurt’s sense). Now, it is indeed tempting to regard this as a case in which my second order volition functions exactly like a commitment, for it seems to register a preference on my part about the kind of person I want to be, in the light of which my not attending political rallies, and my not wanting to attend them, would count as failures on my part for which I should criticize myself. But a more complete elaboration of this real case should help to clarify why we ought to resist this temptation. The question whether I ought, by my own lights, to attend political rallies is not to be settled on a one-off basis, just by noting the presence of my higher order volition concerning this matter; it is to be settled through a more holistic deliberative process that aims to work out what all of my attitudes (=commitments) taken together imply about the matter—an implication that would be captured in an all-things-considered judgment. Here are some considerations (=commitments on my part) that, taken together, entail that I ought not to attend political rallies in spite of the presence of a second order volition to attend them: it is good to engage in some form of worthwhile political activity; attending political rallies is only one of many such activities available to me; I have an aversion to crowds; other things being equal, it is better for anyone to engage in worthwhile activities to which they do not have an aversion. As I’ve noted, I take it to follow from these considerations that I ought not to attend political rallies. (Any reader who doesn’t think that this follows can simply add further considerations into the mix, which they think would suffice to deliver the conclusion.) Insofar as this does follow, it also follows that if I were to continue to harbor a second order volition, to have an effective first order desire to attend political rallies, this would be an occasion for self-criticism! And this is what stands in the way of reducing the idea of a commitment to the idea of a higher order volition. Insofar as our commitments supply our normative standards for self-criticism, they register discords between our considered sense of what we ought to do, all things considered, and various facts about what we actually do (and are disposed to do). This is the familiar gap between ought and is. And my point is that this gap, which is induced by the irreducibly normative character of our commitments, ought not to be confused with the gap between higher and lower order desires. This is not to say that there cannot be discords between our higher order volitions and our actions (and attendant first order desires), for there obviously can be, as my own real example attests. Nevertheless, from a normative point of view, there is no guarantee that, when we deliberate and inquire into matters of value, the objects of our higher order volitions should necessarily always emerge to be better than the objects of our first order desires; and when they do not, we may have reason to criticize ourselves for having certain higher order volitions to begin with, rather than for failing to have the effective first order desires that would accord with them—just as my case illustrates.Footnote 6

Perhaps the foregoing argument will not convince everyone that it is a doomed project to try to reduce normative commitments to higher order volitions. Rather than spend more energy trying to convince the unconvinced let me underscore this: a crucial criterion for the success of this particular reductive project—which I think is, au fond, a project of naturalist reduction—is that it should be able to do justice to all of the phenomena that I shall now proceed to describe in irreducibly normative terms.

To return to my main point: To say that an agent is not a metaphysical given is to say, in part, that the normative commitments that characterize individual agency are not a metaphysical given. All that is metaphysically given is a human being, with a body and a brain and a center of consciousness, along with a capacity for agency—which I take to include a capacity for embracing commitments, and for deliberating, and for acting from such commitments.

This point deserves further elaboration in connection with the idea of the first person point of view from which an agent deliberates and acts. According to the reductionist account of agency, such a deliberative point of view is not a metaphysical given, any more than the agent who possesses such a point of view is.

As I have already explained: an agent requires a first person point of view from which to deliberate; and the aim of deliberation is to arrive at all-things-considered judgments about what it would be best to do in the light of all that it thinks; and to arrive at and act upon such judgments is to achieve the sort of rational unity that defines what it is for an individual agent to be fully or ideally rational. What sets the boundaries of such a deliberative point of view also sets the scope of the “all” in all-things-considered judgments. Thus, to say that I must deliberate from my own point of view, is to say that I must take into account all and only my own commitments when I deliberate, for it is their joint normative significance that my all-things-considered judgments would reflect. If the existence of an agent were a metaphysical given, then the boundaries of the first person point of view from which it deliberates—which is the same thing as the scope of the all in its all things-considered judgments—would be set by the metaphysical condition of its identity. Suppose that this were so. Suppose, more specifically, that the boundaries of an agent’s deliberative point of view were set by the biologically given boundaries of a single human life. Then it would follow that the human being ought to achieve rational unity within its biological life. Yet although the human capacity for rational agency can be exercised so as to achieve rational unity within the boundaries of individual human lives, that capacity can also be exercised so as to achieve rational unity within parts of human lives and within groups of human lives as well. In each case, this is accomplished through an active embrace of various considerations together, as a common basis of deliberation and action; and it is through this active embrace that such considerations come to constitute a deliberative point of view and, therewith, the individual agent whose point of view it is.

There must always be reasons for embracing together the many considerations that constitute a single deliberative point of view. These reasons are given by unifying projects whose pursuit requires significant coordination of thought and effort within some boundary or other—whole human lives, or groups of them or parts of them. These different sorts of unifying projects would then give rise to agents of different sizes—agents of human size, or group agents comprising many human lives, or multiple agents within a single human life. Obviously, the sorts of projects that agents of human size can pursue within a whole human life are very different from the projects that group agents can pursue through many human lives, and they are also very different from other projects that can be pursued by multiple agents within parts of human lives. It is the nature, feasibility and value of such unifying projects—how much rational unity they call for, within or across which human lives, in order to achieve what ends, with what merit—that determines which agents can and should exist. And it is only through the actual embrace and pursuit of such projects, along with whatever rational unity their pursuit requires, that particular agents with particular deliberative points of view come into existence. In short, the individual human being does not qualify as an individual agent unless and until there is an active pursuit of a unifying project that requires rational unity within its whole human life. This gives the basic sense in which the existence of an agent of human size is not metaphysically given with the existence of an individual human being.

So far I have been emphasizing that if there is to be an individual agent who thinks and acts as one, then there must be intentional episodes (where these intentional episodes are to be normatively construed as involving commitments and are not to be construed as mere causal dispositions in the way that functionalists do) that recognize one another, reflect one another, collect one another, track one another, and respond to one another. That is how various commitments come to figure together in a single deliberative point of view. Let me now elaborate these points more specifically in connection with the form of reductionism for which I am arguing.

According to reductionism, there is no further thing, beyond thinkings and doings themselves, that constitute the point of view of an agent. Even the overarching commitment to achieving rational unity among those thinkings and doings, without which they would not together constitute the single deliberative point of view of an individual agent, is a feature of the intentional episodes themselves. That overarching commitment emerges as a feature of how the episodes themselves interrelate with one another, as they seek to discover their joint normative significance for the question, what would it be best to do in the light of them all? But if they are seeking to discover this, it is only because they recognize themselves, and one another, as somehow bearing on a common unifying project that requires unification among them. This is what it is for them to figure as a common basis of the deliberations of a single agent with its own point of view.

