Shared Intention and Reasons for Action

1.1. The (Implicit) Guiding Assumptions for Accounts of Shared Intention and Why They Are Plausible

Let us start by looking at the assumptions that any theory of shared intention must grant and that most views do grant:

So why might most theories of shared intention accept Claims 1 and 2? Accepting Claim 1 does not require that one finds the causal theory of action, which identifies reasons as the relevant antecedent conditions for an action to be intentional, to be plausible. Minimally, it requires one to accept that intentional actions, when they are done for reasons, can be done for reasons of either the explanatory or the justifying type without the type of reason affecting their characterization as intentional actions. An explanatory reason is one that merely explains the action undertaken by an agent without necessarily rendering the action a genuinely good or preferable one to undertake. A justifying reason, by contrast, renders the action good, preferable, or worthwhile. ¹⁹ Of course, explanatory reasons may also be justifying, but they need not be. Typically, theories of intentional action do not require that an agent have a justifying reason in mind for her action to be intentional. The distinction between the two types of reasons is supposed to highlight that agents undertake actions on the basis of considerations that would not necessarily pass muster when subject to evaluation, but nonetheless motivate the agent to undertake the action in question in the first place. As the agent has the same relationship with her actions in both cases (with few exceptions), the type of reason is irrelevant for determining the intentional character of the action.

It is worth noting that, for views such as Bratman’s, actions are not intentional because they are done for reasons; rather, intentions are, for Bratman, a feature of our capacity as planning agents. Here, one might worry that this dissimilarity among accounts regarding the role that reasons play in intentional action forces us to have doubts about whether proponents of these accounts would accept Claims 1 and 2. If not, then there is no problem for these types of accounts of shared intention.

For my argument to go through, however, it is not the case that all views of shared intention must agree on either what intentions are or whether reasons play a central role in making an intention (or intentional action) what it is. My point is that as long as theories of shared intention take acting for reasons to be one important and necessary feature of both individual intentional action and shared intentional action, then they must also grant that whatever is true of individual intentional actions done for reasons will also be so for shared intentional actions done for reasons.

The upshot is that these accounts are alike in their respective view about the relationship between reasons and intentions, not about other features that are characteristic of intentional agency. Even if, say, Bratman and Velleman, differ with regard to their views about what intentions are and the role that reasons play in intentional action, they both share the basic commitment to the truth of Claims 1 and 2 above.

In this regard, Claim 2 appears to be independently plausible and, like Claim 1, does not require that one favors any particular view of intentional action. It simply suggests that if shared intentional action is possible, then the same characteristics that make individual actions intentional also make instances of shared action intentional or, at least, they are necessary conditions for the latter.

Why ought we to take Claim 2 to be independently plausible? First, many accounts of shared intentional action make use of this assumption to guide their accounts. As an example, consider Velleman’s ([1997] 2000, 205) claim that accounts of shared intention must be able to explain how shared intention is possible while granting the following: individuals exert control over their own intentions, they each settle whether to act on their respective intentions and their actions remain their own. ²⁰ Here it is clear that Velleman is arguing that shared intentions must have most of the defining features of individual intentions, where those defining features are individual control and the individual ability to settle on one’s own actions. Although Velleman’s purpose in mentioning these conditions is to identify a tension in many accounts of shared intention, I believe that we can infer from the conditions he proposes that he takes there to be a strong family resemblance between shared and individual intentions.

One may object here that there is no mention of the defining feature on which I focus in this article—namely, the reasons on the basis of which individuals form intentions—in Velleman’s discussion. Furthermore, one may insist that nothing about the conditions that Velleman does mention entails any conclusions about whether the kinds of reasons in light of which individual agents form individual and shared intentions contribute to their family resemblance.

