The Epistemology of Group Duties: What We Know and What We Ought to do

Recently, my friend Jane suffered a bad fall on a joint trail ride. Alice and I, the two other riders in the group, stopped and jumped off our horses to assist Jane. Alice quickly took hold of Jane’s riderless horse and I handed her my reins, too, indicating that I would attend to Jane who was still on the ground and obviously in pain. After checking on Jane I called an ambulance and while I started making my way to the nearest road to meet the paramedics, Alice stayed behind, taking care of our three horses and keeping an eye on Jane. By the time the ambulance arrived and paramedics were loading Jane onto a stretcher, a cyclist had stopped and offered assistance. He then helped the two paramedics carry the stretcher up the embankment to the road where the ambulance was parked.

When Jane had her accident, Alice and I, in an instant, divided up the roles that we each needed to play in making sure that Jane was safe and being looked after. The paramedics, a well-rehearsed professional team, did their best to help Jane and, finally, the cyclist, a passer-by, joined in the effort and assisted, too. Being cooperative is natural to us and we have set up our social world in a way that both presupposes and requires constant joint efforts. We rely on each other all the time, assuming that others will play their part in these shared endeavours as we are playing ours.

Cooperation works really well in many if not most situations we find ourselves in – from basic coordination when sharing and navigating public spaces, to enjoying social activities with friends and loved ones, to working in a team with others in our professional capacities on complex tasks. Still, we regularly encounter collective action problems that challenge us: Whether these arise because we have arranged the social world in a suboptimal way or whether something unforeseen and socially unrehearsed occurs – we can get stuck with problems that we have not developed (and may not even be able to develop) behavioural patterns or effective joint responses for.

Philosophers have spent considerable energy on analyzing small-scale cases of impromptu collective assistance among random bystanders, for instance (Aas 2015; Collins 2013, 2019; Goodin 2012; Held 1970; Schwenkenbecher 2014, 2019). Conceptually even more challenging are those large-scale moral problems that we could substantially improve through distributive collective action (Schwenkenbecher 2020a) – changing our day-to-day activities and adopting certain collective behavioural patterns (such as reducing household waste and plastic usage to address the problem of plastic pollution of oceans and waterways). Some of these behavioural changes are going to be good for us regardless of whether or not collective success is achieved. Others – and that’s the great majority, really – are more complicated: individual efforts are in vain unless (sufficiently many) others cooperate. These changes will generate costs for the unilaterally acting agent while producing no benefit for anyone. Worse, uncoordinated action might even stymie collective efforts to produce such benefits.

In other words, cooperation comes with its very specific set of problems surrounding the uncertainty of others’ actions, intentions, and motives. This uncertainty may undermine agents’ reasons to choose cooperative (multilateral) over non-cooperative (unilateral) options, or their motivation to do so, or even make them fail to perceive of (or frame) a situation as one requiring cooperation at all. This is particularly true of cooperative action outside so-called structured or constituted group agents. Group agents such as a university or a company, but also a team of paramedics on duty, assign roles to its individual members; communication and supervision structures warrant that role-related responsibilities are discharged and that the group achieves its goals (if it is well-functioning). These structures in turn provide group members with the level of certainty concerning others’ actions, beliefs and goals that is necessary for adopting group-based reasons for action (Woodard 2017) and be appropriately motivated to make their own contribution to the joint endeavour.

This is different in groups that are not thus structured and where collaboration is ad hoc. These can include passers-by and random bystanders, but also social groups (who are merely united by a common involuntary feature or some common goal or interest). In small-scale scenarios with groups of manageable size, continued mutual reassurance is often key to the success of any joint endeavor (Bratman 2014). Successful communication can be stymied by a variety of factors and even where it works, uncertainties and disagreements concerning the joint goal, its relative (moral) importance, and the individual strategies that will produce that goal will often jeopardize the collective endeavor (Held 1970). In larger, unstructured groups the problem often lies in the lack of direct communication between group members and uncertainty regarding group membership.

Uncertainty about others’ plans and future actions has moral implications: not only does it make collective action^[1] difficult, but it also poses challenges for moral agency: When do we have moral obligations to act jointly with others? When are we responsible for our failure to have acted jointly with others? These are some of the key questions that have been preoccupying moral theorists who focus on collective action for several decades now.^[2]

In her book, Group Duties, Stephanie Collins aims to develop comprehensive answers to some of those questions by providing an account of the kind of moral duties groups and their members can have, covering all groups from so-called group agents and unstructured groups to groups that are mere mereological sums (Collins 2019). There is much to be commended about Collins’ book, but I will be focusing on the points of disagreement and areas that need clarification, with a view to constructively driving the debate forward.

