Kutz (2000) Complicity (3) Acting Together

Kutz, Christopher (2000). Complicity: Ethics and Law for a Collective Age. Cambridge: Cambridge University Press.

 

3.1 Introduction

Two partners plan to rob a bank. The first recruits a driver while the second purchases a shotgun from a gun dealer. The driver knows he's taking part in a robbery, although not a bank robbery. The gun dealer should have checked his customer's police record before the sale, but failed to do so. The bank is robbed, a guard is killed, and the robbers escape, only to be caught later. 'They committed bank rob bery," a prosecutor will say. But does "they" include the gun dealer, whose lax standards made the robbery possible? 'They conspired to rob the bank" - but does "they" here include the driver, who didn't know it was a bank they were robbing? "They killed a bank guard" but does it matter who pulled the trigger?

The general intuition I shall be exploring at length is that collective action is the product of individuals who orient themselves around a joint project. The particular form of the analysis I will defend makes use of the notion of an individual participatory intention, or an intention to act as part of a group. When suitably combined, individuals acting upon participatory intentions achieve jointly intentional action, and the group of which they are a part can be said to have acted.

3.2

In this chapter, I aim to give an account of collective action, collective intention, and individual participation in collective action. Indeed, I will argue that the first two items on the list can be explained by the third. But collective action (or joint action, as I will often refer to it) is a slippery notion: two drivers jointly navigate a four-way stop without crashing into one another; you and I play a chess match; a baseball team makes a double play in the bottom on the ninth; Exxon posts its third-quarter earnings; a hostile mob storms the Bastille. These joint acts involve very different kinds of groups, whose individual mem bers engage in very different activities, with different degrees of mutual interaction. I want to lay out the methodological structure of my approach to these disparate phenomena: I seek an explanation that generalizes over a broad range of collective acts, explains those acts in terms of individual mental states and, in particular, in terms of individual intentions and beliefs.

3.2.1 For Generality

Are the myriad forms of collective action in which we continuously engage susceptible to a single analytical account? We can easily iden tify at least five dimensions of variation in types of joint action, beyond the infinite variations among particular species of joint action.

1. the number of agents, ranging from the minimal two, to populations of hundreds of millions, as in general elections.
2. task-intricacy: Many coordinated activities, such as negotiating a four-way stop sign, involve few choices (I'll go if you wait). Other activities, such as conversing or playing tennis, involve great responsiveness, the evaluation of many options, and sophisticated individual skills
3. collective activities will vary in cooperative spirit. That is, some activities require degrees of goodwill, or willingness to put forth extra effort or incur extra costs for the sake of others (playing a team sport), while other activities can be conducted with little goodwill on the part of others (merging onto a freeway).
4. joint actions can involve different levels of agent autonomy, or individual (nonresponsive) discretion in how to perform one's task.
5. collective activity can be more or less egalitarian. Individuals can vary in the influence they actually or properly have over the direction taken by the collective, in choice of ends and means.

I will argue in this chapter that although various elements of organization and interaction are essential to particular types of joint action, most forms of collective action share a common structural feature: individual members of a group intentionally do their parts in promoting a joint outcome, or engaging in a joint activity.

Participatory intentions explain both the nature and possibility of joint action, and its distinctive normative contours.

I seek a general explanatory framework in order to make sense of the commonalities in our normative responses to individuals who participate in wrongful or harmful collective acts. Although forms of individual participation may vary greatly - from the role of a criminal kingpin to that of an investor in a predatory corporation - intentional participation generally shapes agents' normative relations to the consequences of collec tive action, as well as their relations to other members of the group. Because of their participation, agents can be accountable for acts and outcomes attributable to the group as a whole, as well as for acts attributable to other participating members.

3.2.2 For Individualism in the Explanation of Collective Action

Perhaps because as a matter of social fact we often hold individuals accountable for what groups, or other members of groups, do, it is tempting to think collective action is in some sense prior and irreducible to individual action. That is, accountability appears to accrue first to the jointly acting group, and then derivatively to its individual members.

This feature of our practices of accountability has suggested to some that individual action is also explained by collective action, so that, for example, my stepping left is explained by our dancing a waltz. On such a holistic view, individual action is seen as a product of a collective will (perhaps embodied in individuals), such that the latter explains the former and not vice versa.