It is true that Velleman’s argument does not entail any direct conclusions about whether the reasons in each case partially constitute their family resemblance. Nonetheless, we also have no compelling reason to exclude reasons as one of those features on which the family resemblance between individual and shared intention is based. More important, we have a compelling reason to include them. The conditions that Velleman does mention—both individual control and an agent’s ability to settle on her own course of action—can be categorized as features of an individual’s relationship with her own intentions and actions. Her reasons for forming an intention constitute an additional, essential aspect of this relationship. Thus, we have a prima facie reason to include rather than exclude the status of the reasons on which agents act as one of the features that constitutes the family resemblance between individual and shared intention.

Bratman’s (1999) work reflects a similar tendency to consider individual and shared intentions to have a basic family resemblance. In light of Velleman’s ([1997] 2000, 205) concerns discussed above, Bratman (1999, 148-52) addresses how his original (1993) account of shared intention is consistent with the claim that individual and shared intentions share many of the same features. Bratman (1999, 148-52) precisifies these conditions as what he calls the “Settle,” “Control” and “Own Action” conditions. For the purpose of understanding how Bratman, like Velleman, takes there to be a family resemblance between individual and shared intentions, it is sufficient to consider only the former two conditions. ²¹ The Settle condition, according to Bratman (1999, 149), requires that that any coherent account of shared intention ought be able to explain how individuals can both share an intention and yet continue to settle their own individual minds. Bratman (1999, 148) understands the Control condition in a similar vein. To meet this condition, accounts of shared intention must explain how agents are nonetheless in control of their respective actions even though the actions are, themselves, shared.

For my purposes, it is sufficient that we understand why Bratman would take these conditions to set limits on any account of shared intention rather than whether he is successful in meeting them. ²² Like Velleman, Bratman too seems to take much of what holds for individual intentions to hold for shared intentions, thereby indicating that Claim 2 above is one that many accounts of shared intention either implicitly or explicitly endorse.

A careful reader might note here that there are two concerns to consider. First, it is not clear that either Velleman’s or Bratman’s claims entail that the status of the reasons for individual and shared intention must be the same. One might claim that though many of the features of individual intentions are also features of shared intention, the status of the reasons that one has for undertaking each one need not be one of those features. Second, even if the proponents of two very different views of shared intention take there to be a family resemblance between the two types of intentions, this does not entail that there are independent reasons to grant that this claim is true.

I believe that there is one response that satisfies both of these objections to the plausibility of Claim 2. Views of shared intention must understand shared intentional action either:

or

Here “analogous to” simply means that the two forms of intentional action share a family resemblance in all of the relevant features that render them intentional actions. Naturally, this allows for the possibility that there are important differences between individual and shared cases, given that they differ in at least one important respect—whether they are possessed by individuals or by groups of individuals.

So which of (a) or (b) is true? If it is the former, we have the incoherent implication that the two types of intentional action are intentional, yet on the basis of a significantly different set of conditions. It seems implausible to grant that actions are intentional but on the basis of a different set of necessary and sufficient conditions. The conditions that render individual actions intentional and shared actions intentional should be the same given that they are conditions for the same feature of each action. Namely, the feature that they are intentional, which has no bearing on whether they are individually or jointly held. Furthermore, there is no conceptually compelling reason to think that instances of shared intention must involve genuinely rational considerations simply because they involve more than one agent. Given that the Claim (a) seems implausible, then we have reason to think that they are indeed analogous and that they should bear a strong family resemblance to one another. Thus, we ought accept (b) as true, and, thereby, Claim 2.

Here one might further object that there is a third option besides (a) and (b) above to consider. Perhaps (b) is true, but both individual and shared intentional actions require justifying reasons as one of their respective necessary conditions. Yet if this claim were true, we would be forced to revise our conception of individual intentional action by accepting the overly demanding conception that it requires as its antecedent genuinely justifying considerations. Thus, we continue to have reasons to grant that (b) is true, and, thereby, to grant that there are independently plausible reasons for granting that Claim 2 is also true.

