The self-reinforcing nature of joint action

Abstract

Shared intention normally leads to joint action. It does this, it is commonly said, only because it is a characteristically stable phenomenon, a phenomenon that tends to persist from the time it is formed until the time it is fulfilled. However, the issue of what the stability of shared intention comes down to remains largely undertheorized. My aim in this paper is to remedy this shortcoming. I argue that shared intention is a source of moral and epistemic reasons, that responsiveness to such reasons on the part of each individual reinforces her own relevant attitudes, and that this enhances the stability of the shared intention as a whole. We thus discover an important self-reinforcing and stability-enhancing mechanism at the heart of shared intention and action. It follows from this proposal that the psychological commitment that an individual exhibits in joint action is normally a function not only of her own participatory intention, as often maintained, but also of her endorsement of the reasons created by her group’s shared intention to so act.

1 Introduction

Suppose that you and I have just decided, and thus have formed a shared intention, to dance the tango together this weekend. Normally, our shared intention will lead us to dance the tango then. It will do this partly by inducing certain forms of coordination between us –for example, about who will lead and who will follow. Yet our shared intention will be able to play these coordinating roles, it is commonly said, only because it is a characteristically stable phenomenon, a phenomenon that tends to persist from the time it is formed until the time it is fulfilled. To coordinate our thought and action, we must be able to count on each other in relevant ways. I could not sensibly count on you if I knew that you would leave me hanging on the dance floor on a whim. But since we have a shared intention, it seems that I can sensibly count on you. For I know that your commitment to our joint action is stable enough to persist throughout the process. And vice versa.

But what, we might want to ask, does the stability of shared intention amount to, exactly? The issue poses problems for familiar views of this phenomenon. According to one of them, the stability of shared intention reduces to the stability of the constituent intentions of each. Unlike mere whim or desire, an individual’s intention possesses a certain inertia: it tends to persist so long as the individual does not acquire stronger reasons to do otherwise, that is, new reasons that outweigh the reasons on the basis of which she formed and presently maintains that intention. Accordingly, this view suggests, when you and I have a shared intention to dance the tango, each of us is disposed to stick to her individual intention so long as her original reasons for forming that intention are not outweighed by competing new reasons to do otherwise. This is an insightful perspective, but is ultimately unsatisfactory. It is clear that the stability of shared intention is partly a function of the stability of the intentions of each. But it is also clear that the stability of the former phenomenon does not reduce to the stability of the latter. Shared intention is more resistant than is individual intention to new reasons to do otherwise. This is because, unlike individual intention, shared intention itself comes along with reasons, which counterbalance and often defeat competing new reasons to do otherwise. This morning you judged that dancing the tango with me would be enjoyable. But now that you have arrived at the party and discovered that Astor is also there, you judge that dancing with him would be even more enjoyable. Yet, you also think that the fact that you and I have a shared intention to dance together puts pressure on you to stick to the original plan. In the end, you decide to stick to our shared intention. That makes sense. But, if so, it seems implausible to maintain that the stability of your commitment to our joint action is to be explained solely by reference to the inherent inertia of your constituent intention.

Another influential view of shared intention offers, in contrast, a nonreductive perspective on the issue. On this view, the stability of this phenomenon is determined by its purportedly special extinguishing conditions. While the intention of a single individual can be extinguished by a mere change of mind, the view holds, the shared intention of a group of individuals can be extinguished only by all such individuals together. Thus, were you to unilaterally change your mind and decide to dance the tango with Astor instead, our shared intention would not be extinguished as a result. Or so this view maintains. It is surely plausible to hold that shared intention exerts greater pressures towards stability than does individual intention. And it is also plausible to hold that, in the normal case, it is both correct and rational for each individual to seek permission from the other before deciding to abandon the shared intention. But the suggestion that it is impossible for any single individual to unilaterally destroy the shared intention seems to me hard to sustain.¹

Although central to the forms of interpersonal coordination and dependance characteristic of joint action, the issue of the stability of shared intention remains largely undertheorized. My aim in this paper is to remedy this shortcoming. Our main question, to be more precise, concerns the rational stability of shared intention, that is, the question of why, if rational, each individual participant will be disposed to stick to the shared intention until the joint action is carried out. This raises in turn a more fundamental question about the nature of shared intention as such and, thus, a question about the kinds of metaphysical structures and relevant norms that should be brought to bear in order to account for its rational stability. In previous work, I have argued for a dual aspect view of this phenomenon (2009, 2016a). According to it, shared intention involves a structure of attitudes of individuals that includes not only intentions of each but also, and importantly, attitudes of reciprocal reliance. This structure of attitudes, I have contended, both guides each individual’s thought and action in the direction of the execution of the joint activity and is a basis for –i.e., defeasibly creates—moral obligations between such individuals. Shared intention thus exhibits two main aspects: psychological and moral.² In this paper, I propose an account of the rational stability of shared intention that builds on that view.