This is not the place to respond at any very great length to various sources of resistance to this reductionist view of agency. But I can at least indicate a general strategy of response. Suppose it is argued that we need to posit some further metaphysical condition for the existence of an individual agent, over and above the existence of intentional episodes standing in the right sorts of relations—by arguing, for example, that an agent must be a member of a natural kind, or possess an immaterial soul, or have a ‘self’ in some other sense that I must confess I have never found very clear. Then the question will arise: Does this metaphysical posit help us to mitigate any of the philosophical perplexities that the notion of agency induces, when we contemplate such matters as freedom and normativity? For my part I have never found that my own philosophical perplexities in connection with these matters are mitigated in any way by supposing that a rational agent is something over and above intentional episodes, normatively construed, standing in the right sorts of relations. But what is more to the point: no such further metaphysical condition can provide for the existence of an agent, unless it provides for a commitment to achieving the sort of rational unity that is characteristic of individual agency; and it cannot be the site of that commitment unless it is the site of intentional episodes through which this commitment’s content is understood and responded to; this is impossible without a recognition of the scope of the commitment; and this in turn is impossible without a recognition of which intentional episodes belong together as the episodes among which rational unity is to be achieved; and this in turn is impossible unless the intentional episodes themselves incorporate this recognition as part of their contents, for it is only through intentional episodes that any such imagined further-thing-that-is-supposed-to-be-the-agent could possibly conceive and implement its commitment to overall rational unity. At the very least, these reflections show that no further metaphysical condition, beyond the existence of intentional episodes standing in the right relations, suffices for the existence of an individual agent with its own deliberative point of view. From the reductionist point of view, these reflections also show that such metaphysical conditions are not necessary either, for to posit them in our account of agency is to posit an idle wheel. All the work of the account is done by clarifying how it is that intentional episodes can stand in the right relations, by virtue of which they together constitute the individual point of view, and indeed the intentional life, of an individual agent—and this emerges to be the condition of individual agency, when we view an agent as something that deliberates and acts on reasons in the irreducibly normative sense.

To sum up: It follows from the reductionist account of agency that a human being can fail to be the site of an individual agent, because the requisite unifying intentional activities and commitments that I have described can be missing—as they are missing in the life of the wanton. In the course of normal psychological development, each human being gradually leaves its initial condition of wantonness and becomes a site of rational agency. This process generally begins with commitments to projects that require rational unity within boundaries much smaller than a whole human life. In our cultural moment, most mature human beings come to embrace larger unifying projects, which are often called “life projects”. But there is no necessity about having to take on the sorts of unifying projects for the sake of which rational unity—and hence individuality—would have to be achieved over time within the course of a whole human life. It is perfectly within human capability to achieve rational unity within different portions of a human life instead. And ultimately, it follows from reductionism that any metaphysical condition can be the site of individual agency so long as the requisite unifying intentional activities and commitments can take place, that give rise to a commitment to achieving rational unity within a body of intentional episodes, where the scope of that commitment determines the boundaries of a particular deliberative point of view from which deliberation and action may then proceed. The argument for the possibility of group agency says that these conditions of individual agency can be met within a group of human lives, so as to give rise to a group agent.

2 Why group agency is not necessarily a social phenomenon

It should already be clear that the reductionist view of agency gives us no reason to presume that group agency must be a social phenomenon. The reductionist claims that it is always the same thing that is going on when a deliberative point of view is forged through intentional activities, regardless of where its boundaries fall with respect to human beings; therefore, since we do not think of a deliberative point of view as a social phenomenon when it falls one-to-one with respect to human beings, there is no reason why we must think of it as a social phenomenon when it happens to fall one-to-many with respect to human beings. The situation would be different if we were using “social” in the biologists’ sense, on which a phenomenon would count as social simply by virtue of involving interactions among many distinct organisms. But as I clarified at the very beginning of the paper, when I ask whether group agency either must be or can be a social phenomenon, I mean “social” in the sense of interest to action theorists, social ontologists, etc., on which social relations arise among individual agents who mutually recognize one another as such, and who therefore recognize the social relations in which they stand as such.

I will approach the task of showing that group agency is not necessarily social in this specific sense by coming at matters from the other side—by assuming for the sake of argument that it is necessarily a social phenomenon, and then considering what would follow.

Among other things, it would follow that the intentional activities through which a group point of view is forged leave the human constituents of the group intact as distinct agents with their own separate points of view. It would also follow that the intentional activities through which a group point of view is formed are really social activities, in the sense that they are really the intentional activities of many distinct agents of human size, each of whom is deliberating and acting from its separate point of view even as they together forge rational unity at the level of the whole group. Finally, it would follow that group agency is a special case of collective agency—it would be the agency of many distinct agents acting together, where this is supposed to constitute the agency of onegroup agent at the same time.Footnote 7

So let me now re-examine these implications through the lens of reductionism. If the human constituents of a group agent were to qualify as distinct agents who have their own separate deliberative points of view, they would have to separately embrace and pursue unifying projects that require rational unity within the entirety of their separate biological lives. But the group point of view, and hence the group agent itself, would not exist unless some of the intentional activities that occur within the biological lives of the group agent’s human constituents were carried out from the group point of view; and so, even if all of the rest of the intentional activities that occur within the group’s human constituents were carried out from separate points of view, those separate points of view would typically be somewhat smaller than human size. Another way to put this point is that the human constituents of group agents are typically sites of rational fragmentation.

It may seem natural to suppose that any sort of rational fragmentation within a single human life must be a rational failing of some kind. But there is no rational failing in the condition I just described above. Commitments to rational unity are very much in place in this condition—it’s just that their scope happens not to coincide with the boundaries of a whole human life so as to range over all and only the intentional episodes within that human life. I have already argued that any such commitment to rational unity follows upon commitments to unifying projects, which set the boundaries within which such rational unity must be achieved. In situations where group agency arises, there will typically be many such commitments in place—some unifying projects will set the scope of the deliberations of a group agent, while other unifying projects will set the scope of the deliberations of other agents who are smaller than human size. This point is of a piece with the points I made in the last section, about the possibility that there can be multiple agents within a single human life. When such multiple agents exist within a single human life, that life is also marked by rational fragmentation. There is no reason to suppose that the resulting rational fragmentation must be a rational failing in those cases either, because human lives may be marked by such rational fragmentation and yet remain sites of perfectly rational pursuit of ends—albeit, by multiple agents of different sizes.Footnote 8

