In the previous section, I suggested that each kind of responsibility cannot be attributed as often as we might like. I now want to scrutinise that thought: just how prevalent are these responsibility gaps? This will require considering each collective-type separately, because some gaps are prevalent for some collective-types but not others.
Above, I explained that it is highly infeasible to assign causal responsibility for the GPGP to any individual, business, or government agency. This seemed to imply that the GPGP has a causal responsibility gap. But there is another possibility: could a diffuse collective, as a whole, have causal responsibility for the GPGP? Consider the collective composed of all plastics-using agents (individuals and agential collectives). These agents are not united by acting responsively to one another as they work towards a common goal (indeed, their goals are in opposition, insofar as they are competitors). Nor are they united under a collective-level decision-making procedure. This collective is diffuse. Can we attribute causal responsibility to it? It is tempting to think not. After all, there is no united entity here for responsibility to be attributed to.
The metaphysics of causation can help here. I shall use Christian List and Peter Menzie’s difference-making account (a version of the probability account) to show that causal responsibility can be attributed to diffuse collectives, plugging causal responsibility gaps.
Consider the following. What matters for the existence of the GPGP is not that Coca-Cola, Pepsi, Greenpeace, a fishing trawler, a local government, and a consumer—alongside innumerable other agents—each performed certain highly specific actions. As explained in the previous section, the GPGP would still have existed if those actions had been different. Those actions are specific instantiations of a more generally describable phenomenon: plastics use by plastics-users. If we insist that a causal explanation of the GPGP cannot refer to this diffuse collective, but must instead list all the specific actions of all the specific entities included within that diffuse collective, then we would get a causal explanation that ‘does not satisfactorily fit the role [of ‘cause’] because it is overly specific and involves extraneous detail.’ (List and Menzies 2009, pp. 479–480). To illustrate, List and Menzies give an example from Stephen Yablo:
consider a pigeon that has been trained to peck at all and only red objects. The pigeon is presented with a red target and she pecks at it. As it happens, the target is a specific shade of crimson. What caused the pigeon to peck? Was it the fact that the target was red or the fact that it was crimson? … The target’s being red is of the right degree of specificity to count as a cause of the pigeon’s action. In contrast, the target’s being crimson is too specific to count as a cause of the pigeon’s pecking: citing it as a cause might give the erroneous impression that the pigeon would not peck at anything non-crimson (List and Menzies 2009, p. 480)
The point is this: the target’s being crimson isn’t what makes the pigeon peck. The pigeon would have pecked if the target were maroon, or scarlet, and so on. What makes a difference to the pecking is that the pigeon’s target is some form of red or other.
Likewise for the GPGP. Coca-Cola’s designing bottles isn’t what makes a difference to the GPGP existing. Instead, what makes a difference is that some-collection-or-other of entities, taken as a whole, in aggregate, produce large amounts of (what ends up being) plastic trash. The diffuse collective ‘plastics-using agents, whoever they happen to be’ is the grouping that has the right level of detail to explain how the GPGP occurs. The causal power of this diffuse collective is ‘realisation insensitive’ (List and Menzies 2009, p. 496; List and Spiekermann 2013). That is, regardless of which particular agents happen to be part of this large plastics-using diffuse collective, the GPGP will result; and if the large plastics-using diffuse collective did not exist, then the GPGP will not result. It is this collective that makes the difference.
If we follow List and Menzies in this way, then we can say the diffuse collective ‘plastics-users’ causes the GPGP, while none of its members does. More generally, we can say that diffuse collectives can have causal responsibility: polluters as a group cause pollution; bargain apparel companies as a group cause the continuation of sweatshop conditions; and so on. Given that diffuse collectives are everywhere and made up of anything, this result has the power to eradicate causal responsibility gaps, assuming we can pinpoint the right diffuse collective in each case. This result also allows us to judge that all statements in the left-hand column of Table 1 are defensible (at least, pending empirical demonstration that the collectives named in that column are indeed the collectives with which the relevant effect does occur, and without which it does not occur). By salvaging our ability to provide collective-level historical causal assertions about all three kinds of collective, List and Menzies’ account salvages our ability to satisfy victims who are concerned to have historical knowledge of how their situation came about.