By contrast, on an individualistic view, collective action is explained by individual intentions and actions: our waltzing is explained by my dancing my part and you dancing yours. Individualists typically claim that collective action can always be "reduced" to individual action.

The possibility of the reduction of claims about groups to claims about individuals.

Take the sentence, "Because Exxon posted a third-quarter loss, its share price fell." This sentence predicates an act of a collective entity, Exxon, and attributes an effect to that act. Methodological individualists suggest that such sentences may be replaced in many contexts, perhaps for the sake of social-scientific explanation, by sentences referring only to individuals and individualistic predicates, that is, predicates that do not relate individuals to social institutions.

Individualists would claim Exxon refers only to a set of individuals (some subset of Exxon employees and share holders); that the sentence could be replaced by a (very large) set of sentences about those individuals, as well as about individuals trading Exxon stock; and the best explanation for the lower share price can be deduced from sentences about those individuals.

Holists, in contrast, might claim Exxon's posting of a loss can only be explained by reference to a collective plan for notifying the public about the company's performance. Though Exxon's act necessarily involves the actions of certain company officials, a holist might say, their actions can only be explained by reference to the collective, insofar as the officials both conceive of themselves as acting as their institutional offices require and because their acts are only regarded as authoritative in virtue of those offices. Holists do not deny that individuals act; they deny that the best explanation of social facts is couched solely in terms of facts about those individuals.

Statements about collective acts may be rephrased always as statements about individual agents, because all collective action is explicable in terms of the intentionality of individuals - their motives, beliefs, and plans. If Exxon posted a third-quarter loss, then this fact may be explained in terms of the acts and intentions of individuals who saw themselves as acting on behalf of Exxon, as well as the expectations and beliefs of other individuals regarding what Exxon is and what acts it is capable of. Furthermore, the corporate policy these individuals saw themselves as promoting can itself be explained as the product of the deliberations and negotiations of and between individuals. Here the adequacy of such a reductive explanation is relative to a particular theoretical purpose, namely understanding the causal his tory of Exxon's act with the smallest gaps between intermediate explanatory events. Other purposes may be better served either by ignoring individual mediating events or looking to more distant original causes. For example, citing unrest in the Middle East might provide a better explanation of Exxon's posting a loss than would simply resting with unexplained facts about individual pricing and buying behavior.

My claim is only that individual intentional action always implicitly mediates the causal explanation of collective acts and events, not that referring to individual acts always provides the most useful explanation.

This very weak form of individualism does not generalize to any strongly individualistic position in the philosophy of the social sciences. Indeed, it is compatible with many moderate forms of holism. In particular, I deny that a full explanation of collective action can be given without reference to collectives or social facts, because refer ence to irreducibly holistic facts and entities must occur in an account of the mental states of individual agents.

For example, if a group overcomes free riderism and collectively provides a public good, the explanation may be that individuals accept fairness norms, modify ing their preferences. However, accepting an individualistic explana tion of the act does not commit one to the further view that accep tance of these norms is in any strong sense a choice of the individ uals. The best explanation of the inculcation and acceptance of fair ness norms might be nonindividualistic, for example, it might be a form of group-level evolutionary adaptationism.

In addition, many social groups cannot be reduced to sets of their members, because some groups can persist through changes in their membership (Exxon would be Exxon with a new bookkeeper). That is, certain structured social groups have nonextensional (or non mereological) identity criteria. In many cases, the identity of a group is grounded in individuals' dispositions to identify themselves (and certain others) as members of that group.9 In other words, group identity is explained in terms of individual participatory intentions. These dispositions include not just inchoate, romantic feelings of group solidarity, but a willingness to assume obligations taken on by other group members, to speak, decide, and act on others' behalf, and to deliberate about how to act so as to further collective plans and intentions. The identity of Exxon is independent of the extensio nal composition of its membership, for example, because newly ar riving insiders understand themselves to be joining the Exxon organ ization, and outsiders attribute representative authority to self proclaimed Exxon members.

3.2.3 Intentionalism and Functionalism as Methods in the Theory of Action

My account of intentional actions generally, and individual intentions in particular, is intentionalist and functionalist in form, and thus in keeping with a large body of philosophy of action.