Let me now offer an illustrative example in order to help clarify the point that group agency need not be a social phenomenon. Consider a philosophy department that recognizes reasons to deliberate and act as one, for the sake of arriving at a reasoned and coherent Ph.D. curriculum. We would normally describe a department as constituted by its human ‘members’, and so to start with I shall use this terminology. But we are about see that the terminology can be misleading, precisely because when we refer to the ‘members’ of a group agent such as a philosophy department, we normally presume that they are, and indeed must be, agents of human size who then socially constitute the group agent. Once I have shown how and why this presumption is mistaken, I will no longer use the terminology of ‘members’, and refer instead to the human constituents of group agents, or more simply, to human beings—making no presumption that human beings are, or must be, agents of human size. But let me return to our more normal ways of speaking for the moment, as I elaborate my illustrative example. Suppose that the ‘members’ of the department recognize that they cannot simply argue and vote, because then, if students were to ask why they must satisfy certain degree requirements, the only available response would be: Well that’s the way the vote cut. What a responsible department might therefore want to do instead, is to work out what degree requirements would be best, all things considered. But then it must identify the things-to-be-considered. These things cannot be confined to thoughts that figure in the biological life of one or another human ‘member’ of the department, but must include all of the thoughts that figure in the biological lives of all of the department’s human ‘members’, that bear on the deliberative question before it. These thoughts, which are scattered across distinct human beings, constitute the proper basis of the department’s deliberation. In the course of this group deliberation, what is to be worked out is the joint normative significance of all of those thoughts, for the question what would it be best for the department to do in the light of them. In working this out, the department forges a group point of view, from which there are real answers to the question, when it is addressed to the department as a whole, why did you impose the requirements that you did? The answers will not reflect the reasoning of any one human ‘member’ of the department, but rather the reasoning of the department as a whole. In order for this reasoning to take place, each of the human ‘members’ will be a site of intentional activities—thinkings and doings—that are carried out from the group point of view. Insofar as each human ‘member’ remains the site of a separate point of view from which non-departmental projects are also pursued, these separate points of view will not include the thoughts that figure in the life of the department, but will only include the various other thoughts that bear on these smaller, separate, pursuits. This is the ‘rational fragmentation’ to which I referred above.

To persist in viewing group agency as necessarily a social phenomenon simply on account of the fact that it involves interactions among many human beings would be highly misleading, therefore. It would give the false impression that the intentional activities—the thinkings and doings—that constitute the group agent’s life must be carried out from separate points of view of human size, whereas I’ve just shown that this need not be so, because the way in which the group agent’s point of view is forged need not leave intact separate points of view of human size. In other words, it would give the false impression that group agency is always, necessarily, a special case of collective agency—the agency of many —whereas it is really the agency of only one.

There can be no doubt, however, that we are highly susceptible to this false impression. One reason why is that we tend to conflate the idea of an agent’s deliberative point of view with two other kinds of point of view: the phenomenological point of view from which a human being enjoys direct conscious access to thoughts, and the bodily point of view from which a human being perceives and moves. Of course, the fact that human beings have separate phenomenological and bodily points of view makes some difference to the processes through which rational unity gets achieved within individual human beings vs. within groups of human beings. But these differences do not threaten the status of a group agent as a genuinely individual agent; and nor do they threaten the status of the relations among intentional episodes through which its rational unity is achieved, as genuinely intra-personal relations.

Take, for example, the fact that a group deliberation requires linguistic communication whereas deliberation within the separate consciousness of a single human being does not. Regardless of whether rational unity is achieved through spoken language, or through silent means, it always involves the same kind of normatively driven effort—which is always an effort to achieve the kind of rational unity that is characteristic of individual rationality, by working out the all-things-considered significance of a body of deliberative considerations. According to the reductionist account of agency, it is sufficient, and also necessary, for the existence of an individual agent with its own point of view, that there be a commitment to precisely this kind of effort within the boundaries of that point of view. So if this commitment should arise within a group of human beings, as we may find in the case of a philosophy department, this suffices to make it a single group agent—and it matters not that the group agent’s efforts to live up to the commitment will require linguistic communication across distinct centers of consciousness. Correlatively, if this commitment should be lacking within the life of a single human being, then that human being will not qualify as the site of an individual agent of human size—and it matters not that the human being is the site of a single center of consciousness. The latter failure may be due to a total absence of rationality, as we find in cases of extreme wantonness or madness; or it may be due to a segregation of rational effort, as we find when a human being is the site of multiple agents with separate points of view; or it may be due to a similar segregation of rational effort, as we find when a human being is a site of rational activity that is carried out from a group agent’s point of view, which exists alongside another point of view, somewhat-smaller-than-human-size, within that same human being.

Take, for another example, the fact that a human being exerts direct motor control over its body, which seems to have no counterpart in group agency. This capacity, though significant in many ways, is not what enables an agent of human size to think and act as one. What affords such oneness is the capacity to coordinate its thoughts and actions so as to achieve overall rational unity within itself. Thus it is far less significant that it can think, “I will raise my arm now” and then happily find that its arm does indeed go up on cue; it is far more significant that it can think “I will give a paper at next January’s meeting of the American Philosophical Association” in the anticipation that its intention will be remembered when the time comes, and then happily finds that it does indeed remember when the time comes, and that it also still has the requisite commitment to unity in the light of which remembering the intention counts as a reason to implement it. It is through such reflectively mediated coordination of thought and action that individual agents achieve the rational unity that is characteristic of the individual.

These last remarks help to flesh out my claim in the last section, that the existence of an agent is always a product of effort and will. If a human being’s life is to be the site of a single, abiding agent, this will require ongoing,voluntary, and reflectively mediated coordination of thought and action over time within that human being’s life, so as to achieve rational unity within it. For the most part, this will not be achieved through direct forms of causal control, akin to direct motor control, that somehow reach across time without the help of deliberate effort. If it were, then our aim in forming long-term intentions would be like Oddyseus’s aim when he tied himself to the mast in order to resist the Sirens’ call. In contrast, the way in which rational agents generally exercise their agency is through an ongoing, voluntary, and reflectively mediated coordination of thought and action.

These last remarks also provide a fresh angle on my claim that the sort of rational unity that characterizes rational agency is always the same thing, no matter how it falls with respect to human bodies. It is because it always requires a voluntary coordination of thought and effort for the sake of achieving overall rational unity—this is just as true when it is achieved within a single human life so as to constitute an agent of human size, as when it is achieved within a group of human lives so as to constitute a single group agent, or within parts of a human life so as to constitute multiple agents within it.

It should now be clear that, if the reductionist account of agency is correct, and if the existence of an agent is, as the account says, a product of effort and will, and if this comes to pass through the active embrace and pursuit of a unifying project that requires the kind of rational unity that characterizes individual agency, and if the intentional activities through which such rational unity is achieved are always the same in structure, regardless of whether they are carried out within parts of human lives, or whole human lives, or groups of human lives, then there is no metaphysical ground on which to deny that group agents are bona fide individuals in their own rights. Likewise, there is no metaphysical ground on which to insist that their human constituents must be sites of individual agents of human size, given that they may be sites of rational fragmentation instead—so that some of the intentional episodes and rational activities that figure within a given human life may proceed from the point of view of an agent of smaller than human size, while others proceed from the point of view of a much larger group agent that comprises some intentional episodes and rational activities that figure in other human lives as well. Since the human constituents of group agents can be sites of rational fragmentation in this way, group agency need not be a social phenomenon.

