Michael Bratman’s magnificent book Shared Agency (Bratman 2014) aims to establish the continuity of “the conceptual, metaphysical, and normative structures central to … modest sociality … with structures of individual planning agency” (p. 8). The strategy is to provide sufficient conditions for modest sociality – the sort of collective intentional activity that finds expression in friendship and love, singing duets, dancing together, having a conversation, building a boat together (intentionally), or the orchestral performance of a symphony – while drawing only on the conceptual, metaphysical, and normative resources already available in our understanding of individual planning agency. To provide sufficient conditions for forms of small-scale collective intentional activity that draw only on these individualistic resources is to show that collective intentional activity as such does not require the introduction of irreducible concepts of collective intentional action, sui generis forms of intention or commitment, or irreducibly group level agents, who are themselves the subjects of intentions directed at their actions.
The central target of analysis is shared intention. Collective intentional activity (at least in the case of modest sociality) is understood as collective activity brought about in the right way by a shared intention. What is a shared intention? It is natural to take it to be what is expressed by a particular reading of a sentence such as .  has a distributive reading on which it is understood as in .
 We intend to sing the national anthem.
 Each of us intends to sing the national anthem.
 says of each one of us that he or she intends him or herself to sing the national anthem, as we might if we were all auditioning for a patriotic musical, and had selected the same audition piece. However, if we are members of a chorus, planning a choral performance of the national anthem, we express by an utterance of  our intention to do something together – intentionally, as we might put it. This gives us what I will call the collective reading of . If  is true of us on the collective reading, it seems appropriate to say that we share an intention to sing the national anthem together. This is more than just each of us intending that we sing the national anthem together. For we might each intend we do something together but by a plan that does not see the others as in on our doing it, and this would not amount to our intending to do it together in the relevant sense. For example: four of us, with apartments opening onto a courtyard, each know the three others have selected the national anthem for audition, but not that the others know the corresponding thing. We know the others practice – windows open – at 6 pm, and expect them to sing their audition piece. Each of us intends that we sing the national anthem together by singing it along with them at 6 pm. Still, when we sing the national anthem together as a result, we do not sing it together intentionally.
However, Bratman does not identify shared intention, at least in the first instance, with what is expressed by sentences like  on the collective reading, and this will be connected with a question I want to raise about the account. Rather, Bratman introduces shared intention as playing a role in collective intentional activity analogous to the role that individual intentions play in individual intentional activity. In particular, shared intention is to play an explanatory role in shared intentional activity analogous to the role played by individual intention in individual intentional activity. While this may sound as if it is tailor-made for the view that groups per se are the subjects of shared intentions, the goal is rather to show that we can find a structure in a group that is constructed out of individual intentions interlocking in content together with interdependencies between them, mediated by belief and knowledge, which plays a functional role in group action analogous to intention in the individual case. Individual intentions are planning structures that coordinate and guide action and deliberation. Analogously, shared intentions are planning structures that coordinate and guide collective activity and deliberation, and support “associated norms of social agglomeration and consistency, social coherence, and social stability” (p. 27).
Given space constraints, I limit my discussion to the central constructive account given in chapter 4 of Shared Agency. This is a major contribution to understanding the nature of shared intention and collective intentional action. The care and rigor with which the materials for the account are developed and put together, defended from objections, and contrasted with alternative accounts, is an impressive intellectual accomplishment. I am in basic agreement with Bratman on all major theoretical points. I believe that the continuity thesis is correct (quite generally). I agree that shared intentional agency does not undermine any plausible own action or control conditions. I agree that neither moral nor sui generis contra-lateral obligations are constitutive of shared intention or shared intentional activity. I agree that shared intention should be understood in terms of concepts already at play in our understanding of individual agency [my own account (Ludwig 2007, 2014a,b), with somewhat different starting and ending points, is very much in the spirit of the approach in Shared Agency]. I will suggest, however, that the account shifts our attention away from what strictly constitutes of shared intention because it focuses on a subgenre of collective intentional activity, with features that are inessential to shared intention in general, and arguably not even necessary parts of sufficient conditions.
The account is expressed by the compressed basic thesis, that conditions A–D below provide sufficient conditions for shared intention, and that condition E provides a sufficient condition for modest sociality in action, that is to say, a particular form of collective intentional activity (p. 103).
Intention condition: We each have intentions that we J; and we each intend that we J by way of each of our intentions that we J (so there is interlocking and reflexivity) and by way of relevant mutual responsiveness in sub-plan and action, and so by way of sub-plans that mesh.