Following Donald Davidson, I will assume intentional action is action (body movements) that is both causally and teleologically explained by an agent's goals, as those goals are embedded in networks of intentions, desires, and instrumental beliefs. Describing an action as inten tional is appropriate because of the logical and causal role of the goal in explaining the actions in question. Goals teleologically explain actions so long as there is a possible deliberative route from what an agent wants or intends to what the agent does; the agent need not actually have deliberated.

As long as what the agent does satisfies a goal nonaccidentally, an intentional action is performed, and the action is intentional under a description appropriately related (or identical) to a statement of the agent's goal.

Intentional actions that are a means to an end may be redescribed in terms of their ends, either as contributions towards that end, or, if closely connected, as realizations of that end.

My intentional turning on of a switch to light a room may be redescribed as my intentional lighting of the room. My opening the refrigerator to get the mayonnaise may be redescribed as my preparing to make a sandwich.

In the collective context, this aspect of intentional action allows us to re-describe individual contributions in terms of a collective end: The musician, for example, is not just playing the viola, but is performing - along with the others - a certain symphony. The musician's intention to participate in a collective act, playing the symphony, both causes and rationalizes the viola playing, and so licenses our redescription. The possibility of legitimate redescription will be central to my account of individual accountability for collective acts.

Mental states in general, and intentions in particular, are defined by their role in a causal theory that maps agents' psychologi cal inputs (perceptions, intentions, beliefs, and desires) onto their outputs (actions, subsidiary intentions, further beliefs, and desires). Intentions mediate between agents' beliefs and desires (their reasons) and their practical reasoning and action.

Functionalists attribute content to individuals' intentions by interpreting their planning and action in terms of reasons that explain and rationalize that deliberative behavior. Although individuals are presumably introspectively aware of the content of their own intentions, we can also attribute intentions on the basis of behavioral observations coupled with a general theory of human rationality.19 Because func tionalists identify the content of intentions according to the best in terpretation (or most coherent theory) they can offer of agents7 plan ning, dispositions, and behavior, functionalist analyses deny a strong first-person epistemological privilege. Other interpreters may be in equally good positions to make sense of an individual's behavior.

Functionalists may, therefore, ascribe intentions in virtue of uncon scious or explicitly disavowed goals and motives if an interpretation making use of those motives is better than the interpretations offered by agents themselves. This point will be important in Chapter 5, when I consider those cases in which individuals claim to be alien ated from a collective activity, but in fact contribute to its realization.

3.3

I will defend an account of collective action in which what makes a set of individual acts a case of jointly intentional action is the content of the intentions with which the individuals act. In particular, I will argue that jointly intentional action is primarily a function of the way in which individual agents regard their own actions as contributing to a collective outcome. I call this way of regarding one's own action, acting with a participatory intention. In this section, I will show why participatory intentions are ineliminable elements of any account of joint action

3.3.1 The Necessity of a Collective Conception

This raises the obvious suspicion that there simply are no univer sal conditions constitutive of collective action as such, that collective action types simply hang together in a familial fashion. I will argue in this section, however, that all forms of collective action share a com mon element in the form of overlapping, individual participatory intentions. My strategy is first to try to elicit the individually neces sary and jointly sufficient conditions for a case which is under detailed enough to generalize plausibly, yet determinate enough to guide our intuitive assessment of the analysis. Then I will show whether these conditions remain necessary for other collective action types.

We can imagine cases of spontaneous collective action in which there is neither mutual responsiveness nor mutual awareness beyond what is implicit in the idea of a participatory intention - a conception that one is doing one's part in a collective project.

Consider first some examples that resemble jointly intentional action but are not genuine - examples where it would be false to say some group of agents jointly did G, where G stands for either a joint act or activity (such as playing softball), or for an outcome brought about by several agents (such as the performance of a symphony). You and I may go to Chicago together by happening to go there in the same plane or train, but it would not be true to say that we went to Chicago jointly if our coming on the same flight or train was sheer coinci dence.20 Jointly acting individuals do not merely act in parallel: Each responds to what the others do and plan to do.

Thus, our going to Chicago jointly requires that the presence of each of us on the plane or train somehow depends upon the presence of the other. Unless my choice of transportation somehow depends upon your choice (or my expectations regarding your choice), and similarly for you, we will not have coordinated our going to Chicago. Let us call agents' intentions strategically responsive if what they intend to do is sensitive to their beliefs or predictions about what others intend to do.