Of course, we do not often find group agency described along these lines. Almost any philosopher who is prepared to discuss the phenomenon at all will simply take it for granted that the human constituents of a group agent are not sites of rational fragmentation but are rather sites of overall rational unity, and that they therefore remain sites of individual agency (agents of human size) even when they are also sites of group agency. Indeed, I think it is fair to say that most philosophers who discuss group agency are prone to conceive the intentional episodes that constitute the life of the group agent as thinkings and doings on the part of individual agents of human size who are thereby making individual contributions to the life of the group agent. To conceive matters in this way is to conceive group agency as a social phenomenon. In the next section I will explore whether this is so much as possible.

3 How group agency might be a social phenomenon: Rousseau’s general will

By the lights of the reductionist account of agency, here is what must happen if group agency is to qualify as a social phenomenon: The human constituents of the group agent must be sites of individual agents of human size; these individual agents must carry out rational activities that are directed at achieving overall rational unity within their own points of view; at the same time, the very same rational activities must aim to have rational effects on the points of view of other agents of human size, by appealing to their internal commitments to achieving rational unity within themselves; these individual agents of human size must have the further aim of achieving overall rational unity at the level of the whole group that includes them all, through their effects on one another.

Much of the philosophical literature on group agency concerns itself with such matters as characterizing the sorts of intentions that individual agents of human size must frame in order to pursue group aims together, and whether such intentions can be well formed given that they are directed at social outcomes that are not under the control of the individual agents who form them. The question that concerns me in this paper is quite different: Can the rational unity of a group agent be constituted in the way I just described in the last paragraph, so that it does not occasion rational fragmentation within the lives of its human constituents, but preserves their individual rational unity? For that is the only condition in which the rational relations and activities that constitute the life of the group agent will qualify as social relations and activities among distinct agents of human size, even as they are supposed, by hypothesis, to be intra-personal relations and activities of the sort that characterize individual deliberations of the group agent.

Many philosophers who write about political rationality, or social rationality more generally, take it for granted that this condition can and should be realized. But my arguments of the last section should give us pause, for they show rather more than I have so far claimed on their behalf. So far, I have only claimed that they show that group agency is not necessarily a social phenomenon. But I think they also give us reason to think that group agency would not typically be a social phenomenon. Why? Because when a group agent like a philosophy department deliberates from its point of view, its deliberations do not proceed from the greater part of the intentional episodes that figure in the lives of its human constituents. Again, why? Because the group agent will deliberate only from what bears on the pursuit of its ends, and many of the thoughts that emerge within the confines of the biological life-spans of its human constituents simply do not bear on the pursuit of the group agent’s ends. Correlatively, insofar as the human constituents of a group agent are sites of agency that is directed towards non-group pursuits, such as marriages, careers and other personal projects, the agents who engage in these pursuits will deliberate only from what bears on their pursuit of their ends—which is to say, they will not deliberate from many of the thoughts from which the group agent deliberates. That is why I describe these agents as being smaller than human size, and why I describe the human constituents of group agents as sites of rational fragmentation. These human beings are sites of rational activities, some of which proceed from the points of view of agents of smaller than human size, and some of which proceed from the larger point of view of a group agent.

I want next to explain how Rousseau’s account of the general will provides a good model for how group agency can be a social phenemonon, even by the lights of the reductionist account of agency. Then I will go on to explain why I think any case in which group agency counts as a social phenomenon by the lights of the reductionist account of agency will share crucial features with the general will.

Rousseau took for granted that human beings are sites of individual agency, or rather that they just are individual agents. And he took for granted the standard liberal view according to which, if individual agents are to be free, then it must be the case that they are subject only to their own wills and to no one else’s. Accordingly, the problem he set himself in The Social Contract was to work out how individuals of human size can be subject to a political sovereign and still be free. His strategy for a solving this problem was to seek a way in which, when individual agents are subject to the will of the political sovereign, they are not really subject to anyone else’s will but their own.

Obviously, Rousseau intended this to be a contribution to the liberal tradition that had already made the so-called consent of the governed central to legitimate political sovereignty. But he understood two problems much better than other liberals of his own time, and I daresay ours as well. First, consent is too easily nominally gained, and not at all legitimizing, in situations of significant inequalities of power. Second, any political will to which citizens are subject must be one with which they are wholly identified, or else its exercise will amount to a form of domination by a greater (i.e., unequal) power. He introduced his conception of the general will precisely in order to solve these problems. Specifically, he aimed to show how it is possible that genuinely individual agents, with their own separate points of view and individual wills, could together forge a common will to which they would all be subject, but without being illegitimately dominated, either by one another or by the group as a whole. I think it is fair to say that if he succeeded in demonstrating the possibility of such a general will, then he also succeeded in showing how group agency can be a social phenomenon. For on his description, the general will would be forged through inter-personal relations among social contractors, and yet these inter-personal relations would also, at the same time, qualify as intra-personal relations within a single group agent’s point of view—an agent he generally referred to as the sovereign, and which he also explicitly described as a self.

A few words are in order about what I’ve been calling “rational fragmentation” as it relates to Rousseau’s views. Rousseau was wonderfully sensitive to the distinction between the general will that would be forged through the social contract and the individual wills of the contractors themselves, sometimes distinguishing these two with the words “public” and “private”. In some of his descriptions, it almost seems as if citizens must have two hats, a public one that they wear for the purposes of the political deliberations that they engage in with others about their common ends as they forge a general will, and the private one that they wear when they are engaged in their own personal pursuits. These descriptions might seem to suggest that his political case would involve rational fragmentation within human lives in much the way I have claimed typically occurs in cases of group agency—in other words, they suggest that the unity of the political community’s general will would come at the cost of thoroughgoing rational unity within the individual lives of its human ‘members’. By the lights of the reductionist account of agency, it would follow that these human ‘members’ would not, strictly speaking, be agents of human size, but rather each of them would be a site of two instances of agency—a private citizen who is somewhat smaller than human size, and (a part of) the sovereign.

But it would hardly be a sympathetic, or even comprehending, reading of The Social Contract, to suppose that the human ‘members’ of a political community might fail to qualify as individual agents of human size. A sympathetic and comprehending reading would emphasize that, in Rousseau’s view, the whole purpose of legitimate political agency is to secure the individual liberty of individual citizens conceived precisely as agents of human size. From a reductionist point of view, this aim would have to go hand in hand with the aim of perpetuating overall rational unity within each individual human life. And I think Rousseau would agree. Hence his somewhat flamboyant talk about how each contractor must give his or her self up entirely to the whole. His idea was not that the individual viewpoints of these human size agents would be entirely absorbed into a larger group will when they enter into the social contract, so that they no longer existed as separate agents with separate points of view. And nor was his idea that each should become a site of rational fragmentation. Rousseau’s idea was that each contractor should be able to embrace a commitment to achieving overall rational unity within the political community, as required by the community’s common ends, while at the same time retaining a commitment to achieving overall rational unity within its own life, by which I mean, a particular human life.