So far, so good. But we cannot extend this vindication of diffuse collectives’ causal responsibility to their prospective and moral responsibility. The main problem can be explained briefly. It is that diffuse collectives lack the volitions necessary for prospective and moral responsibility. To explain, consider that there are numerous outcomes that are possible: sweatshop labour could end; global warming could be limited to 2 degrees above 1990 levels; the GPGP could be cleaned up. But for there to be moral or prospective responsibility over those outcomes (or their absence), the outcomes (or absence) must be tied in some way to volitions (decisions, intentions, willings, tryings). For there to be moral responsibility for an outcome, there must have been a past volition that did (in some way) cause the outcome. For there to be prospective responsibility for an outcome, there must be the possibility of a future volition, produced via moral reasoning, which could (in some way) cause the outcome.
Unlike agential collectives, diffuse collectives cannot have their own, group-level volitions, decision, intentions, willings, or tryings (Lawford-Smith 2015, p. 235)—either in the past or in the future. Of course, we could say that many or most members of a diffuse collective have individual volitions over the same thing. For example, we could imagine that many or most plastics users acquire an individual volition to eradicate the GPGP. But that would be a statistical fact about the individuals; this does not take the group as a united whole. And if all members of the collective were to have the volition, then that would render the collective teleological. (It is innocuous to assume that, when individuals have a goal in common, they will respond to one another, insofar they encounter one another, in their individual pursuits of that goal.) With this in mind, it is to teleological collectives that I will now turn.
Recall that members of teleological collectives (1) act responsively to one another (insofar they encounter one another) as they work toward a common goal, but (2) lack clear procedures for making decisions that are attributable to the collective as such. The arguments above about causal responsibility in diffuse groups extend straightforwardly to teleological collectives: if diffuse collectives can be causally responsible, then so can teleological collectives. The unity present in teleological collectives (as compared with diffuse collectives) makes it more likely that, in a teleological collective, the collective level-of-description will be the difference-maker for some effect. This includes such collectives as ‘the plastics lobby,’ ‘democracy-endorsing states,’ and ‘environmental non-governmental organisations.’
Again like diffuse collectives, teleological collectives are poor candidates for prospective responsibilities. While the members of teleological collectives do have volitions with the same content, these groups lack a group-level decision-making procedure that could process obligations and itself produce a volition, via moral reasoning, to perform the action that an obligation requires of the group per se (indeed, the group per se cannot act at all, in virtue of not having such a decision-making procedure). That is, teleological collectives do not have a central core of agency at which it is possible to reason its way to a genuinely group-level volition, or decision, or action. To see this, it is helpful to compare them with agential collectives. An agential collective can decide something (have a volition over it), even if no member decides that thing—indeed, even if no member would decide that thing. This happens in cases of intra-group compromise. Suppose the Coca-Cola board decides to say something about the GPGP. Half want to say it is regrettable, but not deplorable; the other half want to say it is deplorable. They all agree ‘awful’ is halfway between ‘regrettable’ and ‘deplorable.’ So, as a compromise, the board decides ‘Coca-Cola will say the GPGP is awful.’ This group-level volition has arisen out of the group’s decision-making procedure: no individual member has decided this. What’s more, if any given individual members were to decide for the board, using just her own decision-making procedures, then she would have decided differently (choosing either ‘regrettable’ or ‘deplorable,’ not ‘awful’) and would have used a different procedure to decide (using her own deliberative faculties, rather than a compromise procedure). (If she would have taken others’ preferences into account, then this is more like her operationalising the distinctively group-level procedure, than it is like her making her own decision.)