My proposal involves a trio of interrelated theses. First, I claim that shared intention is a source of reasons, practical and epistemic. Not only does shared intention defeasibly give each individual a moral, obligation-based reason for action. It also gives each a reason to believe that the other will eventually do their part in the joint activity. Second, I hold that each individual’s endorsement of and responsiveness to such reasons reinforce her own attitudes of intention and reliance: they provide further justification for such attitudes and induce in them a tendency to persist across a wider range of circumstances. Third, I claim that this reinforcement of the attitudes of each accordingly enhances the stability of the shared intention as a whole. It follows from this trio of theses that a self-reinforcing and stability-enhancing mechanism lies at the heart of shared intention and action. In advancing this trio of theses about the rational stability of shared intention, in this paper I aim to describe a rational psychology that is stable under reflection, that is, the psychology of a group of individuals each of whom is appropriately responsive to considerations that are in fact normative for her relevant attitudes.

My paper is organized as follows. Section 2 offers a brief outline of the dual aspect view of shared intention and defends the first of the trio of theses mentioned above. Section 3 elaborates on the key notion of the reinforcement of intention and thus provides part of the conceptual framework that informs the second and third theses. Section 4 defends these two theses. Section 5 concludes by calling attention to two corollaries of the present proposal: the emergence of a novel conception of the commitment characteristic of joint action as well as of a more precise characterization of the aforementioned contrast between the stability of individual and shared intention.

2 Shared intention as a source of practical and epistemic reasons

According to the dual aspect view, the shared intention of a group of individuals involves a complex set of attitudes of each of them. Three different kinds of attitudes are central in this set. First, shared intention involves an intention on the part of each individual in favor of the group’s activity. Since you and I have a shared intention to dance the tango, it follows that I intend that we dance the tango and that you intend likewise. It matters that the attitude that each individual has toward the joint activity is one of intention, rather than ordinary desire. Intention is a conative attitude that tracks, or seeks to bring about, some action or result, in ways in which desire does not. Intention plays such an action-tracking role by disposing the agent to, inter alia, act when the times comes, form further intentions about means, and, as mentioned in Sect. 1, retain this attitude in the absence of strong reasons for reconsideration.³ But desire need not involve such a cluster of dispositions. When a group of individuals have a shared intention to do something together, there is some action or outcome that each of them tracks –rather than merely desires or hopes to attain. What each individual tracks is the success of the joint activity itself.⁴

Second, shared intention also includes attitudes of reciprocal reliance between individuals. This is a central claim by the dual aspect view, with important ramifications. In this view, reliance enters into shared intention and action at a very early stage. Reliance on the other is a condition of possibility of the intention of each. It is an intuitive idea that the action-tracking quality of intention imposes a cognitive constraint on the intending agent. The constraint is that an agent may only intend what she takes her so intending as actually settling, or bringing about. An agent cannot intend to do something –say, to run a marathon—if she does not take her so intending as settling the issue of whether she actually does that thing, that is, as being causally efficacious in bringing about that result. Intentions in favor of the joint activity are ordinary intentions of individuals and seem as such subject to this constraint, too. However, such intentions of individuals also have a special target –they track the group’s action rather than just the individual’s own—and this raises a question as to whether an individual can sensibly see her own intention as settling the issue of what her group does. So, how can I see my own intention that we dance as settling the issue of whether we dance? The answer to this question builds on a premise about the causal efficacy of the intentions of each. Barring obstacles, you and I will dance the tango together if and only if I intend that we dance and you intend that we dance. In other words, normally, the intentions of each are individually necessary, and jointly sufficient, to bring about the joint action.⁵ Therefore, it is plausible to suggest, in light of this premise, that in order for me to see my own intention as settling the issue of our dancing together, I must see you as intending likewise and your intention as playing a similar causal role in settling that issue. But to see you in this way is nothing but for me to rely on you to so intend and act.⁶ In shared intention, each individual relies on the other to intend likewise and to do her part in the joint activity.

This connection the dual aspect view establishes between intention and reliance gains further support from an independent theory of the latter phenomenon (Alonso, 2014, 2016b). According to this theory, reliance is a cognitive attitude whose main function is to serve as a fixed point on the basis of which an agent may deliberate, make plans, and act. This function of reliance supports the connection established above. For it is partly against the cognitive framework provided by my reliance on your (eventual) intention and actions that I get to form my own intention that we dance. Clearly, in providing this general sort of cognitive guidance, reliance resembles belief. But reliance also differs from belief in important ways. A crucial difference is normative. While belief is correct just in case it is true (Williams, 1973; Shah, 2003), reliance is correct just in case it cognitively guides the agent’s reasoning and action in ways that are instrumental to her relevant ends and values (Alonso, 2016b). The correctness of belief is thus a purely epistemic matter, a matter of whether this attitude accurately represents the facts. The correctness of reliance, by contrast, is usually of a mixed sort. It is fundamentally a practical matter, since it has to do with whether this attitude is instrumental in promoting, or partly constituting, some relevant end or value. And it is usually an epistemic matter, too, since the instrumentality of this attitude is in many cases a function of the truth of what is relied upon. This theory of reliance therefore gives us an insight into the conditions under which it is correct for an individual to rely on another for the purposes of joint action. My reliance on your intending that we dance will be correct just in case it is conducive to our eventually dancing the tango, which is something I value or care about; and it will conduce to this only if you actually intend that we dance and my reliance accurately represents this fact.