It appears as though Claims 1 and 2 are independently plausible. But if we are to take these two assumptions seriously, then it should follow that

Any account of shared intention that denies Claim 3 but accepts Claims 1 and 2 over-rationalizes, as I will call it, the phenomenon of shared intention. This means that the account takes one necessary condition for sharing an intention to be that the agents in question have genuinely good reasons for acting as they do. This diverges from Assumptions 1 and 2, both of which seem plausible on face value. It stands to reason that there is a problem with any theory of shared intention that diverges from them in denying Claim 3.

1.2. How We Get from the Guiding Assumptions to Over-Rationalization

But in what way do accounts of shared intention deny Claim 3 while accepting Claims 1 and 2? I have already shown why it is reasonable to accept (1) and (2) and that most accounts of shared intention take Claim 2 to set limiting conditions on the coherence of an account of shared intention. Let us now turn to see how they deny Claim 3.

Typically, accounts of shared intention argue that for intentions to be shared, each participating agent must understand each other agent as intending to φ and as understanding each other agent as similarly intending to φ. This, in various forms, is the common knowledge condition. ²³ This condition, I will argue, justifies the formation of intentions that are capable of being shared, and, as result, it can only serve and it must serve as a justifying consideration in forming these types of intentions. In what follows, I demonstrate that if we ask about the reasons that agents have for forming intentions that they will share, the common knowledge condition comes up as one such reason for doing so. As a result these accounts unwittingly end up arguing that justification is necessary for shared intention. In this regard, these views make the false claim that shared intention is possible only if the participating agents form their respective intentions on the grounds of genuinely rational considerations. The more serious issue is the implicit claim that instances of shared intention must be instances of genuinely rational intentional action, while individual intentional actions need not be.

Let us consider Bratman’s account of the role of common knowledge. As common knowledge plays roughly the same role in both types of views of shared intention described earlier, we can, for the sake of simplicity, take one view as the exemplar. On Bratman’s view, an intention can be shared if, in part, the participating agents take as the content of their respective intentions that they (together) act jointly. That they each take it as common knowledge that they each will so intend is an additional necessary condition for the possibility of shared intention. ²⁴ Here Bratman (1999, 121) understands the common knowledge condition in roughly the terms that David Lewis ([1969] 2002, 56) defends: a state of affairs can be that about which agents have common knowledge when it is reasonable to believe that such a state of affairs is the case. It is also reasonable, by extension, to believe that others have reason to believe such a state of affairs holds and so on ad infinitum (Schiffer 1972, 32). Although Lewis’ view is more demanding insofar as it is aimed at explaining agents’ game theoretic preference to conform to the convention that common knowledge implies, Bratman’s own use of this condition approximates the idea that Lewis sought to defend—namely, that there are states of affairs in the world, social or otherwise, that are open to others as objects of reasonable belief and such belief is, itself, open to others in the same way (Lewis [1969] 2002, 58, 74-75). The same can be said of Gilbert’s (1989, 189) use of a common knowledge condition, although her version is surely more demanding than Bratman’s use of this condition. ²⁵ For my purposes, however, the basic features of the common knowledge condition remain the same in both of these views of shared intention, so any problem that affects this condition will affect both views.

To illustrate the role that this condition plays, consider the following example. Jane and Bob agree to help their friend Karen move her furniture into her new apartment. Karen owns a particularly heavy armoire that cannot be moved without the help of at least two additional people. Jane, Bob, and Karen share an intention to move the armoire into the apartment just in case they each take it as their respective intentions that they act together to lift the armoire, and they form these intentions against the common knowledge background condition that each other agent will do the same. ²⁶

So what role does common knowledge play on these accounts? On face value, it is a necessary condition for shared intention. It explains how it is possible for individual agents to make judgments about shared matters of concern that require knowledge of what other participating individuals will do. But, from the looks of it, it also does more work than simply explaining how it might be possible for individuals to rely justifiably on what others will do. ²⁷