Here is how this might be accomplished (and I do think it is what Rousseau had in mind): on the one hand, the very reason for being of the political sovereign is to perpetuate the individual life and liberty of its citizens, each of whom is an agent of human size; on the other hand, each citizen wholeheartedly believes that the best way to pursue its own ends is by entering into and abiding by the social contract through which the sovereign will of the political community is formed. Thus, because the very purpose for the sake of which the political sovereign exists is to secure the life and liberty of individual citizens, it exists only as a means to their individual ends. This is why entering into a social contract is not an occasion for rational fragmentation within human lives. There is no occasion for conflict, or even any significant divergence, between the general will of the political sovereign and the individual wills of the agents who comprise it, because the former is wholly in the service of the latter. So if such a sovereign were to exhibit the sort of rational unity that is characteristic of individual agency, it would appear to be a case in which group agency was indeed a social phemomenon, or in other words, a case of collective agency—it would be the will of one which was also, at the very same time, the will of many.

I will not go so far as to say that I know, with certainty, that the general will as Rousseau conceived it is a real possibility. Or, even if it is a real possibility, I will not go so far as to say that it would be the will of a group agent, for it might be better conceived as a set of abstract principles that citizens might all have reason to embrace—much as Rawls portrayed his principles of justice. If that is how it should be conceived, then it wouldn’t strictly speaking count as a case of group agency at all, and so it would not vindicate the claim that group agency can be a social phenomenon.

However, let us suppose for the sake of argument that Rousseau’s general will is a real possibility, and that it really would be the will of a group agent. It does seem to me that this would be a case in which the group agent would count as a social phenomenon even by the lights of the reductionist account of agency. For the pursuits of this group agent would not engender rational fragmentation within its human constituents, owing to the coincidence, nay, the very identity, of the group agent’s goals with the goals of individual agents of human size who exist within it, and who form it, and who carry out its deliberations and actions, for the sake of their individual ends.

Let us next consider, by contrast, any other unifying project for the sake of which a group agent might be formed, besides serving the ends of individual agents of human size, who together constitute the group agent expressly for the sake of all of their respective individual ends. Take, for example, the project of answering scientific questions, or developing technologies, or making music, or building buildings, etc. Projects like these might be undertaken and pursued for their own sakes, simply because they are worth doing. If they were, then the group agents who pursued them would not exist for the sake of other agents of human size, nor would they exist as mere means to their ends. These group agents would exist for the sake of their own ends, i.e., their unifying projects; and the rational unity that their unifying projects call for across human lives would typically come at the cost of rational unity within them, in the way that we saw in my example about the Philosophy Department. The human constituents of such group agents can of course be sites of significant rational unity that is achieved for the sake of other, smaller, unifying projects—projects that are pursued by agents somewhat smaller than human size, who exist alongside the group agent. But the important point to appreciate is that the unifying projects of these smaller agents would not figure among the ends of the group agent itself, and nor would the group agent’s unifying project figure among the ends of these smaller agents. That is why the human constituents of the group agent will be sites of rational fragmentation. Once we recognize the source of this rational fragmentation, we can easily see why group agency would not typically be a social phenomenon.

The agents who I claim would typically exist side by side in these situations in which human lives are rationally fragmented—group agents and agents somewhat smaller than size—may themselves stand in all sorts of interesting social relations. They are bound to know about one another; and they are likely to have positive evaluative attitudes towards one another as well. They may welcome the ends for the sake of which their companions exist, and they may even see their respective ends and activities as mutually reinforcing. Indeed, this would likely hold in my own example about the Philosophy Department, whose project of running a Ph.D. Program would naturally harmonize with, and be reinforced by, the individual projects of teaching and research that are carried out by its ‘members’—who I am claiming are not agents of fully human size but are smaller than that. But such relations—approval, harmony, mutual reinforcement, and the like—among the ends of distinct agents does nothing to threaten their distinctness. The same is true even when agents go so far as to give aid to one another. To take action to promote someone else’s ends does not generally involve appropriating their ends as ends of one’s own—for it is one thing to reason from one’s own point of view that one has reason to help another to do X,  and it is quite another thing to reason from one’s own point of view that one has reason to do Xoneself. Think, for example, of how a parent might have reason to help their child to master ballet, without thereby becoming someone who has reason to master ballet. The line I’m pushing in this paper is that this is a good way to think about some of the relations that hold between certain group agents and certain agents smaller than human size who co-exist within the same human lives that are the sites of the group agent’s intentional activities. For just as what the child is doing, and even can do, is quite different from what the parent does and can do, likewise, what the group agent is doing, and even can do, is quite different from what a smaller agent of less than human size can do. The latter does not become someone who is deliberating and acting from the group point of view just by virtue of being aware of the group agent’s deliberations and actions; and the distinction between the two is not collapsed or threatened if each should welcome the projects of the other, or if each should regard their respective projects as harmonious and reinforcing, or even if each should do things to promote the others’ ends.

I have taken the trouble to clarify these points about the social relations that would naturally arise in cases where, according to my arguments, group agency is not a social phenomenon, in order to clarify how difficult it is to envisage a case in which group agency is a social phenomenon. In order to envisage this, we must find a way to model relations that are at once social, because they hold among distinct agents of human size who deliberate and act from their own points of view, and yet also intra-personal, because they constitute the life of a group agent who deliberates and acts from a single group point of view. Unlike the cases I just described above, these would be cases in which the human constituents of the group agent are all agents of human size, and each actually embraces the ends of group agent as its own—so that when it deliberates and acts from its own point of view it is (among other things) deliberating from and acting for the ends of the group agent.

As far as I can see, the case of Rousseau’s general will is the only clear case in which a group agent would satisfy this condition. Because the group agent exists for the sake of ends that are separately embraced by the individual agents of human size who together form it, the internal unity of those individuals will not be compromised by the social activities through which they also forge the unity of the group agent, and carry out its deliberations and actions. From a first person point of view, each can think of its contributions to the life of the group agent as a means to its own ends.

The question arises whether there is a middle ground that lies between the two cases of group agency that I have been discussing. On the one side are cases of group agency that definitely are not social phenomena, because the group agent’s unifying project is pursued from its own point of view and not from the points of view of other agents, smaller than human size, who co-exist alongside the group agent, and who pursue their own unifying projects independently of the group agent’s pursuits; on the other side are cases of group agency that would qualify as social phenomena because they share the crucial feature that distinguishes Rousseau’s general will—in these latter cases, although there is a sense in which the group agent’s project is pursued from a single group point of view, the group project is nevertheless embraced and pursued by many distinct agents as means to their individual ends. What, then, could occupy a middle ground between these two cases? I myself do not see anything that could fit the bill. If the group project is to serve the separate ends of individual agents who embrace and pursue it from their separate points of view, it is Rousseau’s case. If the group agent has some other purpose, then it is being pursued for its own sake. If we posit other agents, smaller than group size, who have their own projects, we can easily imagine that their personal projects stand in all sorts of interesting relations to the group project—compatibility, harmony, reinforcement, and even mutual aid. But as I just explained above, this will not suffice to make it the case that the group project is literally being embraced and pursued by smaller agents of human size, from their own separate points of view, for their own individual ends. On the contrary, the pursuit of the group project would require another agent, whose life is not constituted by the thinkings and doings that constitute the lives of smaller agents of human size. Precisely insofar as the life the group agent is not constituted in this way, it is not a social phenomenon. And to reiterate, I see no other way in which the life of the group agent can be socially constituted, except insofar as it shares the features of Rousseau’s general will: each participating agent must embrace and pursue the group agent’s end from its own point of view, as a means to its own ends.