Belief condition: We each believe that if the intentions of each in favor of our J-ing persist, we will J by way of those intentions and relevant mutual responsiveness in sub-plan and action; and we each believe that there is interdependence in persistence of those intentions of each in favor of our J-ing.
Interdependence condition: There is interdependence in persistence of the intentions of each in favor of our J-ing.
Common knowledge condition: It is common knowledge that A–D.
Mutual responsiveness condition: Our shared intention to J leads to our J-ing by way of public mutual responsiveness in sub-intention and action that tracks the end intended by each of the joint activity by way of intentions of each in favor of that joint activity.
Condition A expresses the key idea, and pertains to what is special about the participatory intentions of agents who engage with others in collective intentional activity (their we-intentions). There are two components. The first is that if we intend to J, each of us intends that we J and that we J because of and by way of the intention of each that we J. In earlier work, Bratman expressed the second component as the requirement that we each intend that we J by way of meshing subplans associated with our respective intentions that we J. In Shared Agency, this is derived from another requirement, that we intend that we J “by way of relevant mutual responsiveness in sub-plan and action,” that is to say, that we each be responsive to what the other is doing and planning in executing his intention to J in executing our own. The requirement of meshing subplans helps to cash out the idea that when we jointly intend to do something together, we intend to do it together intentionally, as I put it above. The trouble with this intuitive expression of the idea is that it is circular. Part of the idea is captured by the thought that to intend to do it together intentionally we each intend that we each intend it. But this does not rule out each of us intending to trick the other into doing it in a way that would frustrate the other’s plan for how to do it. The idea that we intend not merely that we do it, or do it by way of our intentions to do it, but by way of meshing subplans associated with them, seems to capture a minimal sense of essential cooperativeness in joint intentional action. This is, in my view, one of the central insights of the account.
What role do conditions B–D play? Condition B ensures the participants see themselves as (i) interdependent, so that they do not think of themselves as going it alone, but also as (ii) individually and jointly practically effective. C ensures that these beliefs are true. D helps to ensure that members of the group will function together as a deliberative body about how to carry out their joint intention. B–D together help to ensure that the structure of intentions expressed in A function in a way analogous to individual intention in individual action. Finally, E expresses the satisfaction conditions for the shared intention.
Now I turn to some questions about the account. Although B–D play a role in spelling out the idea that shared intention is a structure of interrelated intentions in a group that functions analogously to individual intention, it is not clear that, even in terms of the account itself, they should be seen as necessary for shared intention, or even necessary parts of a sufficient condition for shared intention. Consider the question whether condition A entails conditions B–D. It seems clear that it does not. Moreover, it does not appear to be a condition on rationally having intentions in A that B–D obtain. It may be rational for members of a group to form intentions of the sort identified in A, in a high stakes context, though their degree of belief in success is quite low, and their degree of belief in persistence interdependence is low, and in fact the intentions are not persistence interdependent, though they do persist. And it is clear in these conditions they would not have common knowledge of conditions A–D. But now suppose that their intentions in A are in fact satisfied. Then they J-ed together. Did they not thereby J together intentionally? If not, it is difficult to see how adding B–D could make any difference. But if so, assuming that shared intention is what, when satisfied, results in collective intentional activity, B–D are not necessary for shared intention, nor necessary parts of a sufficient condition. A alone would give conditions sufficient for shared intention.
It might be said that even if B–D are not necessary, nor necessary parts of a sufficient condition, for shared intention, if A is sufficient, we still have a sufficient condition if we add B–D. That’s true. And we would have a sufficient condition if we added that all the members of the group wear orange hats, or are members of the Rotary Club. But B–D, of course, relate to A more intimately because they support the stability of the structure of intentions it expresses. Similarly, of course, knowing that one can do what one intends supports the stability of one’s individual intentions. We could then analogously say that there is no harm in giving as a sufficient condition for having an individual intention that one intends to do something and knows that one can. But there is no point to doing that. What we want is a minimally sufficient condition, for this is what gives us an account of what constitutes having a shared intention. The extra conditions are not necessary conditions on shared intention, or necessary parts of a sufficient condition. (We could also add a requirement that shared intention involve mutual obligations among members to continue to contribute; but as Bratman rightly says, this should not go into an account of shared intention, because although present in some cases, it is inessential in the sense that what makes the shared intentions it supports shared intentions has nothing to do with the fact that it supports them.)