Such uniquely group-level procedures (e.g. compromise) for producing uniquely group-level volitions (e.g. ‘awful’) are possible in agential collectives, but these features cannot arise if members merely act responsively toward a common goal. To illustrate, suppose Algalita and NOAA each adopt the goal of cleaning up the GPGP, and each start to do so responsively (e.g. NOAA starts at the East because Algalita has already started at the West; Algalita then heads South when it notices NOAA heading North). There is not yet any mechanism in place for producing an Algalita-NOAA distinctively group-level decision on who cleans which areas. This is because, unlike in an agential collective, any alleged group-level Algalita-NOAA decision: (1) cannot conflict with the individual companies’ private decisions; (2) cannot be other than a stringing together (with ‘and’) of the two companies’ private decisions (‘NOAA will clean the East and North and Algalita will clean the West and South’); (3) is compartmentalised: each company has influence only over their own ‘part’ of the decision. Contrast this with our imagined Coca-Cola board decision, where the group decision (1) conflicted with all members’ private decisions, (2) was not made of parts conjoined by ‘and,’ so (3) did not have parts that were divided amongst members.
One might think decision-making is not the only agential capacity: perhaps teleological collectives have some other agential capacities. But decision-making—and, more specifically, the ability to make decisions via moral reasoning—is necessary to bear prospective responsibility, which is ultimately at issue here. And the above three points apply just as much to beliefs, desires, and intentions as to decisions. Thus, teleological collectives lack many agential capacities that are crucial to prospective responsibility.
Of course, teleological collectives are not entirely like diffuse ones, for at least three reasons. First, many teleological collectives are such that they would be transformed into agential collectives if all the members aimed at this. Isaacs (2011) uses ‘putative collective’ to mark collectives with this potentiality. Until this has happened, though, there is no distinct source of distinct decisions, at which we can target duty attributions.
Second, when a member of a teleological collective fulfils a prospective responsibility to bring about some good, they will often have to act responsively to other members of the teleological collective, as Algalita and NOAA do in the imagined example. This inter-relatedness of the content of members’ duties might be thought to suggest that the duties are held by the collective as such (Young 2011, pp. 180–181). However, we should not infer this. After all, all prospective responsibilities of all agents require their bearers to react to other agents. This might be a more pressing consideration within a teleological collective than within a diffuse one, but this is a matter of degree, not of kind. What matters for the existence of a group-level duty is not interrelated content of individuals’ duties, but the possibility of group-level reasoning when deciding whether to produce a volition to act on the duty at all. Teleological collectives are, then, like diffuse collectives in that they can have causal responsibility but not prospective responsibility.
Third, teleological group members are disposed to rely upon and reinforce one another’s endorsement of, and actions towards, the shared goal. It is also rational (at least, more rational than in diffuse collectives) for members of teleological groups to engage in ‘we-reasoning’ around their common goal. When someone we-reasons, he ‘considers which combination of actions by members of the team would best promote the team’s objective, and then performs his part of that combination.’ (Sugden 2003, p. 167, emphasis original). We can thus view teleological groups thus hold a weak type of shared agency, though this doesn’t rise to the capacity for collective-level reasoning. In the imagined example, Algalita’s cleaning action encourages NOAA’s specific cleaning action, which encourages Algalita’s subsequent action, and so on. Each chooses their individual action with the common goal in mind (‘cleaning the GPGP’) and does their bit of the collective pattern that will produce that result.
In light of this, what should we say about moral responsibility in teleological collectives? Neither option is costless. First, we could decide that the weak type of shared agency is enough for blame. (Using different conceptions of weak shared agency, Sheehy (2007) and Pasternak (2011) each argues along these lines.) However, if we say this, then we must take one of two further options: (1) Impute prospective responsibility to teleological collectives. This is something I have argued firmly against (though this is the option Sheehy and Pasternak each endorse). (2) Deny an attractive general claim about the connection between moral and prospective responsibility. This attractive general claim concerns symmetry: ‘If there is symmetry, then whenever there is [moral] responsibility for an action after the fact, there will also have been an obligation to not have performed that action before the fact.’ (Lawford-Smith 2015, p. 241). In other words: to impute blameworthiness to teleological groups, while endorsing my argument that they lack obligations, requires saying that they can be blameworthy for some outcome without having had an obligation not to produce that outcome.