In order to do that, let me return to the trail riding accident. This is a joint necessity^[3] case in that the contributions of both Alice and myself are required in order to effentively assist Jane. What are we – individually and collectively – required to do? In “Collective moral obligations: ‘we-reasoning’ and the perspective of the deliberating agent”, I suggest that:

“in scenarios where the contributions of two or more agents are jointly necessary in order to produce a morally desirable outcome, we sometimes reason from the top down, so to speak, starting with the most desirable option even though that option is not available to us in the way that individually achievable options are available to us” (Schwenkenbecher 2019, p. 161).

And further:

“agents should we-reason when the collective option is their best (moral) bet given the evidence. If both agents are equally in a position to view the same problem in the same way, and if to the conscientious agent pattern-based reasons outweigh individual-based reasons then both become the joint subject of a collective obligation, that is, they jointly hold that obligation” (2019, p. 163).

In other words, to the extent that Alice and I had reason to view the collectively available outcome (cooperating in assisting Jane as we did) as optimal compared to alternative, individually available outcomes, we would determine our individual contributions by way of inferring them from that optimal outcome. This involved taking plausible steps towards producing that outcome. On my view, Alice and I had a joint obligation to assist Jane and we had corresponding individual obligations to do our share in that joint endeavour.^[4]

Collins, in her book, opts for a view that is not altogether dissimilar, but which introduces an additional (and – as I will be suggesting – not altogether necessary) ontological element. According to Collins, what obligations we have in scenarios such as the one above comes down to the type of group we are in: depending on that we need to employ different approaches to our moral reasoning and our obligations differ, too. There are only three kinds of groups when it comes to collective action and moral obligations, according to Collins’ “tripartite model”. Those that she calls “collectives” are united by a decision procedure. When collectives have moral obligations they hold such obligations as a group and members must play whatever role the decision procedure allocates to them. Then there are “coalitions” of agents who are united by some common goal but have no decision procedure. Members of coalitions share a goal, each has a belief concerning the others’ adoption of that same goal and each is willing to act responsively to others. At the same time, there is no decision-making procedure, not even in the most minimal sense of someone being the “dictator” and simply giving instructions that others follow. Coalitions cannot have obligations (as a group) but individuals can acquire coordination duties that require them to enact the result of deliberation via “coalition-reasoning”, a modified form of “we-reasoning” (Hakli, Miller, and Tuomela 2010)). Finally, there are “combinations”, which are all the groups that fall short of being coalitions. Their members can have coordination duties to do whatever so-called I-mode reasoning tells them to do. Collins claims that these categories are exhaustive.

According to Collins, what kind of group were the horse riders and what kind of reasoning were we required to apply? And what kind of moral obligations did we have? In applying Collins’ categories, it would first seem that the three horse riders were a non-moral coalition: we shared a goal – to go on a trail ride together – but this was not a moral goal; this fact was out in the open – we knew we all wanted to go on a trail ride –; and we were each disposed to act responsively to others with regard to that goal. In that we resemble the two beach-goers, one of Collins’ examples, who share the goal of eating a tasty lunch together (2019: 147).

At a second glance, however, it appears that we may actually have been a “collective” – a category that also includes groups such as business corporations, states, and other kinds of organizations. Collins is fairly permissive with regard to what falls under “collectives”. What distinguishes collectives from mere coalitions, on Collins’ account, is whether members are united under a group-level decision-making procedure. That decision-making procedure can be informal:

“To illustrate the permissiveness, imagine a group of three friends who are at the beach. Numerous decisions must be made: where to lay their towels, where to go for lunch, and so on. Such a group is probably composed of agents that are united under a rationally operated group-level decision-making procedure that can attend to moral considerations. The procedure is probably conversation-based consensus. This procedure can become established simply by each member’s taking a conversational and consensual stance to the various decisions—each asking the others where, for example, they would prefer to go for lunch, and why, until all agree. Such a procedure can be rationally operated, just so long as the group doesn’t decide, for example, both to get burgers for lunch and not to get burgers for lunch. By taking part in the conversations (which can be done simply by staying silent to indicate indifference), each member tacitly commits to abide by the procedure’s results.” (Collins 2019, p. 14)

On this characterization, the horse riders were a collective, at least at the beginning of the ride. There was conversation-based consensus: we more or less jointly decided when to ride, where to ride, and at what pace. But did our group status change later, when we were forced to abandon our goal and Alice and I, instead of continuing on our ride, cooperated in helping Jane? Were the two of us also a collective, albeit different from the original one? On Collins’ account, the answer to this question would seem to depend on how Alice and I communicated. Because we made decisions about assisting Jane in the way described above, Alice and I were a collective, albeit one that soon merged into a larger collective comprising also the paramedics – a professional team and certainly a collective in Collins’ sense – and the cyclist passing by, before eventually being reduced to the smaller collective of just Alice and myself again. In other words, our collective changed multiple times over the two or so hours that passed.