To identify shared intention with what (minimally), when satisfied, results in collective intentional activity is to identify it with what is expressed by  on its collective reading, for if an intention of the sort expressed is carried out, then the group in question does the thing in question together intentionally. But one could say that shared intention, of the sort that the project aims to characterize, should not be identified with what (minimally), when satisfied, leads to collective intentional action, and so should not be identified with what is expressed by  on its collective reading. Rather, as the going in idea was to capture a structure that functions analogously to individual intention, we need to add further constraints to ensure that we capture something about a group that plays a more robust role in guiding action and planning than A alone provides. However, individual intention is what (minimally), when satisfied, leads to individual intentional action, and if we are looking for its analog in the case of groups, it would seem that we would want what we construct to meet this condition. In addition, individual intention can guide action even when one’s degree of belief in success (and even in its persistence) is low. To preserve the analogy, then, it would seem that we would want to find something that functions analogously in the case of group action. Is A enough, though, to support norms of social agglomeration, consistency, coherence and stability? There is of course still pressure from the normative requirements of the individual intentions in this structure for social agglomeration, consistency, coherence and stability. So we do not need more to explain the source of these norms in the social case. What B–D add here is the analog of true belief in the stability of individual intention and belief that success conditions obtain, which helps put us in a position to implement the norms.
It might also be said that the target is not collective intentional activity in general, but instead modest sociality, and so not shared intention per se but shared-intention-together-with-the-features-it-has-in-the-case-of-modest-sociality. Modest sociality is introduced initially with a list of activities (of the sorts given above). Its further specification comes from the conditions that are given for its expression in A–E. This raises the question whether there is an independent target in terms of which we could assess the account. It may be suggested that it is a kind of paradigm or prototype of collective intentional activity, idealized in certain ways, and that other forms are understood in relation to it. But it seems to me that, if we are after what makes even modest sociality collective intentional activity, which is required to vindicate the continuity thesis, we are looking for a minimal sufficient condition for shared intention.
Another consideration in favor of seeing shared intention as constituted more minimally than A–D is that collective intentional activity in general should be seen as the result of shared intentions being satisfied. It is implausible to suppose that in different cases – for example, one in which condition D is not satisfied – what constitutes shared intention is different.
With this in mind, I want to turn to look more closely at condition A. Suppose that we are interested in what constitutes shared intention, that is to say, in minimally sufficient conditions (if not in necessary and sufficient conditions), so that we do not include conditions not needed to constitute shared intention. Are all the requirements in A necessary parts of a sufficient condition? Is A in fact sufficient?
The requirement of mutual responsiveness is not necessary for collective intentional action, as Bratman points out. There is the case, as he puts it, of the Ayn Rand singers, who sing a duet together intentionally, but do not intend to help the other if he or she falters (“I am not a servant of their needs”). There are also cases of joint intentional action in which mutual responsiveness is off the table, as when members of a team of saboteurs are instructed (or agree) not to try to contact each other to reduce the risk of discovery on their mission, or when their conditions are not met, as in Bratman’s example of synchronized diving, or people carrying out parts of joint tasks at locations remote from one another. This suggests that the requirement of meshing subplans should be put back in as a basic component. Then we can ask whether our each intending that we J by way of corresponding intentions and meshing subplans associated with each provides a minimally sufficient condition for shared intention.
Here a question arises about how to interpret the requirement of meshing subplans: all or some, and if some, how much mesh is required? To show that in general meshing subplans all the way down is not required, Bratman cites the example of competitive games like chess, in which each player has subplans that require for their success the frustration of corresponding plans of the other. For in intending to play chess, we intend an activity in which it is constitutive of the play that each player pursues divergent aims with respect to the final distribution of pieces on the board (that is, with respect to who gets checkmated). But for that reason, it is also unclear that the intentions these divergent aims give rise to are incompatible with any of the subplans aimed at realizing an instance of the action type: playing chess. For your carrying out successfully your plan to win does not frustrate any of the subplans I formed pursuant to the goal of playing chess (as opposed to my winning). So we can specify that the subplans associated with the intentions that are to mesh all the way down are those pursuant to instantiating what the shared intention aims at realizing.
Still, meshing subplans all the way down seems too strong a requirement in general. Suppose Sue and Jake intend to paint the house together and agree that they will each paint half, starting on opposite sides, painting clockwise, but miscommunicate a bit, and Jake starts a ways clockwise of where Sue thinks he is to start. He paints until he comes to where Sue begins. Sue paints until she comes to where Jake began (she may or may not recognize at that point that their plans do not mesh). Plausibly their intention to paint the house together has been satisfied and they painted it together intentionally. But not because of and by way of subplans that meshed all the way down. If the content of their shared intention is to match its satisfaction conditions (one could deny this), we should not think of it as requiring that their subplans mesh all the way down. We need, we might say, a kind of margin for error, built into the content of the shared intention.