So much for the first option: imputing blame to teleological collectives. The second option is to point out that the common goals, mutual reliance, mutual reinforcement, and common rational availability of we-reasoning do not make for genuinely group-level volitions. The group cannot have a volition that differs from, and is produced in a way different from, the sum of members’ decisions. This is important, because a wrongful volition seems necessary for blame to be justified. What’s more, the symmetry claim is highly attractive and (I have argued) teleological collectives lack obligations. Thus, we might deny group-level blame for teleological collectives. This will seem unsatisfactory when some teleological group (such ‘the plastics lobby’) seems to be particularly connected to a harm. But notice all that we can say in such cases: we can say that each member of the plastics lobby is blameworthy for having the goal of promoting plastics production and consumption, and for encouraging others to have that goal. To explain exactly why this is wrong for each agent, we can appeal to the wrongness of being part of the group that is causally responsible for a bad outcome. This doesn’t mean that we must blame each lobbyist for the entire GPGP [as Kutz (2000, p. 122) would be inclined to do]. Instead, each agent (whether individual or collective) has performed the wrongful action of making itself part of a harm-producing group.
In my view, denying teleological collectives’ blame is the best option. At least, this is true on the assumption that irreducible group-level volition is necessary for group-level blame to be justifiable. It will also be the best option if (1) we should retain symmetry between moral and prospective responsibility and (2) I am right that teleological collectives lack obligations.Footnote 8
So far, I have discussed attributing responsibility to diffuse and teleological collectives. These are the trickiest cases for collective responsibility attributions. When we turn to agential collectives, we can see that they can straightforwardly be attributed all three kinds of responsibility. Indeed, this might seem to follow from arguments for the irreducible moral agency of agential collectives (e.g. given by Copp 2007; Pettit 2007; Rovane 1998; Tollefsen 2015). But it is worth spelling out exactly why. This will allow us to see that there are still some gaps, particularly when it comes to prospective responsibility.
If we apply List and Menzies’ argument to diffuse and teleological collectives, then we can also apply it to agential collectives. However, this means that assigning causal responsibility to an agential collective is defensible only when the collective-level phenomenon is what makes a difference to the presence or absence of some effect. For example, imagine that BP’s role in the Deep Horizon spill was in fact dependent on the actions of just one employee, acting beyond the scope of their role. In that case, the collective-level is not the difference-making level. By contrast, if BP’s oversight procedures (for example) more-or-less ensured that some disaster like this would happen at some point, then the collective-level is the relevant causal level. The same is true for diffuse and teleological collectives. But this limitation is particularly worth noting for agential collectives, since we might too readily attribute causal power to the agential collective as such.
For moral responsibility, there are other necessary conditions to add. As alluded to above, moral responsibility requires a certain quality of will—a volition, intending, willing, or trying—alongside harm-causing actions. Also alluded to above, agential collectives can have qualities of will even if some, most, or even all members lack that quality.. I mentioned compromises earlier, but discursive dilemmas are the most well-known cases of this kind. Imagine a three-person group that makes decisions via majoritarian voting. It votes on the motions P, Q, and R in turn. Motion P is that humans are not sentient, Q is that humans are nutritious, and R is that humans are a sustainable food source. Assume these are each necessary and together sufficient for it to be permissible to kill and eat humans. A votes for P, for Q, and again R; B votes for P, against Q, and for R; and C votes against P, for Q, and against R. So, by voting on each of P, Q, and R in turn, the group comes to believe in all three motions.Footnote 9 The group, thereby, comes to believe that it is permissible to kill and eat humans. Being rational, it forms the volitions entailed by this (such as being unfazed if it starts to do so, and intending not to interfere if others do so). In this kind of case, no individual member is morally responsible for the agential collective’s coming to this belief—after all, they each individually deny that that all three motions are true—so we must place blameworthiness the group itself (Pettit 2007, pp. 197–198).
However, if we are to blame the collective itself, the collective itself must be a moral agent. This requires that the agential collective can respond appropriately to moral reasons. If a group’s members are human moral agents, and can recognise moral reasons, then it is natural to assume they can design a group decision-making procedure so that moral reasons are treated as desires, in the sense of having a ‘world-to-mind’ direction of fit. And it is natural to assume they can design the procedure so that the group, by and large, responds to moral reasons, such that the group takes its own measures in response to these reasons (Pettit 2007, p. 187). Of course, it’s unlikely that the group will do this infallibly, but if infallibility were the standard for moral agency, then no human would be a moral agent. So, this is not a conceptual roadblock to agential collectives’ moral responsibility. It is, however, a practical roadblock: if some agential collective is constitutionally incapable of recognising moral reasons (if its procedure is set up so that members cannot present these reasons to the agential collective) then the agential collective cannot be held morally responsible.