To Collins, the big difference between “collectives” on the one hand and “coalitions” as well as “combinations” on the other hand is that only in the first type of group are members able to jointly deliberate, to communicate, to (jointly) form shared goals and devise strategies. This is what they need in order to make decisions, according to Collins, and only where an agent can make decisions can they hold moral obligations:

“collectives are able to make decisions to produce multilateralism (coordinated role-performance) amongst members. Combinations and coalitions lack this ability, because they cannot make decisions at all.” (Collins 2019, p. 154)

Coalitions’ and combinations’ members (need to) reason independently; they cannot make joint decisions. Note, that the lines between combinations, coalitions, and collectives are at times razor-thin: Two unconnected beachgoers who merely exchange a “concerned look” when they see someone drowning are therewith transformed from a mere combination (two random people at the beach) into a coalition, on Collins’ view. They now have a “team’s objective” in saving the drowning person and should infer their individual actions from the team’s objective (section 5.3). However, if among the same unconnected beachgoers one then instructs the other regarding what to do, therewith acting as the group’s leader, and the other one abides, then the beachgoers have transformed into a collective with an irreducible group obligation (section 6.2.1). Analogously, if Alice and I had gone about our respective individual actions in assisting Jane without communicating, we would have been “merely” a coalition. In that case, we would not incurred a group obligation to assist Jane, but merely individual-level obligations, which Collins calls “coordination duties”. The duty to assist Jane would have remained unallocated. This is remarkable, because in collective ethics, including in Collins’ earlier work, often a stark contrast is drawn between group agents and those groups that are not agents in terms of their internal structure and moral responsibility. This line is becoming much more permeable in Collins’ new account: with only a few words, I can change the ontological status of my group with implications for the moral obligations its members hold and the way they ought to reason about their own course of action. Her approach suggests that it often is very easy to turn a combination into a coalition and into a collective, but we only are required to do so if we are able “to form a collective that will produce that morally valuable outcome—and if no other group will do so in a better way” (p. 97). This transformation, as we have seen, boils down to communicating a plan for the joint activity to others such that the ontological categories, which Collins claims to be central to her “tripartite” account, appear superimposed on more fundamental but tenuous epistemic distinctions.

It is difficult to make general statements on how intuitive any philosophical theory ought to be. Yet, it is generally seen as a virtue of a theory if it concurs with widely held intuitions. On Collins’ account there is a serious ontological and moral difference between a first scenario where Alice and myself act as we did, dividing up the tasks without losing many words over it and merely exchanging glances, and a second scenario where I say (as I, in fact, did): “You take care of the horses while I attend to Jane”. In the latter case, we are a collective (perhaps a dictatorial one, but Collins is fine with those, and I think she is right to be) and in the former we are “merely” a coalition.

Why do these distinctions matter? They matter, because they have implications, firstly, for what we as members groups are morally obligated to do, and, secondly, for how we should reason about our obligations in collective contexts. As to the first issue, if we are a coalition then we cannot not be blamed for having failed in producing a joint outcome, because that duty cannot be allocated – there are individual coordination duties only. On Collins’ account, the moment we communicate with each other verbally and not merely through gestures the ontological status of our group and, therewith, its moral status change and so do our moral obligations. But why would Alice and I not have had an obligation together (or jointly) to help Jane before I made my plan explicit? What if we had done the exact same thing without talking to one another, each just taking cues from the others’ actions or simply inferring our individual courses of action from what we perceived to be collectively optimal? If there is a morally optimal pattern of actions that is salient, on my view, we can be jointly obligated even if we cannot communicate with the other members of our group (Schwenkenbecher 2019, 2020b).

This takes me to the second issue. Collins is right, in my view, that members’ obligations differ depending on the epistemic set-up of a group. But I do worry that her account of group duties may regularly fail to yield clear answers. Take the aforementioned cyclist who offered assistance when passing by the accident scene. Initially, he is only a member of a combination comprising himself and everyone else who is at the scene. Members of combinations, on Collins’ account ought to reason in I-mode about their obligations – that is, to pick the best response available to them, individually. They do not “we-frame” scenarios they are confronted with. That is, they do not frame them as issues for the group. Our cyclist, however, in offering his assistance and joining in the collective effort, appears to have done just that: he saw himself as part of the group that should assist Jane. And in doing so, did he discharge coordination duties or did he become part of the collective and was covered by the group duty and the collective decision-making procedure? That these questions have no straightforward answer provided by the tripartite account is problematic. My sense is that one of the issues here is the ontological approach itself, as opposed to a purely epistemological approach. What really seems to matter is the agents’ epistemic position. What are agents obligated to do given what they (can) know and how they reasonably perceive the situation? Even though he was only passing by, the cyclist recognized the events he witnessed as calling for collective action and he volunteered to join in the effort. In doing so, it appears that he inferred his course of action from the best collective pattern of action (coalition-reasoning) even if he was not part of a coalition? Collins would reject that, but it is not clear to me that we should. Further, in assessing his actions after the fact, I am not sure if her account allows us to make a call on whether or not he did the right thing.