In Shared Agency, Bratman aims to sidestep this issue by focusing on the goal of providing sufficient conditions for shared intention, and so specifies that in his construction the parties to shared intention intend their subplans to mesh all the way down. I have two concerns about this. One is connected with the theme I have been pursuing, that it does not give minimal sufficient conditions for shared intention. It is rather like asking what makes for a good person and being told what it takes to be supererogatory. But perhaps for the purpose of establishing the continuity thesis this is not a problem. If you want to know whether it is possible to be a good person, and I show you that supererogation is possible, and it follows from this that someone is good, I’ve given you what you asked for. A second concern, though, is connected with the case of Sue and Jake. The potential for what happens in their case seems to be omnipresent. When there are failures of mesh, but the joint action is completed in the spirit in which it was undertaken, we still hold that the participants did it intentionally. But then we should not read into the content of their intentions that their subplans mesh all the way down – ever. But could that be the form of such an intention? Yes. And this may be enough in principle for the continuity thesis. But would the account of modest sociality then capture much of the territory that it was intended to cover?
If this is right, it does not seem to me to constitute any fundamental criticism of the model, but instead simply to require adjusting the setting of a knob here or there, and it is intended in that spirit. How to adjust the knobs, though, is not so clear. One suggestion is to require not that the parties intend that they J by way of meshing subplans all the way down but require that they intend to J by way of meshing subplans, and that neither have any subplans that aim to subvert subplans (pursuant to the joint goal) of the other. If playing chess involved having subplans to carry out the intention to play chess that aimed to subvert the other’s subplans to carry it out, then this would be too strong. But that turns out not to be what is going on in chess. But it still seems too strong, for it would rule out cases in which someone working with another occasionally does something to subvert his contributions (maybe on the sly) – to keep the other from outshining him, for example – but never so much as to prevent their succeeding in their joint goals (think about cases of sibling rivalry). Here, it seems to me, they do the thing together intentionally, and intended to do it together. But we cannot say that it is enough that they intend to do it by way of their intentions that they do it and just some of their associated subplans. Two mafia wise guys from rival crime families in Miami may each be ordered to kidnap the other and take him to New York. They have each infiltrated the other’s organization, and know that each plans to invite the other to a certain deserted lot to overpower the other and put him in the trunk of his car, and this is common knowledge between them. Each of course is confident he will succeed. They each intend that they go to New York together in part by way of corresponding intentions and meshing subplans – right up to the point that their intentions bring them to the deserted lot. But they do not share an intention to go to New York together. So we are left with a question about how to draw the line between this last case and a case like the imagined one of sibling rivalry. We want something in the ballpark of their intending to be on the same page, at least at the time of action, with respect to what their roles are in what they are doing, within a certain margin of error.
To summarize, I agree with Bratman about all the major theoretical claims he makes in Shared Agency. I have raised several questions. First, I raised the question, internal to the account, whether conditions B–D should be seen as necessary parts of a sufficient condition for shared intention. If the intentions in A are satisfied, does that not result in joint intentional action, and, hence, is not A all that is needed for shared intention (and if not, how can B–D help)? Second, I raised the connected question what the relation is between the notion of shared intention expressed in A–D and that expressed on the collective reading of a sentence such as . If we think of shared intention as what, when satisfied, results in collective intentional activity, surely we get at what constitutes shared intention by specifying what the truth conditions are for sentences like  on their collective readings. This is also the typical target of analyses of shared intention – our intending to do something together, to line dance, to sing a duet, have a conversation, lift a piano, and so on. Third, focusing on condition A, I have asked whether mutual responsiveness should be seen as the ground of the requirement of meshing subplans, or whether that requirement, even if required by mutual responsiveness, is not more basic. Fourth, I have asked whether we should think it characteristic of shared intention, even in cases of modest sociality, that participants intend mesh all the way down. And I have asked, if we put this aside, how we can correctly characterize the degree and kind of intended mesh characteristic of shared intention.
Bratman, M. (2014): Shared Agency: A Planning Theory of Acting Together. Oxford: Oxford University Press.
Ludwig, K. (2007): “Collective Intentional Behavior from the Standpoint of Semantics”. In: Noûs 41. No. 3, p. 355–393. [Web of Science]
Ludwig, K. (2014a): “The Ontology of Collective Action”. In: S. Chant, F. Hindriks & G. Preyer (Eds.): From Individual to Collective Intentionality. Oxford: Oxford University Press.
Ludwig, K. (2014b): “Proxy Agency in Collective Action”. In: Noûs 48. No. 1, p. 75–105. DOI: 10.1111/nous.12013. [Crossref]