Finally, in shared intention the cited attitudes of intention and of reliance of individuals are common knowledge or “public” between them. It is reasonable to assume that in this context individuals get access to each other’s minds, and thus eventually arrive at such common knowledge, through reciprocal acts of signaling or communication between them, intended or otherwise (Alonso, 2009, 458, n.46). Perhaps you responded “yes” to my invitation to dance; or maybe you signaled to me, just by entering the dance floor, that you were ready to engage. Several rationales have been offered for including a common knowledge condition in our account of shared intention and action. It is plausible that common knowledge of the attitudes of each be necessary in order for a group of individuals to arrive at shared intention, or sufficient to cognitively frame certain forms of planning and coordination between us.⁷ The dual aspect view is compatible with such rationales, but also calls attention to another, having to do, as we will see below, with the normative implications of such common knowledge.

This view accounts for central features of shared intention. It is apparent, we mentioned above, that shared intention brings about joint action partly by coordinating the thought and action of individual participants. The dual aspect view holds that relevant attitudes of intention and reliance of each are largely responsible for the performance of such coordinating roles. Given that each of us intends that we dance the tango, each will be disposed to form subsidiary intentions about means. And each will be disposed to form such subsidiary intentions in light of her assumptions about the subsidiary intentions and eventual actions of the other. It is partly because you rely on my intending to lead and on my actually leading the dance, that you intend to follow and will follow when the time comes. It is also partly because I rely on your intending to follow and on your actually following my movements, that I plan to lead the dance and will do so at the right times. And it is because each of us so intends and relies, basically, that we normally come to dance together.⁸

The dual aspect view helps to explain another important feature of shared intention. Suppose that you attend the tango party and I fail to show up. Suppose that you discover, after the fact, that I had unilaterally decided to change plans. It seems that in such circumstances it would be appropriate for you to resent me, and for me to feel guilt, for what I have done. What makes such reactions appropriate is the fact that in such circumstances I had an obligation to you to act in a certain way, which I simply decided to disregard. As Margaret Gilbert has emphasized (1997, 2009), cases such as this strongly support the idea of the existence of a tight connection between shared intention and interpersonal obligation. The dual aspect view offers an account of this connection (2009). It claims that shared intention defeasibly creates moral obligations of each to the other. The argument for this claim can be given in three steps. First, we recall that in having a shared intention to dance the tango, you and I have common knowledge of the attitudes of intention and reliance of each. Second, we note that an important transaction takes place in such a context, a context in which the attitudes of each are public between us. In intending that we dance, I thereby signal to you that I so intend and that I will act accordingly when the time comes; and, in signaling you this, I reinforce your reliance on my relevant intention and eventual actions. And vice versa. Finally, we argue, building on ideas by Scanlon (1998) and MacCormick (1982), among others, that in reinforcing each other’s reliance in this way we normally incur relevant moral obligations to one another. In particular, since in sharing an intention to dance the tango with you I have reinforced your reliance on my intention and eventual actions, and since I have good reason to believe that you will suffer relevant losses –of, say, time, energy, and resources—if I do not act as relied upon, in that context I incur an obligation to you to prevent such losses. This obligation is non-strict, in that it does not demand strict performance. For I may comply with it, and thus prevent such losses, not only by performing as relied upon, but also, for example, by giving you a timely warning that performance is not forthcoming or by providing compensation for the losses I might have created in you.⁹ And vice versa.