To understand the additional roles that common knowledge plays, let us focus on how an individual participant—say, Karen—comes to decide to participate in such an action. The following features describe Karen’s deliberative process such that she decides to intend to act with the others to move the armoire:

The above represents what must minimally be the case for an individual agent, in a situation of possible shared intention, to decide that she will form the intention to act together with the other agents in question. Common knowledge, in (4), leads Karen to form the intention to act jointly, although it is surely not what leads her to (want to) form the intention to move the armoire. In this context, common knowledge of Jane’s and Bob’s respective intentions provides overriding considerations in favor of forming the intention to act jointly.

How so? Karen, Jane, and Bob either will or will not act jointly. A necessary condition for their acting jointly is that they all so intend to act jointly. It is necessary but not sufficient insofar as other things must hold true—such as Jane not being impeded from realizing her intention to act to move the armoire by, say, a sudden wave of nausea—for the intention to be successfully realized. But Karen can reason that, given that these other conditions obtain, the fact that there is common knowledge between them provides an overriding reason to believe that one can realize the intention to act jointly. This, then, provides an overriding consideration to form the intention in question. Karen ought not form this intention if it were not a matter of common knowledge among them that they will each do their respective parts. She would be wrong insofar as she could not realize the intention to act jointly if others would not also be so committed. If she cannot realize the intention, then she ought not, rationally speaking, form the intention in the first place.

This entails that, in instances of shared intention, agents’ respective intentions to act jointly are necessarily formed on the basis of genuinely justified considerations. In this regard, these accounts over-rationalize instances of shared intention. Any account of intentional action ²⁸ that requires justifying reasons as necessary conditions for the action to be intentional over-rationalizes the type of action it seeks to explain.

Here one might object that although most accounts of shared intention require the common knowledge condition in some form, none requires that common knowledge provide an explicit reason for participating agents to intend to act together ²⁹ ; that is, common knowledge is typically taken to be a constituent condition for shared intention rather than a reason that individuals form intentions that will be shared. I think that this is a fair worry, particularly given that common knowledge is typically used to explain how agents can coordinate their actions rather than why they undertake them in the first place.

I do not deny that common knowledge is a constituent condition for shared intention. Or, at least, nothing about my account challenges this claim or rules it out. In fact, what my argument demonstrates is that by virtue of the fact that common knowledge is postulated as a constituent condition for shared intention, it comes to serve a second, implicit role, namely, it plays the secondary role of providing the justifying reason for agents to form intentions to be shared.

Although common knowledge plays one explicit role in accounts of shared intention, it plays another role that becomes explicit when we ask about why the participating agents formed the intentions to act together. It is a necessary condition for agents to share an intention that they intend to act together. If they must so intend, then they must individually form intentions to this effect. In explaining how and on what grounds individuals form intentions to this effect, we run headlong into the problem of the reasons that individuals can plausibly have to form these intentions. If accounts of shared intention are to maintain the family resemblance with accounts of individual intention, they must allow for the possibility that individuals can form intentions on the basis of either motivating or justifying reasons. But the only plausible reason that one has to form the intention to act with others, assuming that common knowledge obtains, is that common knowledge does obtain and thus that the intention to act with others is realizable.

Why is this the only plausible reason, for surely Karen can form the intention to act with others on the grounds that it, say, promotes camaraderie? Of course Karen can form the intention to act in light of this reason and this reason may even be based on a false belief, thereby saving the family resemblance between individual intentional action and shared intentional action. My point, however, is that at least one of the reasons must be that others so intend and that they understand that Karen so intends (i.e., the common knowledge condition). As a result, the accounts of shared intention fall prey to over-rationalization even admitting that agents may have additional reasons beside the common knowledge condition to form the intention to act together.