However, it is important to see that, even if I were overreaching in this last claim, the other claims for which I have argued would all still stand. They are: group agency is not necessarily a social phenomenon owing to the fact that the it may be an occasion for rational fragmentation within the lives of its human constituents; it is nevertheless possible for group agency to be a social phenomenon, if something with the structure and point of Rousseau’s general will is possible; Rousseau’s general will is not a good model on which to understand many of the cases of group agency that philosophers who are interested in the phenomenon often have in mind, such as teams of scientists, sports teams, orchestras, etc.; we ought not to presume that these latter are social phenomena—for insofar as the condition of individual agency really is realized at the level of a group, in such a way as to forge a group point of view which is distinct from all other points of view, we should not expect that there is any sense in which its projects are being pursued by a plurality of agents of human size, rather than simply by it.

4 Some tendencies in the philosophical literature on group agency

In spite of all that I have said in this paper, it may be difficult to shake off the mistaken belief that group agency either must be a social phenomenon, or typically would be. In this final section of the paper, I’d like to address this mistaken belief, insofar as it appears in the work of some fellow travellers, all of whom acknowledge the possibility of group agency.

There is a great deal of similarity between the account of individual agency I have recommended here and the account that Christine Korsgaard has developed and defended in the course of her career.Footnote 9 We both regard agency as irreducibly normative in a sense that can only be made sense of from a first person point of view—what she refers to as the “deliberative standpoint of agency”; we both hold that it is characteristic, and indeed constitutive, of individual rationality to be committed to achieving overall rational unity within oneself; we both regard the existence of an individual agent as a product of effort and will; and we both allow that the condition of rational unity that characterizes individual agency can be realized within groups of human beings. Yet there are some important differences. Her earliest published argument in favor of group agency (that I know of) appeared in the context of her very impressive critical discussion of Derek Parfit’s psychological reductionist account of personal identity, according to which a person is “nothing but” a series of events standing in the right sorts of psychological relations.Footnote 10 She rightly pointed out there that Parfit had left entirely out of account the fact that persons are agents. As a result, her sustained attack on his reductionism leaves an impression hanging in the air that she is against reductionism in general—and that would commit her to rejecting the particular variety of reductionism about agency that I have defended here. Yet it seems to me that her own constructivist account of agency is implicitly reductionist.

I shall have to leave it for another time to sort out this possible point of disagreement with Korsgaard, in order to focus on another point of disagreement that matters much more for my purposes in this paper. While her constructivism does entail that individual agents come to exist only insofar as the human capacity for agency is actually exercised, she nevertheless insists that whenever that capacity is actually exercised, it is bound to issue in agents of human size. In her view, there is a kind of practical necessity involved here, by virtue of which each individual human being must be a site of individual agency if it is to be a site of agency at all. If she were right about this, then I would have to withdraw my main arguments in this paper—I would have to allow not only that group agency can be a social phenomenon, but also that it must be.

Korsgaard makes two related claims in support of her view that the individual human being must be a site of individual agency. The first is that a human being must be unified at a time because it must act, and it has only one body with which to act. The second is that when different psychological functions are realized within the same body, they must become unified in such a way as to constitute a unified agent. These claims are both mistaken in my view—committing mistakes of exaggeration and conflation. What is true is that any human being is bound to be a site of significant psychological unity; but this may fall short of the sort of the normative commitment to rational unity that is characteristic of individual agency, as I shall now argue.

It is true that any normal human being will exhibit a high degree of perceptual-motor organization by virtue of which it is able to locate itself within its environment and move in relation to other objects; and it is equally true that this perceptual-motor organization must be supported by significant unity in brain function, through which the human being is able to keep track of its bodily movements. In his paper “Brain Bi-Section and the Unity of Consciousness” Thomas Nagel explores whether these natural forms of practical unity are disrupted when a human being undergoes cerebral commissurotomy—a procedure in which the human being’s corpus callosum is severed, so that the two hemispheres of its brain no longer directly communicate with each other.Footnote 11 Since different sides of the body are controlled in different ways by the two hemispheres of the brain, it might naturally be expected that a human being who has undergone this procedure would display a significant lack of perceptual-motor coordination. But in fact, it does not take very long after the procedure before such a human being regains its normal perceptual-motor functioning. It is only in highly artificial conditions, in which sensory inputs to the left and right sides of the visual field are carefully orchestrated and segregated, that such a human being ever displays significant lack of perceptual-motor coordination. And yet, as Nagel notes, even in these highly artificial conditions, the human being will still make an effort to achieve whatever degree of perceptual-motor coordination remains possible. On the basis of Nagel’s description, we might fairly conclude that this ‘effort’ is involuntary in the following sense: the effort is made automatically, without giving any thought to whether it is worth making, and moreover, there doesn’t seem to be any recognition that it might be an option not to make the effort. Prima facie, it might seem that this conclusion brings us fairly close to Korsgaard’s claim that there is a kind of necessity about the way in which different psychological functions within a single human being become unified. But whatever necessity there may be for a human being to achieve these forms of psychological unity that afford perceptual-motor coordination, there are many reasons to doubt that such unity will necessarily provide for, or require, the sort of rational unity that characterizes individual agency. For one thing, these forms of psychological unity are present in other species of animals that cannot achieve such rational unity, because they lack the requisite rational capacities. For another thing, the mere fact that a human being possesses such rational capacities does not ensure that it will exercise them, for it could instead function as a wanton. But finally, and this is the most important thing of all: Even if it were true that a human being could never be a wanton, because it could never avoid exercising its rational capacities (something Korsgaard seems to believe, but which I myself doubt), this still would not ensure that the human being must be a site of individualrational agency. That would be so only if the human being was bound to exercise its rational capacities in such a way as to achieve rational unity within the boundaries of its biologically given life. The burden of my arguments throughout this paper is that the boundaries within which rational unity is to be achieved can be drawn more narrowly or more widely than that. These arguments cannot be rebutted simply by pointing to the fact that human beings are sites of various forms of involuntary psychological unity that fall short of rational unity. In fact, my arguments show that we have no reason to accept Korsgaard’s other, much weaker, claim, which is that a human being must at least achieve rational unity at a time, if not within the whole of its life, because it has only one body with which to act. For rational action requires a deliberative point of view from which to act, and a single human being need not be the site of a single deliberative point of view, even at a time. To see that this is so, recall the sort of rational fragmentation that follows upon participation in many kinds of group agency, such as a philosophy department. It simply isn’t the case that all of the thoughts that occur at any given time within the brain of a single human constituent of the department must constitute a whole, single, deliberative point of view. Some of those thoughts will figure in the department’s point of view, while others will figure in a point of view somewhat smaller than human size. So there really is nothing, as far as I can see, to support Korsgaard’s contention that human beings must be sites of individual agents of human size. I conclude that her work does not give us any reason to suppose that group agency must be a social phenomenon.