Joint action will often - and coordinated action will always - require strategic responsiveness: Individuals acting jointly decide to act, and do act, in light of their beliefs about other potential or actual joint actors.

You and I satisfy another plausible condition on joint action: We share a goal. Let us say two agents share a goal if there is at least one token activity or outcome involving the actions of the other whose performance or realization would satisfy the intentions of each.21 Both of us share the goal of going to Chicago on the same mode of transportation. Even competitive forms of joint action involve some shared goal. You and I may each be trying to beat the other at chess, and hence no ending of the game will wholly satisfy the goals of each of us. But, at a lower degree of resolution, there are some goals we do share: We each seek an orderly game with the other, played by the rules, and so on. If we did not share these goals, we could not com pete with one another, for there would be no background against which to assess the other's performance.

Sharing goals, in a sense that I specify further in Section 3.7, is a necessary condition across all forms of joint action.

Conversely, it seems that for our going to Chicago together to be joint, we each must believe it at least possible the other knows of or will try to predict our choice, and be favorably disposed to the other's knowledge or anticipation of that choice at least in the sense that no one would modify his or her plans in virtue of disclosure. As friends rather than spies, each knows or hopes the other got notice of the planned means of transportation, or was otherwise able to predict the other's choice. If, in fact, the messages did get through, or if we are able successfully to anticipate the other's choice, then each will be acting consistently with the other's preferences: our individual aims are furthered rather than frustrated by the other's awareness. So for our trip to be a joint production, each must not only act in light of beliefs about the other's plan, but each must also be favorably disposed towards the other's possible knowledge of this strategic sensitivity. More simply, each must be open to the possibility of joint action. Call this a condition of mutual openness concerning our interaction. Mutual openness is a much weaker condition than common knowledge of our situation: it can accommodate those cases of joint action that come off despite inchoate expectations about the other's plans or awareness.22 Yet it is strong enough to exclude cases of fully adverse strategic interaction, cases like the spy example that are not plausibly regarded as joint.

By contrast, if we are friends trying to go to Chicago together, our goals will not be fully satisfied by a third party's intervention. For we do not (presumably) simply want it to be the case that we arrive in Chicago on the same plane. Rather, we want our traveling together to be the product of the decision of each. That is, we go to Chicago jointly when we go there together, and our going there together is the product of each of us acting with an intention of contributing to our joint project of getting to Chicago together. We act jointly when we act as members of a group who act together.

The important point is that this additional requirement in the cooper ative case distinguishes it from competitive game theory. In competitive contexts, each attempts to achieve the most preferred outcome, based on the most likely choice of the other (or, in the absence of any further information, based on the choice of the other that would make for the worst outcome). Each player is like Spy and Count erspy: the other's choice is simply part of the background against which strategy is formed.28 In cooperative contexts, by contrast, each must act in accordance with a conception of the other as committed to joint resolution of the problem. In Gerald Postema's expression, each "deliberates from the first-person plural/'29 Each sees the other as an intentional participant in a collective action.

If each of us deliberates in this manner and we do both take the plane together, then it will be true that we went to Chicago jointly. What makes this claim true is how we each conceived of the choice, as one in which each intends to do his part in promoting our group act, not any prior agreement or concert on our part. Call this way of conceiving of action a participatory intention: an intention to do my part of a collective act, where my part is defined as the task I ought to perform if we are to be successful in realizing a shared goal. This conception of oneself as contributing to a collective, as manifested in one's deliberation and action, is what lies at the heart of collective action generally, from simple coordination to complex cooperation.

3.4

A participatory intention has two representational components, or sets of conditions of satisfaction: individual role and collective end. By individual role I mean the act an individual performs in order to foster a collective end; and by collective end I mean the object of a descrip tion that is constituted by or is a causal product of different individ uals' acts. This is to say that individual participatory action aims at two goals: accomplishment of a primary individual task that con tributes to a secondary collective achievement, be it an activity or an outcome. The collective end might be a state of affairs whose realiza tion depends upon several agents acting together, such as the move ment of a heavy object; or an activity, such as dancing a tango; or it might be a social group with characteristic behavior or internal culture, such as a university faculty. Some joint activities can be performed jointly intentionally or unintentionally, such as going to Chicago together. Others can only be performed jointly intentionally if they are performed at all, such as playing chess or dancing the tango.