Clearly, the foregoing remarks serve to elucidate the sense in which shared intention is, according to the dual aspect view, both a psychological and a moral phenomenon.¹⁰ But, importantly, such remarks do more than this. They also provide support for the first of a trio of interrelated theses about the rational stability of shared intention I defend in this paper, namely, the thesis that shared intention is a source of practical and epistemic reasons. It is plain from the normative story outlined in the paragraph above that shared intention is a source of obligation-based reasons for action. Yet, further reflection on the common knowledge condition mentioned above indicates that shared intention is a source of reasons for belief, too. When you and I arrive at our shared intention to dance the tango, a radical shift is produced in the access that each of us has to the attitudes of the other. Attitudes that were perhaps previously lodged in the private sphere of our own minds now become public between us. We come to have knowledge –and, indeed, common knowledge—of the attitudes of intention and reliance of each. This has important normative implications. Consider a crucial step in the process leading to such common knowledge: my gaining knowledge of your intention. One way to register the normative significance of this epistemic achievement is in terms of the notion of a (normative) reason. Not only does this knowledge clearly give me, or constitute, a conclusive reason to believe that you intend that we dance. It also gives me a reason to believe that you will eventually follow through on your intention and do your part. For, to gain knowledge of your intention is to get to know, first, that absent strong reasons for reconsideration, your intention will tend to persist until the time of action and, second, that if your intention persists and nothing interferes, you will try to execute your intention then. Therefore, we see that shared intention is a source of epistemic reasons in that it provides each individual with a conclusive reason to believe that the other has the relevant intention as well as with a (non-conclusive) reason to believe that they will do their part.¹¹

The thesis that shared intention is a source of practical and epistemic reasons, I said in Sect. 1, figures as a premise in an argument for the other two theses I defend in this paper. These are, recall, the thesis that in shared intention the intention and reliance of each are reinforced by the practical and epistemic reasons created by the former phenomenon –which we identified in the paragraph above—and the thesis that the stability of this phenomenon is enhanced as a result of such a process of attitude reinforcement. In the rest of the paper, I motivate and defend these two theses.

Before I proceed, two clarifications are in order. First, the second thesis advances in fact a more precise claim about the reasons that influence the attitudes of each in shared intention. It claims that while the reliance of each is reinforced by the abovementioned reasons for belief, the intention of each is reinforced both by such reasons for belief and by the abovementioned obligation-based reasons for action. I defend this thesis in Sects. 3 and 4 below. Second, this two-part thesis rests on two corresponding assumptions about the normativity of such attitudes, respectively, that epistemic considerations –i.e., evidence of truth—can be normative for reliance, and that both epistemic and practical considerations can be normative for intention.¹² These assumptions, if true, underwrite the idea that the process of attitude reinforcement we have associated with shared intention is stable under reflection, that is, that for an individual participant to adjust her intention and reliance in response to such considerations is for her to adjust her attitudes in response to considerations that are in fact normative for them. The preceding observations about the norm of correctness for reliance support the former assumption. Evidence of truth can be normative for reliance. The latter assumption, however, requires some elaboration and defense. The idea that practical considerations can be normative for intention seems uncontroversial. But the idea that evidence also can be normative for it appears less so; at least, it does so from the perspective of a conativist conception of this attitude of the sort endorsed here.¹³ In the next section, I provide support for such an assumption. I also elaborate on the general idea of the reinforcement of intention, which is key to the discussion of the stability of shared intention in Sect. 4.

3 Some preliminaries: on the reinforcement of intention

The idea that practical reasons for action can be normative for intention finds support in our ordinary notion of intention and in the idea that practical reasons can transmit from ends to means.¹⁴ Part of the point of intention, we said, is to causally bring about action. It is to serve as a means to a relevant end.¹⁵ The fact that an action is valuable is a consideration in favor of bringing it about. It is also –given the instrumental transmission of practical reasons and the point of intention—a consideration in favor of intending to bring it about. More schematically, we may say that an agent has a reason for intending to φ if (a) her intending to φ is a means to her φ-ing and (b) she has a reason for φ-ing (cf. Kolodny, 2008, 2018).¹⁶ Reasons for intention can therefore be instrumental in structure –since they derive from reasons for the intended action—and practical in nature –since the reasons from which they derive are practical reasons.

This conception of reasons for intention also helps to explain why evidence can be normative for this attitude, too. On this conception, an agent’s reason for intending to φ can be partly constituted by the more basic fact that her intention to φ is a means to her φ-ing. But to say that an agent’s intention to φ is a means to her φ-ing is to say, roughly, that there is nonzero probability that the agent will φ as a result of that very intention (Kolodny, 2018). Clearly, this probability assessment is based on facts that are evidence for and against the proposition that if the agent intends to φ, she will φ in fact.¹⁷ Therefore, we see that in such a case the agent’s reason for intending to φ is partly constituted by facts that are evidence for the proposition that her intention to φ will lead to her φ-ing. These remarks suggest a specific view of the normative bearing that evidence of future success can have on intention. This view involves two main ideas. The first is that such evidence cannot be an independent reason for this attitude. The mere fact that intending to act in a certain way will be causally efficacious does not by itself speak in favor of so intending. After all, the agent may have no reason for,  or even have reasons against, acting in that way; perhaps performing the action in question is silly, imprudent, or immoral. The second, and more positive, idea is that although evidence of future success cannot be an independent reason for intention, it can nonetheless be partly constitutive of an instrumental practical reason for it. It follows from this view that evidence can be normative for intention, without constituting an independent reason for it. There is more to be said in defense of this view, but the underlying intuition should be clear. Whether or not an agent’s intention will lead her all the way down to action is a relevant fact for her to consider when deliberating about whether to form, retain, or abandon that very intention. I may really want to go skiing this weekend. But if I know that there is not enough snow yet, I might not have a reason to form the corresponding intention.