The common attitude among philosophers who write about group agency is that Korsgaard is right: individual human beings are, necessarily, sites of individual agency, and group agency must therefore be a social phenomenon that arises through social interactions among individual agents of human size. It seems to me that any philosopher who does not explicitly disavow the common attitude, should expect to be read as holding it. This is the situation with Michael Bratman. I say this even though he has offered an account of individual planning agency that harmonizes very well with reductionism. In any case, let me clarify my own view of his various claims, both about individual planning agency and about group agency.

Bratman’s account blends some of the normative considerations that drive both my own and Korsgaard’s accounts of agency with other considerations, some of which are analytical and some of which are empirical. In broad outline, here are some relevant highlights: an agent is able to conceive and plan a project with many different component steps; the agent can keep track of how its current thoughts and actions fit into its larger project; this keeping track is made possible by relations of reciprocal recognition among different intentional episodes within the agent’s life (the agent anticipates later remembering what it is thinking and doing in the present, and moreover, anticipates that its future memories will incorporate a recognition that they were anticipated); the agent can deliberate as needed about how best to proceed in the light of new information; this enables the agent to form and implement appropriate intentions as needed in order to carry out the project.

To say that this account harmonizes well with reductionism is to say that it invites us to think that anything that realizes the structure it identifies should qualify as an individual agent, including groups and parts of human beings as well as whole human beings; it is also to say that when groups of human beings satisfy the structure, the resulting group agent need not be a social phenomenon, because the human constituents of the group agent need not be sites of individual agents of human size, but may be sites of rational fragmentation instead.

But this is not the line that Bratman himself explicitly takes when he takes up the issue of group agency—which he takes up in the context of developing an account of the social phenomenon that he calls shared agency. According to him, shared agency arises when many distinct agents plan a project together on the basis of shared intentional attitudes, and implement it through shared intentional activities. As he conceives them, such shared intentional episodes are always possessed by individual agents, as opposed to the “plural subjects” that Margaret Gilbert has posited. (N.B. This is one of the places where a reader would naturally assume that what Bratman means by “individual agent” is an individual agent of human size.) But he insists that these individually owned attitudes can nevertheless be shared in the context of shared agency, insofar as they incorporate the right sorts of mutual recognition on the part of the individual agents who are involved—where this mutual recognition makes possible the kind of coordination of thought and action that is required by their shared project. If we bracket the explicitly social contents of shared intentional episodes, there is otherwise a remarkable similarity of structure in shared agency and individual planning agency, as Bratman portrays them—a similarity that he himself emphasizes by declaring that shared agency is continuous with individual planning agency. The question arises, might the similarity ever be so great, that shared agency would actually amount to group agency?

Bratman takes significant steps towards affirming this possibility in the context of his discussion of group subjects. He does not employ the term “subject” with its usual phenomenological sense that implies a center of consciousness, which most anyone would agree is missing in cases of group agency. What he means is a Davidsonian subject ofinterpretation. Davidson argued that such a subject must meet various holistic constraints on belief, and rational constraints on agency (Davidson 1984). It is clear that such a subject would qualify as an individual agent in the sense that I have been discussing throughout this paper—an agent with its own deliberative point of view that can be engaged in conversation, argument, etc. I have argued elsewhere (Rovane 1999) that, contrary to Davidson’s own assumptions, his approach to interpretation does not place any restrictions on how agents must fall with respect to human beings; that is, anything whose actions are interpretable as the actions of a single rational agent will count as an individual agent, and this includes groups and parts of human beings who would then constitute group and multiple agents. These arguments are of a piece with the point I made above about how Bratman’s account of individual planning agency invites a similar conclusion. But that is not Bratman’s focus when he takes up the issue of group subjects. His focus is on the empirical question whether, as a matter of fact, shared agency generally involves a group subject—a question to which he answers no. Interestingly, he does not go so far as to claim that it is impossible that shared agency could ever involve group subjecthood, and thereby meet the condition of group agency. Instead he remains cautiously open to the possibility in certain “special” cases. I am not sure what he has in mind by “special”. But one thing I am sure of is that he is envisaging such special cases as social phenomena—for the context makes clear that he is conceiving them as special cases of shared agency, which is by definition a social phenomenon.

As I’ve said, I think this is the common attitude among philosophers who recognize the possibility of group agency. They simply take it for granted, as something that needs no particular elaboration or defense, that human beings are sites of individual agency, and that group agency arises through social interactions among them. I think they also take it for granted that it would be a relatively straightforward and unproblematic task to portray group agency in this way, as a social phenomenon. The main burden of my arguments in this paper is that it is not at all a straightforward or unproblematic task. And I want to reiterate one last time the reasons why—setting them against the backdrop of Bratman’s vision of group agency as a special case of shared agency.

By hypothesis, a group agent’s deliberations would proceed from intentional episodes that are spread across different human lives, in accord with the normative requirements that define individual rationality. This suffices to ensure that these intentional episodes all figure in the same deliberative point of view—which is not the point of view of any agent of human size, but rather, the point of view of a single group agent. And then we must ask: Do these intentional episodes also figure, at the same time, in the lives of individual agents of human size? Can they so figure? As we answer these questions, we must bear in mind that the mere co-presence of different intentional episodes within a single human life does not suffice to bring them under a common commitment to rational unity, by virtue of which they would figure in the same deliberative point of view, and hence in the life of the same rational agent. To say that this does not suffice is equivalent to saying that there is a possibility of rational fragmentation within a single human life—conceived not as a rational failure, but as a reflection of the fact that there is no commitment to achieving overall rational unity within the whole of that life, even as there are commitments to achieving rational unity within other boundaries, which are wider or narrower than that biological boundary. The upshot is that the interactions across human beings, which constitute the deliberative life of the group agent, typically are not social interactions at all, but are, rather, interactions that take place within a single point of view—the point of view of the group agent. As I’ve pointed out, the situation would be different if the rational unity of a group agent did not typically come at the cost of rational unity within each of the group agent’s human constituents. But let us consider again what would follow if this were so. In such a case, each thinking and doing on the part of the group agent would have to be, at the very same time, a thinking and doing on the part of an agent of human size; and correlatively, some of the intra-personal relations that hold within the group agent’s point of view will also have to qualify as inter-personal relations, for they would hold among intentional episodes that figure in the lives of distinct human beings, each of whom is the site of an individual agent of human size. But it is not at all straightforward or unproblematic to portray intentional episodes, and the relations in which they stand, in this dual way—as figuring, simultaneously, within the life of a single group agent, and within the lives of many distinct agents of human size. This is why it is so hard to portray group agency as a social phenomenon. In order to do so, we need to identify conditions in which the rational unity of the group agent does not come at the price of rational fragmentation within the group agent’s human constituents—which I have argued are the conditions that Rousseau identified in his account of the general will.