If an agential collective is a moral agent—if it can use its decision-making procedure to respond appropriately to moral reasons—then it can have prospective responsibility. This might look like great news for attributing duties regarding such issues as cleaning up the GPGP. However, before we can assign prospective responsibilities to agential collectives, we must ask about ability gaps and justification gaps. Agential collectives seem powerful when compared with individual agents, so it is tempting to think that they can do anything and everything. Patently, this is not the case. Ability gaps abound for agential collectives. In short, this is because agential collectives’ actions must be enacted by their individual members. Many agential collectives (e.g. states and business corporations) have a fair amount of control over their members, but this control only goes so far.
Inter-governmental organisations (IGOs) are a clear example here. Very roughly, an IGO is a collective, created by agreement between states, of which the members are states or organs of states. State members are bound by the decision-making procedures of IGOs, so are obliged to follow the role instructions that they are given by the IGO’s decision-making procedure (unless the state has entered a “reservation” on that point, upon acceding to the treaty establishing the IGO). This includes abiding by the conventions that the IGO endorses on behalf of its members. However, given that IGOs are run by their members (states), IGOs can only distribute to states those roles that states have given the IGO decision-making procedure the power to distribute—and IGOs can only distribute those roles in the way that states permit through the formulation of the decision-making procedure. Yes, IGOs can and sometimes do provide ‘warrants,’ ‘approval,’ and ‘legitimacy’ to states’ actions [as Harbour (2004, p. 67) argues in defence of their status as moral agents]. But they are not the ones executing the actions and their instructions are not always taken as authoritative. As James Crawford notes, “[u]nlike states, international organizations do not possess general competence: they may only exercise those powers expressly or impliedly bestowed upon them” (2012, p. 184, emphasis added). IGOs are a particularly clear case. But to some extent or other, this is a problem for all agential collectives. They are at the mercy of their members or some subset of their members. They can be disbanded or have their procedures overridden if enough members (or the right ones) act to make it so. (Of course, members might have obligations to do what they can to ensure their collectives can, and do, discharge their obligations. I argue for these member obligations in Collins 2017a).
Even if an agential collective is able to perform some action, we might face a justification gap in giving an account of why it is obliged to perform that action. In business ethics, this has long been framed in terms of the debate between shareholder and stakeholder theory. In the terms of this article, that is a debate about when, and why, there are justification gaps in businesses’ responsibilities. According to shareholder theorists, there are justification gaps whenever an action is not in the interests of shareholders (though see Mansell 2013). According to stakeholder theorists, justification gaps are far less widespread, since a prospective responsibility to perform some action can be justified if it is in the interests of a wider set of stakeholders. I do not wish to settle this dispute here. My purpose in mentioning it is to point out that this debate occupies a tiny portion of what we might think about under the rubric of agential collectives’ responsibility, and to show that we can frame this dispute in terms of a dispute about the prevalence of responsibility gaps—more specifically, prospective responsibility gaps, and even more specifically, justification gaps.
Last but not least, consider fulfilment gaps. Almost no matter what one’s view of justification gaps (specifically, whether one is a shareholder or a stakeholder theorist), one will look out into the world and see fulfilment gaps: cases where an agential collective was able to bring about some good, and had a duty to do it, but failed to fulfil that duty. In such cases, we can place censure upon the agential collective itself. The question will remain of whether other agents (collective or individual) have duties to ‘pick up the slack.’ This is another debate that I cannot settle here, but rather, again, wish to place within my framework: the debate about whether agential collectives have duties to pick up the slack for one another’s failures is a debate in fulfilment gaps, which will, in turn, depend upon one’s position on justification gaps.
The conclusions of this section are summarised in Table 3.