The defining characteristic of a participatory intention, then, lies in the form of relationship between individual act performed and the group act or outcome that rationalizes the part.

Contributory relations might take instrumental form if what the agent does helps cause the collective outcome (my pushing helps to move the car), or if the agent's part is a constitutive element of the group act (stepping this way is part of dancing a tango). The relation might be expressive if by doing one's part, one thereby exemplifies one's membership in a group or participation in an activity, as when by voting I express my membership in a political community. And the relation might be normative if one performs one's part because of norms internal to some group or institution that demand certain behavior (I wear a dark suit as an IBM employee). Of course, a single act may stand in many contributory relations to a group goal or activity. If I am a member of a criminal conspiracy, my refusal to cooperate with the police furthers the success of the conspiratorial objective, adheres to the norms of criminal honor, and expresses my solidarity with my co-conspirators.

What makes my behavior participatory is nothing more (and nothing less) than my conception of what I do as related to the group act, whether that conception is explicit in my deliberations, or func tionally implicit in my actual or counterfactual behavior.

Given a suitable background story, attributing participatory intentions requires only that individuals regard themselves as acting for the sake of some joint goal; no more content is necessary.

3.5

I will now argue the content of agents' intentions can be irreducibly collective so long as the structure of their intentions is straightfor wardly individualistic.

John Searle offers such an objection against the reductive account offered by Raimo Tuomela and Kaarlo Miller. Tuomela and Miller explain jointly intentional action in terms of individuals who act with the intention of doing their parts in a collective act, although they add conditions I believe are superfluous.45 Searle argues that jointly intentional action can only be explained by positing a distinct, "irreducible" form of intending that he calls "we-intending." We intentions are individual intentions to engage in collective activity, distinct from "I-intentions" to perform one's own acts. The difference is not just the difference in content I have already mentioned - the goal of participating in collective action differs, tautologically, from the goal of engaging in individual action - but also in form.

Searle's argument is important, not because it shows that collective action demands a separate form of intending, but because despite his actual claims, it reveals how collective content is necessary to distinguish cooperation from merely parallel behavior. He proposes the following counterexample: a group of business-school students has been indoctrinated to believe one can best help humanity by pursuing one's selfish interests when others do so also. Furthermore, each student believes that each student believes this to be true and will act upon it. Since "each believes that his selfish efforts will be success ful in helping humanity," Searle claims each "intends to do his part" of helping humanity by pursuing selfish interests.46 He objects that even if the business students knowingly do something that, taken together, helps humanity, they are not acting jointly. Contrast this with the case of business-school students who form a pact to help humanity by individually pursuing their selfish interests.

There surely is an important difference between the two cases. If the first students pursue their selfish interests in the belief that they jointly will help humanity, but without intending to promote that end, then they will not be acting with the intention of jointly benefiting human ity. By contrast, if the second group of students forms a pact to help humanity by acting selfishly, then they are acting with the intention of jointly benefiting humanity. But the difference between the two groups is not one of form. It is instead the familiar difference between intending a certain result and acting with the knowledge the result will obtain. The first group of students believes humanity will be helped because of their (collective) selfish efforts, but they do not intend that end; their actions are not counterfactually sensitive to its achievement. Presumably they would act selfishly even if they came to believe humanity was rendered worse off. Members of the second group, however, act selfishly in order to promote an end that can only be brought about collectively. They would not act selfishly if their selfish acts were likely to be fruitless, since they aim at the end of benefiting humanity and not merely of acting selfishly.

So the difference we need to capture is simply one of intentional content: Jointly acting groups consist of individuals who intend to contribute to a collective end, whether outcome or activity. Groups of individuals, all of whom merely know they happen to be contribut ing to a collective outcome, cannot be said to act jointly. So long as we see individual actions as aiming at the achievement of a collective end, we can attribute to them participatory intentions, defined in terms of their goals rather than their form. Or, to put the same point another way, we can have irreducible content and reducible form.

3.6

I have claimed jointly intentional action is fundamentally the action of individuals who intend to play a part in producing a group out come...In this section I will defend this conception of joint action against the more demanding accounts of some other philosophers.