Here another related objection crops up. One might object that although the common knowledge condition may be a genuinely good reason for agents to form intentions to act together, it need not function as such. Why not? It is not necessarily the case that the common knowledge condition must be a genuinely good reason for agents to form intentions to act together, given that agents seem to be able to form these intentions when they are wrong about whether the common knowledge condition does indeed obtain. Might not Karen be able to form the intention to act with others on the basis of a false belief (which she takes to be true) about whether common knowledge obtains? If this is possible, then it is not the case that accounts of shared intention necessarily fall prey to over-rationalization by virtue of the fact that they take the common knowledge condition to be one necessary condition for shared intention.

Although this is an important objection, I believe that there is a natural response to it. For this objection to work, it would not be enough to claim that, from the perspective of Karen’s deliberative process, she need only believe that the common knowledge condition obtains for it to provide (from her perspective) an overriding consideration to form the intention to act together. For if this were the case, then the instance under consideration would no longer be an instance of shared intention, given that no one else intends to act together. And, if so, then the purported counterexample is no counterexample at all. In this regard, to share an intention, Karen, Jane, and Bob must each hold a belief about common knowledge that is, in fact, true. ³⁰ Thus, anytime there is a genuine case of shared intention, each participating agent has a true belief about the common knowledge condition. It is the fact that this condition obtains that serves as one of the reasons each agent has for forming the intention to act together. Thus, we run headlong into the problem of over-rationalization. Citing the possibility that agents may hold mistaken beliefs about common knowledge and yet still form the intentions to act together does not save these accounts from this problem, for the cases in question simply would not be instances of shared intention in the first place.

Here one might wonder about the implications of this claim for cases where individuals act on the basis of a common illusion rather than on the basis of genuine common knowledge. Consider the following case: Jacques, Juan, and Joe are weaving a rug together on a large, old-fashioned loom. Each believes that the others will intend to use the red wool to create the dominant parts of the pattern they are weaving. As it turns out, all three are color-blind and thus pull the green wool through the loom when they believe that they (and each other person) are pulling the red wool through the loom. The agents act on the basis of a common illusion about what others will intend rather than genuine common knowledge. Is this a case of genuine shared intention? If the answer to this question is “yes,” one might contend that this is evidence that my claim about over-rationalization is false—namely, agents can share an intention and do so on the basis of bad, but explanatory, reasons. I think that there are two possible responses here. First, one can argue that this is not an instance of shared intention precisely because the common knowledge condition does not obtain. One might also grant that it is an instance of accidental coordination. Alternatively, one can argue that it is an instance of shared intention, given that common illusion can stand in for common knowledge. If one chooses this route, then one can grant that shared intention, without the problem of over-rationalization, is possible. At this point, my suggestion is that the former rather than the latter response is preferable, given that my argument targets views of shared intention for which common knowledge is one necessary condition. Yet nothing about this argument entails that we must be committed to these types of views as the only views of shared intention; in fact, the argument I advance here hints that there may be reason to abandon the common knowledge condition as a necessary condition for shared intention.

4.1. The Inapplicability of the Charge of Over-Rationalization

First, one may claim that the charge of over-rationalization is inapplicable to the accounts under consideration. One might contend that the rationalizing attitudes in question are beliefs about what others will do, not attitudes that serve as considerations in favor of undertaking the action in question. ³⁷ The beliefs (plus the relevant desires) either do or do not warrant the formation of the intention. If they do, it is as a side effect of the specific content that they have. It is, one might claim, their content that makes them required for shared intention and not their (incidental) justification for the formation of the intention that we act together.

I think that there is a straightforward response to these concerns. It is true that citing common knowledge as a necessary condition for shared intentions does not entail that it is the reason that participating agents must have for forming the intentions in question. But my argument does not suggest that because common knowledge is a necessary condition for shared intention, it is therefore a reason that agents must have to form the intentions in question.