It is quite understandable that all of this is not really visible to philosophers who write about group agency. For it wouldn’t be visible to any philosopher who didn’t have the reductionist account of agency firmly in view, along with its corollary about the possibility of multiple agency as well as group agency, which is to say, the possibility of rational fragmentation within a single human life.

This is certainly the situation with Philip Pettit and Christian List, who have co-authored a book length defense of group agency, not just as a possibility but as a reality—one that they argue it is morally and politically important to acknowledge.

Pettit and List share the common attitude that individual human beings are, necessarily, sites of individual agents of human size—an attitude they mark by calling human beings natural persons. Group agents, by contrast, are according to them artificial persons. What the two cases have in common is that they both satisfy the very same functionalist criteria for mindedness that human individuals do. When this is so, group agents are things toward which we can successfully adopt Dennett’s intentional stance —a stance from which we predict and explain a thing’s behavior by attributing beliefs, desires and other mental attitudes towards it.Footnote 12 According to Pettit and List, not only can we predict and explain a group agent’s behavior by attributing mental attitudes to it; we can also address it and engage with it in distinctively interpersonal ways, more or less in line with the Davidsonian interpretive picture on which Bratman draws in his discussion of group subjects. That is why they too are to be counted as persons, alongside human individuals.

Because Pettit and List believe that human beings are natural persons, there is little scope for them to argue that group agency need not be a social phenomenon. It may appear otherwise, because they explicitly distinguish a group agent’s point of view from the points of view of its individual human constituents—which they characterize as the autonomy of a group agent. But if each human constituent of a group agent is a natural person, then every thinking and doing that occurs within the life of a group agent must also be a thinking and doing on the part of one or another of its human constituents, each of whom is an individual agent in its own right. As I just explained above in connection with Bratman, this suffices to render the life of the group agent a social phenomenon—for it will be through thinkings and doings on the part of individual agents of human size that rational unity would be achieved at the level of the group. In other words, the art through which an artificial group agent comes to be is the art of natural persons—human individuals—as they exercise their agency together—a process that is bound to be social.

Matters would be different if Pettit and List were prepared to allow that human beings can be sites of rational fragmentation, for then they could allow that the intentional episodes that constitute the life of a group agent are not, in any sense, thinkings and doings on the part of agents of human size who are exercising their agency together in order to form a group agent; they could allow instead that such thinkings and doings on the part of the group agent occur alongside other intentional episodes within the lives of the group agent’s human constituents, which are thinkings and doings on the part of other agents who are smaller than human size—as I have described at many points throughout this paper. But Pettit and List expressly reject the possibility of such a non-pathological form of rational fragmentation within the lives of human beings—and indeed this rejection is very much part of what they have in mind when they insist that human beings are natural persons.Footnote 13

But let me set aside the question whether there is room for Pettit and List to allow, consistently with their view of human beings as natural persons, that group agency need not be a social phenomenon. It is in any case abundantly clear that they presume that group agency typically would be a social phenonenon. Much of their book on group agency is devoted to exploring different social processes through which individual agents of human size might function rationally as a group—such as different voting methods. I have argued elsewhere for a two-pronged conclusion about these explorations, which is very much in line with my arguments in this paper: on the one hand, unless such social processes are designed to conform to the normative requirements of individual rationality, they will not issue in anything that qualifies as a single group agent with its own point of view; but on the other hand, when such processes do conform to the normative requirements of individual rationality, this puts into doubt that they really are social processes at all, as opposed to perfectly individual deliberations on the part of a single group agent.Footnote 14 But as I have said, none of this will be in view for philosophers unless they also have in view the reductionist account of agency, and its implications about the possibility of rational fragmentation within a human life, and the reasons why group agency would typically be an occasion for such fragmentation—as Pettit and List manifestly do not.

In closing, I want to shift my attention away from these metaphysical problems of “social ontology”, in order to address some of the normative issues that Pettit and List bring to the fore in their discussion. They subscribe to a liberal vision on which the individual human being is a locus of individual rights. And they agree with Rousseau that there is always a danger that the formation of group agents—whether it be for explicitly political purposes or for other purposes—may result in forms of domination over individual human beings that are inconsistent with liberal ideals. So their own exploration of various social processes through which group agency might be achieved comes with a recommendation that any such process ought to be expressly designed in such a way as to prevent such domination by group agents over their human constituents. It should be clear from my arguments in this paper, that I think anyone who is moved by Pettit and List’s recommendation should be interested in designing group agents that meet the conditions of Rousseau’s general will.

But let me also add: whatever merits their recommendation may hold, we should not undertake to implement it without a clear-eyed sense of some of the difficulties that the reductionist view of agency poses for the liberal point of view that animates it. First, reductionism undercuts the ground on which Pettit and List would draw a moral distinction between individual agents of human size, whom they regard as natural persons, and group agents, whom they regard as artificial persons—for according to the reductionist account, all agents, no matter what their size, are products of effort and will, and in that sense artificial. Second, the classical liberal arguments for individual rights entail that we must grant rights to any agent who can understand the arguments for them, and who can then, in the light of such understanding, claim rights on its own behalf. Third, this means that the liberal framework doesn’t really provide a good ground on which to reject the rights claims of group agents—and alas, in particular, it doesn’t provide a good ground on which to reject the rights claims that Citizens United made for itself in the U.S. Supreme Court. Fourth, just as the liberal framework cannot easily dismiss rights claims of group agents, it cannot easily explain why all human beings should be regarded as loci of rights—in liberal terms, a human being will qualify as a locus of rights only if it forges a point of view of human size from which it can deliberate and act, and also claim rights on its own behalf. Fifth, Pettit and List’s recommendation that we form group agents in such as way as to protect their human constituents from domination could not be implemented unless we formed agents of human size to begin with, and so this is really an implicit part of their recommendation!

In order to justify this last, implicit, part of their recommendation, Pettit and List cannot fall back on the mistaken metaphysical assumption that human beings are natural persons. They must argue instead that the projects that can be pursued by forming agents of human size are better, and worthier, than other projects that could be pursued by forming agents of other sizes (smaller or larger) who might exist in their stead. I can envisage some lines along which they might mount such an argument.Footnote 15 But I hope it is clear that, whatever line such an argument might take, it will do nothing to undermine my arguments in this paper, which show that group agency is not necessarily a social phenomenon, and is not typically a social phenomenon, and cannot be a social phenomenon except in some very special conditions that share certain features of Rousseau’s general will.Footnote 16