Finally, it is useful for the purposes of the forthcoming discussion to reformulate the foregoing points about reasons in terms of the associated notion of justification. We said above that a reason for intention can be partly constituted by evidence of future success and by practical reasons for action. Similarly, we may say that an agent’s intention to φ can be justified on the basis of both epistemic reasons for believing that she will φ (as a result of her intention to φ)¹⁸ and practical reasons for φ-ing.¹⁹

Now we are in a position to flesh out the idea of the reinforcement of intention. Suppose that I am deliberating about what to do this evening. I know that running contributes to my overall health. I judge that this gives me a reason for running 5 miles then and, on the basis of this judgment, form the intention to do so.²⁰ Soon after this, you offer me a significant monetary reward if I do run that distance, and I judge that this gives me an additional reason for running those miles. Initially, I endorsed certain health considerations as normative for my intention. But now I endorse some financial considerations as normative for it as well.²¹ This raises a question. We know how the first reason influenced my practical reasoning. It provided some justification for forming, and accordingly led me to form, the aforementioned intention. The second reason seems to have an important impact on my practical reasoning, too. But what is its impact, exactly? It is plain that this reason provides further justification for the intention I have formed. But there is an additional, and less obvious, influence that this reason exerts on my practical reasoning. This latter influence concerns the rational psychology of intention.

Provided that I am rational –i.e., appropriately responsive to relevant considerations—my endorsement of this second reason as normative for my intention enhances the stability of this attitude. Suppose that after a long day at the office I start to wonder whether stopping by the local bar on the way home is not more desirable than going for a run. At first, I judge that in this context considerations of entertainment outweigh considerations of health. But then I remember your financial incentive and decide to stick to my prior plan. Your offer tips the balance and accordingly plays an important role in my decision. Had I not been offered that additional incentive in that scenario, I would have likely abandoned that plan. In other words, your financial incentive has made my intention more resistant to new reasons to do otherwise and thus more robust in response to relevant changes in the circumstances. Of course, the stability of my intention to run 5 miles could have been enhanced by other sorts of practical reasons, too. Had I promised you to run that distance, for example, I would have also had an additional incentive to so act and this would have also had an impact on my intention. Indeed, we will see in Sect. 4 that obligation-based reasons enhance the stability of shared intention in a distinctive way. But what needs emphasizing here is the form of psychological influence that any such reasons can exert on an agent’s intention. When an agent endorses a reason for action as normative for her intention and is appropriately responsive to it, her intention becomes more counterfactual robust, that is, it acquires a tendency to persist across a wider range of circumstances than it otherwise would. This increase in the counterfactual robustness of the agent’s intention is a by-product of a causal and rational process. It is an effect produced in the agent’s intention, and it is an effect that goes by way of the agent’s responsiveness to a consideration she takes to be normative for her intention. Therefore, I will say that an agent’s intention is reinforced when it becomes more justified and counterfactually robust. Intention reinforcement, as here understood, has both normative and psychological dimensions.²²

The preceding observations concern the practical reinforcement of intention. But parallel observations can be made about the epistemic reinforcement of this attitude, too. This morning I read that a snow storm was expected to hit town this evening, but this afternoon I learned that it had changed course. The consideration that an obstacle to the causal efficacy of my intention to go running this evening has been removed reinforces my intention. It makes this attitude more justified and counterfactually robust. Imagine that after I found out about the weather change this afternoon, I learned also that you had decided to withdraw your financial incentive. In such circumstances, I think, I would have still persisted in my intention. I would have done so on the basis of both health considerations and evidence of future success. However, it is likely that had you withdrawn your offer this morning, when the weather service predicted a snow storm, I would have abandoned it.

Return now to the topic of shared intention and action. So far, our observations about the reinforcement of intention centered on intentions whose target is the agent’s own action. But now we need to consider the reinforcement of the intentions of individuals that are partly constitutive of shared intention and action. Our present target, therefore, is the reinforcement of an agent’s intention in favor of her group’s action. Let us concentrate first on the normative dimension of the reinforcement of such intentions. I discuss the psychological dimension of this process in Sect. 4. It is plausible to suggest, in line with our observations above, that an individual’s intention in favor of her group’s action can be reinforced both by practical reasons in favor of that action and by epistemic reasons for believing that her group will carry it out.²³