Rather, my argument begins from the assumption that common knowledge is a necessary condition for shared intention. With this condition in place, how then are we to answer the question of the reasons that participating individuals have for forming the intentions in question? I argue that the participating individuals cannot form the intention to act with others without considering the fact that others will also so act. They cannot do so because one of the realization conditions on their intention formation is that others also so intend. Thus, although the common knowledge condition may not be individuals’ only reasons for intending to act with others, it must be one of their respective reasons by virtue of its role as a realization condition that figures in agents’ formation of the intention. ³⁸

We typically think that agents can only form intentions in light of whether the intention is realizable within conditions over which they have some control. It thus follows that the common knowledge condition is one of agents’ reasons for forming the shared intention whether or not they identify it as such.

4.2. Might the Over-Rationalizing Condition Be a Structural Condition on Shareable Intentions?

Another possible objection to my argument suggests that the purpose of the putative over-rationalizing criterion is to explain how intentions can be shared rather than why the actions in which they result are intentional. These views do not, one may reasonably claim, entail that the intentions must be rationally justified. They merely suggest that for these intentions to be shared, they must be the product of justifying, rational considerations.

Does this objection undercut my argument? There are a number of reasons to think not.

First, it makes the unjustified assumption that the conditions under which an intention is shared have no bearing on the conditions under which it is formed. In this sense, it suggests that each individual forms her relevant intention, after which each decides whether to share it with others. But, of course, this is not the view. Individuals’ respective intentions are shared as a result, in part, of the way that each individual forms them.

Second, genuinely justifying considerations are not generally thought to be required for the formation of intentions, individual or shared. If they were, many actions that are plainly intentional—albeit irrational or the product of false beliefs—would not count as such. This is clearly an overly demanding conception of intentional action, and thus, it is unlikely to be a structural condition on the formation of intentions that are able to be shared.

Finally, it is unclear why the type of intention that one forms is somehow changed by the kinds of reasons one has when one forms it. The person who defends the objection under consideration here, however, makes just this claim, namely, over-rationalization is not a conceptually incoherent feature of shared intention because only a certain type of intention can be shared. The type of intention, this objector argues, is one that is formed only in light of genuinely good reasons. Yet making this claim in response to the over-rationalization charge requires that we grant that intentions are different in type when they are formed in light of justifying reasons rather than explanatory, but bad, reasons. I think that we have reasons to reject this claim. When I intend to drink coffee because I believe it has magical powers and when I intend to drink coffee because I believe that I prefer it to tea, my intentions do not appear to be different in any significant, structural way. More important, it is not clear why intentions formed on the basis of justifying reasons are metaphysically or conceptually more able to be shared than those that are formed on the basis of explanatory, but bad, reasons.

4.3. Reasons for the Collective versus Reasons for the Members ³⁹

The purpose of the argument that I have advanced is to show that accounts of shared intention fall prey to what I have called over-rationalization. One might object here that I have failed to distinguish between those considerations that are reasons for the individual participating agents and those considerations that are reasons for the individuals qua collective or group.

There are three ways to understand what the proponent of this objection might mean. One way to interpret it is that my argument commits the fallacy of composition. Here the idea is that though individual agents may form intentions to act with others on the basis of genuinely justifying reasons, this does not entail that the individuals qua group have such reasons when they act together. To draw this inference is to commit the fallacy of composition.

In response to this interpretation of the objection, I suggest that my argument only shows that the individuals who form intentions to act together must form those intentions in light of genuinely justifying reasons. Naturally, this does not entail that the individuals, when taken as a group, have the same set of reasons for acting together.

It does entail, however, that the following conditional is true: if individuals must form intentions to act together to share an intention, their reasons for forming their respective individual intentions to act together are necessary conditions for the individuals, as a group, to act together. This suggests that my argument does not commit the fallacy of composition, but it does show that whatever is necessary for individuals to form intentions to act together will also be a necessary condition for individuals, as a group, to share an intention.