The superior descriptive coverage of the minimalist conception is only part of my reason for favoring it. My principal reason is norma tive. Ethically complex cases of joint action rarely involve perfect common knowledge, wholly shared conceptions of the joint act, or highly responsive strategic interaction. Indeed, the genius of orga nized criminality lies precisely in obscuring the interrelations of par ticipants by removing the need for frequent interaction. And the enterprises responsible for significant unintended harms are like wise typically distinguished by the dispersion of task responsibility. Conspirators, for example, often compartmentalize knowledge of individual tasks and identities; corporate officers and engineers may understand themselves to pursue very different goals; and executive officials may shield themselves from specific knowledge of the acts or omissions of subordinates. If joint action is to have special norma tive significance in such cases, either as a basis for holding individ uals accountable for the acts of others, or for aggravating the serious ness of individual offenses, then an account of joint action must not rely upon high degrees of interaction or mutual knowledge.

Gilbert, Bratman, and Tuomela and Miller therefore suggest that mutual expectation and responsiveness are essential to jointly intentional action as such.

Gilbert argues that a participating member of a "social group" must first accept that others have committedthem selves to the group project.

Bratman requires that each agent intend that the group effort be realized in part because of the intentions of the other participants, with these intentions commonly known.

And Tuomela and Miller claim that jointly acting agents must believe a sufficient number of themselves intend to do their parts, and that this is commonly known.

But it is crucial to recognize that mutual and universal responsive ness are not necessary to joint action as such, at the stage either of intention formation or execution. Two types of ostensibly collective action show that individuals need only have very weak hopes or beliefs about each other's plans. Some collective acts emerge when an entrepreneurial agent begins doing what will be part of a joint effort, but only if others follow suit. A member of the Bastille mob starts throwing rocks at the prison guards; soon others join him, and they collectively storm the prison. While the subsequent participants intended to do their parts of storming the prison in response to the sight of the first, the actions and intentions of the first stone thrower were neither responsive to nor predicated upon definite expectations about the others' conduct. Nonetheless, the leader's intention must be characterized as participatory, for he surely would not have thrown stones if he thought it impossible that others would join in. It is unnecessary to attribute to him the determinate expectation that others actually would join in, fomenting revolution being the risky business it is. Rather than requiring participants to have positive beliefs about the prospect of others joining in (i.e., regarding others' participation as more likely than not), we need only require they not regard the prospect as impossible.

No member of a group need form an intention in the light of expectations about the others. Suppose while we are having a picnic, it begins to rain. I jump up, grab the sandwiches, and head for the car. I intend to do my part of our saving the picnic, hoping you will simultaneously grab the drinks and the blanket.53 If you do, then it is reasonable to say we will have jointly saved the picnic. We might not have acted jointly, if, say, you had been dozing when the rain hit. But if we do both act with participatory intentions, then we will have jointly intentionally saved the picnic though neither had formed an intention to save the picnic in the light of expectations about the other's intentions.

Clearly, some degree of overlap is necessary in order to character ize an act as jointly intentional. It might even be thought that in paradigm cases of joint action, overlap is perfect: There is only one state of affairs that satisfies each agent's intentions, and it is com monly recognized. But perfect overlap is rarely, if ever, the case. If we intend to go for a walk together, then the state of affairs in which you and I stroll together, responsively and conversationally, satisfies the intentions of each. But I may further intend that as we walk, we talk about my ill health, while you have no set conversational agenda. Thus, while a broad range of states of affairs will satisfy your inten tion that we stroll together, only a subset of those will satisfy mine.

So overlap is essentially a pragmatic concept and always a matter of degree, given inevitable differences in each agent's expectations and conceptions of the group act. Agents will have more or less determinate conceptions of the group act, they may be more or less willing to compromise after bargaining on the character of that act, and they may have very different ideas about the scope of the group act, its duration and membership. As a result, a group act can be jointly intentional under one description and not jointly intentional under another. You may believe we are going to a friend's house for a quiet dinner, while I believe we are going for a surprise party. While our going to the surprise party is not jointly intentional, our going to the friend's house is.

3.7

The central reason that Gilbert, Bratman, Tuomela and Miller claim that participating individuals intend to act in light of their positive expectations about one another's plans is that they explain joint action in terms of individuals' intentions that their group perform an act. I will refer to these intentions as group-intentions.