This characterization makes apparent the existence of obvious similarities in the reinforcement of intentions in favor of individual and joint action. But there is also a key difference between them. The epistemic reinforcement of an individual’s intention in favor of the joint action depends on her information about the eventual contribution of other individuals in a way in which the reinforcement of an individual’s intention in favor of her own action need not. The explanation lies partly in the causal efficacy, or success conditions, of such intentions, namely, in that the contribution of others is necessary for the success of the former intention, but not for the success of the latter. This observation suggests that we must expand on our initial characterization of the reinforcement of the former intention, so as to reflect more fully its distinctive epistemic dimension. Here it is important to identify the considerations that can constitute reasons for believing that the joint action will be carried out. And to do this, it is useful to appeal to a more comprehensive idea of the causal efficacy of such intentions of individuals, as mentioned in Sect. 2 above. The idea, recall, is that normally the intentions of each are individually necessary, and jointly sufficient, to bring about the joint action.²⁴ Put succinctly, for any joint activity φ and a group of agents consisting of you and I, it seems true that,

(*) Barring obstacles, we will φ if and only if I intend that we φ and you intend that we φ.²⁵

With this idea in hand, we can now proceed to provide the desired characterization. To focus on my own intention, we have that,

(RI) My intention that we φ can be reinforced by:

• (a) practical reasons in favor of our φ-ing;

• (b) epistemic reasons for believing that we will φ, where these include, among others,

  • (i) reasons for believing that I intend that we φ;

  • (ii) reasons for believing that you intend that we φ;

  • (iii) reasons for believing that if I intend that we φ and you intend that we φ, then we will φ.

In Sect. 2, I defended the first of a trio of interdependent theses about the rational stability of shared intention, viz., the thesis that shared intention is a source of moral and epistemic reasons. This included a defense of the claim that the epistemic reasons created by shared intention are, specifically, reasons for belief in the other’s intention and eventual actions. I also tentatively assumed that these epistemic reasons can have normative force for the intentions of individuals that partly constitute the shared intention. Now we see that the reasoning leading to schema RI above vindicates this latter assumption, as the just mentioned reasons are nothing but the reasons that figure in clause (b)(ii).²⁶

4 Attitude reinforcement in shared intention

The discussion in the previous section provides us with conceptual resources that allow us to substantiate our second thesis. While our first thesis holds, we just said, that shared intention is a source of epistemic and moral reasons, our second thesis holds, recall, that when an individual endorses and is responsive to such reasons, her own attitudes of intention and reliance are reinforced as a result. In what follows, I describe how such effects of attitude reinforcement are produced in the context of shared intention. In the previous section, we reflected on the idea of the reinforcement of an individual’s intention and it is clear that such reflections extend quite naturally to the idea of the reinforcement of an individual’s reliance as well. That noted, my main focus here is on the reinforcement of the intentions of each, for this has, as we will see next, important effects on the rational stability of shared intention.

So, suppose that you and I have just arrived at our shared intention to dance the tango. Suppose, also, that each is rational in the sense mentioned in Sect. 3 –i.e., appropriately responsive to relevant considerations. For expository purposes, let us state in schematic form what our shared intention involves.

Schema 1

The fact that you and I have a shared intention to dance the tango entails that:

(1) (a) I intend that we dance the tango.

 (b) You intend that we dance the tango.

(2) (a) (i) I rely on (1)(b) and also (ii) rely on your doing your part in our dancing.

 (b) (i) You rely on (1)(a) and also (ii) rely on my doing my part in our dancing.

(3) (1)-(2) are common knowledge between us.

So, how are the attitudes of intention and reliance of each of us reinforced as a result of our having arrived at a shared intention to dance? It is plausible to say that an initial effect is produced when we acquire common knowledge of each other’s intentions and, in particular, when each acquires knowledge of the other’s intention. We said in Sect. 2 that the acquisition of such knowledge has normative significance for each of us. Importantly, my gaining knowledge of your intention gives me a reason to believe that you will do your part when the time comes and thus a reason to believe that we will bring about the joint action. But what effects, if any, we may now want to ask, does my gaining knowledge of your intention, as captured by condition (3)(1)(b),²⁷ have on my own intention, (1)(a)? The account of the reinforcement of such intentions (RI) proposed in Sect. 3 offers a straightforward answer to this. My knowledge of your intention gives me a reason to believe that you will eventually do your part and in so doing provides some epistemic justification for my own intention that we dance. For, as RI indicates, my intention is epistemically supported by reasons for believing that we will dance and these reasons include reasons for believing that you will do your part. Therefore, provided that I am rational, my discovery of your intention, (3)(1)(b), epistemically reinforces my own intention, (1)(a). Furthermore, it is plain that the acquisition of such knowledge also epistemically reinforces my reliance, (2)(a). It provides conclusive epistemic support for my reliance on your intention, (2)(a)(i), as well as some support for my reliance on your eventual actions, (2)(a)(ii). Something similar can be said about the way in which your knowledge of my intention, (3)(1)(a), affects your own intention and reliance, (1)(b) and (2)(b). We can summarize these effects as follows. Once we arrive at a shared intention to dance, the partly constitutive attitudes of intention and reliance of each are epistemically reinforced; they are so reinforced as a result of each individual’s recognition of and responsiveness to the epistemic reasons created by the former phenomenon.²⁸