A second interpretation of the objection above takes the proponent of such an objection to claim that if over-rationalization affects individuals who form shareable intentions, it ought also thereby to affect group’s reasons. ⁴⁰ Given that the question of whether we can attribute intentional states and, in particular, reasons to group agents is a question that cannot be fully addressed here, I will provide only a tentative answer to this objection

Let us assume that we can attribute intentions and reasons, among other attitudinal states, to groups and that these attitudinal states are not simply the attitudinal states of the individuals who comprise the group. Even with this assumption in place, it remains an open question whether groups act for reasons in the traditional sense of the term. If not, then the over-rationalization problem ought not to affect group agency.

Following List and Pettit (2011, 58), I take it that the only appropriate domains in which to properly speak of group reasons are those domains that are structured by specific institutional rules, such as corporations, appellate courts, and the like. ⁴¹ If this claim is correct, then it follows that the problem of over-rationalization would affect group agency. But it would not do so in a way that is unacceptable. Why not? Any reasons that group agents would possess would be generated by a decision-making procedure ensured to generate good reasons or, at least, reasons that are good by virtue of being representative of the individual attitudes of the members who comprise the group (List and Pettit 2011, 64).

This form of over-rationalization is relatively harmless given that group agents who are defined by institutional decision-making procedures are very much unlike individual agents. Given this fact, we have no reason to attribute the assumptions to this view of group agency that I attributed to most views of shared agency in Section 1 and that resulted in the worrisome form of over-rationalization.

If we do not grant List and Pettit’s claim regarding the domain in which group agency is possible, then we get different results regarding the question of whether the problem of over-rationalization affects group agency. Here again, as above, the answer is that group agency would indeed be affected by the problem of over-rationalization. When we do not grant List and Pettit’s claim, we have no reason to suppose that group agents are radically unlike individual agents. As a result, as over-rationalization affects instances of shared agency where the individuals in question do not comprise a group agent, it should also affect instances where they do comprise a group agent.

There is a third possible reading of the objection under consideration here, that is, might it be the case that group agents (if there are such entities) can possess both explanatory and justificatory reasons when they act even if the individual members of the group can only have justifying reasons for forming their intentions to act as group? Returning to the case of Karen and the armoire above, a proponent of this objection would make the following claim. It may very well be the case that Karen, Jane, and Bob each has exclusively justifying reasons to form their intentions to act together. Yet what matters is that when they act as a group agent who intends to move the armoire, they may have merely explanatory reasons for doing so. If this is right, then my objection that there is a problem of over-rationalization would not apply in the case of the group agent even though it applies for those individuals (Karen, Jane, and Bob) who comprise the group agent.

Given that the question of whether group agents are metaphysically and ontologically distinct from the individuals who comprise them is far from resolved, I can offer only a tentative response to this third and final interpretation of the objection at hand. As I describe in my introduction, the common knowledge condition is widely regarded as a necessary condition for group and/or individual agency and thereby for group or individually shared intentional action. If this is right, then, by transitivity, any common knowledge conditions that obtain for individuals qua individuals also obtain for individuals qua groups. It would then follow that if common knowledge must be a reason for individuals to form an intention to act together as a group, it must also follow that, when they comprise the group, these reasons also hold for the group.

Of course, one may argue that the group can have reasons for acting that are entirely distinct from and need not overlap even minimally with those reasons that the individuals have,. If one takes this approach, my argument about over-rationalization is orthogonal to the question of over-rationalization in the case of group agents. My sense, however, is that the argumentative burden in this case would be significant and difficult to discharge, given that one would have to show that the group agent has no genuine metaphysical or ontological ties to the individuals who comprise it. If one could make such a claim stick, then instances of group agency would be phenomena outside of the bounds of the traditional phenomena that accounts of both shared and group agency typically seek to explain. Still, if such a group agent exists, then I am willing to grant that there may be some instances where over-rationalization is not a problem. This claim is cold comfort, however, to the proponents of the standard views of shared and group agency, given that my charge of over-rationalization would still apply to their accounts.