The introduction of group-intentions is permissible, and perhaps even required in order to explain the practical reasoning and plan ning of some members of jointly acting groups. In general, when agents act so as to realize the collective outcome, to the extent of aiding others in their contributions, we should attribute to them the group-intention to achieve that collective end. However, I think that Bratman and Tuomela and Miller have been misled by their reliance on examples of collective action where such planning is universally shared, namely cases of small-scale, highly interdependent, and non hierarchical cooperation, in which each participant plausibly aims at everyone's achievement of the group goal. Although individuals who intend "that we do G" intend to do their parts of "G-ing," the converse is not true. Individuals may intend to do their parts of our "G-ing," and thus jointly G, without intending that we G. The result is that collective action need involve only individuals acting upon participatory intentions, not upon group-intentions.

In large groups, individuals whose contributions are marginal will typically not have an executive intention with respect to producing the total outcome or activity Instead they will have a subsidiary, participatory intention, an intention to do their part of achieving the executively determined goal. They may have an intention regarding the whole but they don't need such an intention to identify with and act for the sake of the main goal. Their individual participatory inten tions will in turn serve as executive with respect to further intentions and actions. The cellist, for example, has a subsidiary intention to perform the cello part of the Eroica, which generates further intentions to play in tune and tempo, and to show up for rehearsal on time. And the cellist's participatory intention may be subsidiary not only to the music director's intention that the orchestra perform the symphony, but to the cellist's own self-regarding intention to make a career out of music, to play as much Beethoven as possible, and so on.

By contrast, a conductor's planning and action is aimed at the goal that the symphony together perform the Eroica. The conductor is therefore disposed to choose suitable rehearsal times, to ensure that replacement players can be found in case of illness, to study scores and previous performances so this performance can be novel or traditional, and so on. I do not want to suggest that the cellist could not have these dispositions as well, but only that there is no independent justification for positing such an intention. By contrast, it is hard to make sense of the conductor's behavior unless we posit such a group-intention.

Collective action always involves intentional participation; it does not always involve group intentions.

3.8

It will be helpful to distinguish between two sorts of jointly acting groups: ephemeral groups, and institutional groups. Ephemeral groups are groups whose identity as a group consists just in the fact that a set of persons is acting jointly with overlapping participatory intentions. When we push a car out of a snow bank, we are an ephemeral group, whose identity is given by our mutual goal of pushing the car. Our overlapping participatory intentions distinguish each of us as in siders of that group, while excluding from membership those watch ing from the sidewalk.67 But there is no further criterion of member ship. Institutional groups, by contrast, have identity criteria that do not wholly consist in the presence of overlapping participatory in tentions. I cannot make myself a member of the Giants by running out onto the field and catching a line drive, or of the U.S. Senate by intentionally participating in its deliberations. In the case of some institutional groups, recognition of one's membership by other mem bers may be sufficient. Other groups, like the U.S. Senate, have addi tional necessary membership conditions: Even if other members re gard me as a Senator, if I did not win a majority of the vote, I am not a member of that group.68

3.9

I want to restrict discussion of genuine intentions to individual mental states, rather than absorbing functionally characterizable in dividual and interpersonal planning and action under the same term.

By collective intention I mean essentially a figure of speech, referring not to a supra-individual mental or functional state, but to the region of overlap among individual participatory intentions.74

A collective intention is attributable to a group when the follow ing three conditions are met:
(1) Members of the group are intentionally members of that group. That is, they are disposed to participate as members of the group in deciding upon a shared plan and then in acting in conformity with that plan.
(2) There is an explicit or implicit collective-decision rule by which a collective intention may be assigned to the group in virtue of individuals' intentions to participate in forming and abiding by that collective intention.
(3) The participatory intentions of the individuals overlap suffi ciently to meet the constraints of the collective-decision rule.

3.10

I have argued for a conception of collective action that is both indi vidualistic with respect to agency and irreducibly holistic with re spect to the content of agents' intentions. Groups are nothing more or less than agents who intend to participate in collective action. I have also argued for a minimalist conception of the conditions of collec tive action. Although particular types of joint activities may require high degrees of responsiveness and robust mutual expectation, joint action as such merely requires there be sufficient overlap among the objects of agents' participatory intentions. From these few elements, we can do much: We can explain what we do together.