It deserves emphasis that these effects of epistemic reinforcement are particularly significant in cases in which in coming to form her intention in favor of the joint activity, an individual decides to rely on the other’s intention and eventual actions, without her also being in a position to believe –or, to put it more precisely, to believe on the basis of sufficient evidence—that the other will so intend and act. Initially I had serious doubts about whether you would be interested in dancing with me but nonetheless relied on your eventually doing so and formed, partly on that basis, the intention that we dance. True, at that moment I went out on a limb, epistemically speaking, in forming such attitudes.²⁹ But now that we have arrived at our shared intention and I have discovered that you intend likewise, my intention and reliance have been substantially reinforced as a result.³⁰

As noted in Sect. 2, in addition to creating epistemic reasons for belief, shared intention defeasibly creates moral obligations of each to the other to act in certain ways.³¹ It is plausible that in having a shared intention to dance the tango, you and I have common knowledge not only of the attitudes of each, but also of some of the consequences of our having such a shared intention. This includes our having common knowledge of the obligations we have reciprocally incurred in this context. This points to a second –and perhaps most important—effect of attitude reinforcement produced by our shared intention. Namely, in this context the intentions of each are practically reinforced as a result of each individual’s recognition of, and responsiveness to, the obligation she has incurred to the other. Perhaps I formed the intention that you and I dance the tango this Saturday partly because I thought that this would be fun. Whatever reasons I originally had for dancing the tango with you, now that we have arrived at a shared intention to that effect, I have acquired an additional, obligation-based reason for so acting.³² I recognize this additional reason for action and endorse it as normative for my intention. This endorsement of reasons has two main effects. First, it provides further justification for my intention. And, second, and more importantly, it makes my intention more counterfactually robust: there is at present a wider range of circumstances under which I will tend to persist in this attitude, if rational.

In Sect. 3, I claimed that an agent’s responsiveness to an obligation-based reason enhances the stability of her intention in a distinctive way. It is important now to make good on that claim. It is a familiar idea that obligations impose more stringent restrictions on the practical reasoning of a rational agent than do ordinary prudential reasons. R. Jay Wallace articulates this idea in a perspicuous way in recent work (2019). As Wallace observes, we usually understand the role that prudential reasons play in an agent’s practical reasoning partly in terms of their relative weights. When I decide to stick to my intention to go for a run after work rather than to the alternative of going the bar, I do so basically on the basis of the judgement that considerations of personal health and financial gain jointly outweigh considerations of personal entertainment. Moral obligations, however, work differently in the practical reasoning of a rational agent. They play two interrelated roles: first, they serve as what Samuel Scheffler calls “presumptively decisive” reasons for action, in that they “present themselves as considerations upon which [the agent] must act”, rather than as considerations to be weighed against competing reasons (Scheffler, 1997, 196. My emphasis); and, second, they operate as what Joseph Raz calls “exclusionary” reasons for action (1990), in that they “defeasibly block the normative force” of competing reasons (Wallace, 2019, 26). In these ways, Wallace sums up, moral obligations function as “presumptive constraints” on a rational agent’s deliberation (2019, 26).³³

It is feasible that when I originally deliberated about what to do this Saturday, I saw our dancing the tango as one among several options. Perhaps I also saw that option as best and accordingly formed the intention that we so act. But now that you and I have in fact arrived at a shared intention to dance the tango and I have endorsed the obligation I have thereby incurred to you, my practical perspective has changed. I no longer see the option of dancing together as optional but as required, and am accordingly disposed to disregard conflicting options from consideration in my deliberation going forward. To put it in Wallace’s terminology, the obligation I have incurred to you has placed a “presumptive constraint” on my practical reasoning. But there is another important effect to highlight here. It seems that such a constraint has affected not only the input side but also the output side of my practical reasoning. For my responsiveness to such a constraint has enhanced the counterfactual robustness of my previously formed intention that we dance in a special way, in that this attitude now displays a tendency to persist even in the face of ordinary prudential reasons to do otherwise. This enriched idea of a presumptive constraint thus gives precise content to the claim that proper responsiveness to the obligations created in shared intention makes a distinctive impact on the stability of the intentions of each.