Or take a non-moral coalition: If the members encounter a moral problem they need to reason as follows, according to Collins:

“What they should do is their part of the optimal pattern for pursuing the morally important outcome, where that pattern includes the other coalition-members taking the optimal means to pursue the non-moral goal. Former members of non-moral coalitions are entitled to make certain presumptions about other members’ actions. These presumptions can form a premise in their reasoning, such that that reasoning yields a definite conclusion about what they should do (rather than yielding only the dilemma of I-reasoning).” (Collins 2019, p. 212)

Assume for argument’s sake that the horse riders were a non-moral coalition (that is, sharing beliefs concerning all group members’ non-moral goals such as trail riding while not having a joint decision-making procedure). According to Collins, if one of us suffered an accident the two remaining riders should each acted on the assumption that the other would continue on her trail ride. But this, in my mind, should apply only if we have reason to assume that others are not updating their beliefs concerning the moral issue at stake and are not adjusting their goals. And, arguably, the riders should adjust their goals and transform the group. It seems then, like in the previous examples, the key to answering the moral question of who is obligated to do what is not so much in assigning ontological categories, but in focusing in on agents’ epistemic positions: their beliefs concerning the moral problem and concerning other group members’ beliefs.

The preceding discussion suggests that an epistemic approach to determining group duties and deliberative strategies may well be preferable to an ontological approach. I have proposed such an approach elsewhere (Schwenkenbecher 2019, 2020b) and have briefly hinted at it above. I argue that we should approach these questions from the perspective of the deliberating agent:

“… two (or more) agents hold duties collectively when they each have reason to consider the collective option best, to include it in their set of options, and to actually take steps towards acting on that option. They will regularly have reason to do so when the collective activity is a contextually given default position or where they have some positive indication of willingness to cooperate from the other agent(s). The latter case includes those cases where one person takes the lead and distributes tasks.” (Schwenkenbecher 2019, p. 164).

Collins aims to produce a social ontology of groups as obligation-bearers. However, I believe that it is really the epistemic conditions agents find themselves in that do the work in Collins’ argument and that the three ontological categories she proposes in her book – the “tripartite model” – are surplus to the epistemic components of her argument, exposing the account to the risk of being unnecessarily rigid and at times counter-intuitive. I have tried to show as much above.

On a different note, to the extent that the trail riders are in fact a collective and therefore incur group-based duties, there is a significant amount of agreement between Collins’ and my account as well as several other accounts in the literature (Aas 2015; Björnsson 2014; Pinkert 2014; Schwenkenbecher 2014, 2019; Wringe 2005, 2010, 2016). Like Collins, all these authors agree that in a case such as the one described above there is a collective duty to do what is best and group members must perform those actions that form part of (or are constitutive of) a collectively available action or outcome. In fact, Collins’ aforementioned permissive ontology means that a lot of groups that have been discussed in the literature as so-called loose, unstructured, or unconstituted groups are collectives. Collectives, according to Collins can have irreducible moral obligations at the group level. I am in complete agreement with Collins that individuals in groups that are not group agents in the strict sense (as, for instance, discussed by List and Pettit 2011) can hold moral group-level obligations (Schwenkenbecher 2014, 2019, 2020a, 2020b). Other authors such as Bill Wringe (2010, 2016) have defended a similar, though not identical view that such groups can hold obligations “as a group” despite not being agents. What is perhaps remarkable here is not so much the agreement between authors, but the fact that Collins sets out on her project in explicit opposition to the aforementioned views. Yet, upon closer scrutiny, it turns out she would agree with what I have called “revisionist” scholars on collective obligations after all (Schwenkenbecher 2019, p. 151) and as such her account possibly fits more seamlessly into the existing scholarship than she realizes.

Having focused mainly on the points of disagreement let me conclude by emphasizing that Collins’ is a sophisticated and rich contribution to the fast-growing literature on collective action and responsibility. The reality of our social interactions is often messy and it is difficult if not impossible to neatly categorize the multitude of groups we each find ourselves in. This is particularly true given that Collins’ meticulous account aims to cover all types of group duties. In doing so, she perfectly rightly, in my view, opts for paying special attention to the epistemic positions of group members and for scrutinizing the type of reasoning that moral agents (ought to) engage in when deciding on cooperating with others.