We have just seen how the intentions of each of us, (1)(a) and (1)(b), are practically reinforced as a result of the responsiveness of each to the obligation she has incurred through the shared intention. Yet, this latter phenomenon produces an additional reinforcement effect. It is plausible that in this context each of us is aware of the practical reinforcement of the intention of the other. This has further consequences for the attitudes of each. My coming to know about the practical reinforcement of your intention, (1)(b), epistemically reinforces my own intention and reliance, (1)(a) and (2)(a); and vice versa. These effects add to the initial effects of epistemic reinforcement mentioned above. Learning about your intention gave me a reason to believe that you will do your part. But the discovery that your intention has been practically reinforced gives me a further reason to believe in your eventual actions. Not only do I presently know that you have the requisite intention, but also know that your intention now rests on an additional and distinctive reason for action, an obligation-based reason. Therefore, I know that your intention has accordingly become more counterfactually robust. And, since I know this, I have more reason to believe that you will persist in your intention and do your part when the time comes.³⁴

Let us take stock now. We saw that our shared intention to dance the tango sets in motion a process of reinforcement of the intention and reliance of each. Three reinforcement effects were identified. First, we explained how each individual’s intention and reliance are epistemically reinforced by her acquisition of knowledge of the intention of the other. Second, we saw how each individual’s intention is practically reinforced by her recognition of the obligation she has incurred to the other. And, third, we described how each individual’s intention and reliance are epistemically reinforced, in addition, by her acquisition of knowledge of the practical reinforcement of the intention of the other.³⁵ The identification and explanation of these reinforcement effects constitute an argument for the second and third theses proposed in this paper. We have thus discovered an important self-reinforcing and stability-enhancing mechanism at the core of shared intention and action, a mechanism through which the constituent attitudes of each come to rest on firmer epistemic and practical grounds, and subsequently become more counterfactually robust.

In seeing this mechanism as contributing to the characteristic rational stability of shared intention, the present account runs counter to an influential account by Bratman (1993, 2014), gestured at in Sect. 1. The present account shares with that of Bratman two key ideas. One is the natural idea that the rational stability of shared intention is partly a function of the rational stability of the constituent intentions of each;³⁶ the other the idea, lucidly articulated by Bratman, that the rational stability of each constituent intention of an individual is in turn a function of her responsiveness to a relevant norm of diachronic stability, which enjoins her to persist in her intention so long as she does not acquire stronger reasons to do otherwise. However, there are two main differences. First, unlike Bratman’s account, the present account sees such a norm as representing normative relations that exist between intention and substantive reasons, practical and epistemic, rather than as a wide-scope norm of formal coherence for intention that is normative as such (Bratman, 2014, 2018). Second, and more importantly, the account on offer differs from Bratman’s in claiming that the rational stability of a constituent intention of an individual is characteristically a function of her responsiveness to the epistemic and practical reasons that are created when the shared intention is arrived at, and therefore in claiming that the formation of a shared intention makes the partly constitutive intentions of each more counterfactually robust, that is, more resistant to reasons to do otherwise than they were before the shared intention was formed.

This leads to a final comment. It seems undeniable that in many cases of shared intention the stability of a constituent intention of an individual will also be a function of the reasons on the basis of which such attitude was formed –where this includes the initial reasons such an individual had for participating in the joint activity. The present proposal recognizes this, but it also acknowledges that such reasons are prior to and independent of the shared intention itself, and concludes from this that it would be a mistake to see them as contributing to the characteristic stability of this phenomenon. On this proposal, we associate the latter with the attitudes shared intention necessarily involves and the reasons it defeasibly creates.

5 Concluding remarks

I would like to conclude with two important corollaries to the trio of theses about the rational stability of shared intention proposed in this paper. First, it follows from our remarks in the last paragraph above that normally an agent is under normative pressure to persist in her intention across a wider range of circumstances in cases of joint action than she is in cases of mere individual action. This gives content to the natural idea, mentioned in Sect. 1, that normally shared intention imposes stronger normative constraints on an agent to stick to it than does individual intention. Of course, this does not mean that shared intention is, in every respect, more robust than is individual intention. Indeed, there is a respect in which it appears to be less robust. Shared intention is a more complex phenomenon and so more can go wrong than in the individual case. A shared intention is dependent on more than a single agent’s intention, an agent can stop relying on the other, and common knowledge of the attitudes of each can be lost.³⁷ That noted, the main idea remains. Shared intention is more robust with respect to the normative pressures it exerts on an agent.

Second, a novel conception of the psychological commitment of joint action emerges from this proposal. According to it, the commitment that an agent exhibits when acting jointly with others is typically a function not only of her own participatory intention, as commonly maintained,³⁸ but also of her endorsement of the reasons created by the shared intention –especially, of her obligation-based reason—as normative for her own intention. We can properly talk of these two elements as being partly constitutive of an agent’s “commitment” to action, as this latter notion is ordinarily understood, since, as we saw, it is in the nature of an agent’s intention to track, or seek to secure, action, and it is an upshot of such normative endorsement by the agent that her intention becomes more resistant to certain reasons to do otherwise. It follows from this proposal, then, that the rational stability of shared intention is largely a function of the psychological commitment of each, thus understood.