Rowan Mellor’s “Joint Ought” with a critical précis by David Estlund

43 Replies to “Rowan Mellor’s “Joint Ought” with a critical précis by David Estlund”

1. Joint obligations are weird. Normally, obligations are held by an agent, and normally, they can be action-guiding for that agent. Joint obligations are not like that. Those who defend them can avoid the charge they’re held by no one by insisting that they’re held by many someones. They might be action-guiding—they are when it’s common knowledge that the joint thing needs to be done, and when no coordination is needed to decide who does what, because it’s the same thing that everyone needs to do. (An example: a tyrant will be elected mayor unless a supermajority vote against him, and we all know that close to a super-minority are knocked out with the flu and so unlikely to make it to the voting station). But equally they might not be action-guiding, as they are not when the joint thing that needs to be done has many complex components, and parts need to be distributed carefully. Virgina Held distinguishes between these different kinds of cases in her 1970 Journal of Philosophy paper ‘Can a random collection of individuals be morally responsible?’ which many of those cited by Mellor themselves cite, although it does not make it into his reference list.

   I say an obligation is held by an agent, so I do not believe in ‘joint’ or ‘plural’ obligations, held by many agents ‘together’. I think this is messy language that obscures that these obligations are held nowhere. Back when I was more actively involved in this debate, the disagreement seemed intractable. I would stare at Wringe / Bjornsson / Pinkert / Schwenkenbecher / Schweikard (whoever it was on that occasion) in bafflement and try to say, but who has this obligation? And they would stare with equal bafflement back; how many times can I explain this, the aggregate of agents has it jointly.

   So I came to Mellor’s paper with scepticism, wondering what could possibly be said that hasn’t already been said to vindicate the idea of a ‘joint ought’. A claim that we can refer to things jointly seemed to be doing a lot of work: the trees (together) provide shade in the clearing. But that was never in doubt; many people believe in joint causation, for example. A lot of the difference-making debate and its many applications, e.g. to climate change and global labour injustice, starts from the premise that many agents act in ways that together cause large-scale outcomes like an increase in the concentration of global greenhouse gas emissions or for many people in developing countries to be paid inadequate wages to work in terrible conditions. The question is whether properties can be attributed to pluralities, it is whether pluralities of individuals are the kinds of things that can have obligations.

   Getting to the problem Mellor solves requires accepting joint obligations. Those who accept them say that once we have them, they are ‘distributive’ in Thomas Smith’s sense in his 2009 paper ‘Non-Distributive Blameworthiness’, and so tell us what the individuals of the plurality must do. Mellor denies that, and says that joint obligations and individual obligations coexist; the joint obligation exists and informs how wronged individuals may feel about the individuals in the plurality that fail him. He seems to assume that individual obligations take priority, but I am not sure why. If they both exist, surely in some cases there is a question about which takes priority, and sometimes the joint obligation will win.

   If we reject joint obligations, of course, we don’t get the satisfaction of agreeing that the Slice/Patch patient has been wronged by the failure of slicing and patching. But I’m left wondering why we couldn’t move more directly to the claim Mellor seems to want to vindicate, namely that it might be a specific individual of the plurality who it is most appropriate to resent when you have been wronged by a failure of joint action. Can’t an idea like Held’s ‘mutual responsiveness’—taken up later by Stephanie Collins in various places—get us that same intuition, and without the mysteriousness of an obligation held ‘jointly’?

2. Thank you David for providing a clear and accurate summary of my paper, and for offering such probing, careful, and thoughtful questions. There’s a lot to think about here. But I want to respond to one particular aspect of the precis, in the hope of clarifying my position for participants of this discussion.

   David’s precis raises a question which I don’t think I did enough to address in the paper: namely, the relationship between wronging and wrongdoing. I said that “I wrong you when I act in a way which makes it fitting for you to resent me” (p. 61), while I do wrong if I act in a way which violates a moral requirement. Wrongdoing is insufficient for wronging, since I can fail to comply with a moral requirement without warranting resentment (e.g. if I’m non-culpably ignorant). The question is: is wrongdoing necessary for wronging? Is it the case that a person warrants resentment only if she violates a moral requirement?

   As David points out, I never explicitly answer this question in the paper. But his comments have convinced me that I should. So, here’s my answer. In order for the view I table in Section IV (‘Wronging and Resentment’) to work, I need to deny that wrongdoing is necessary for wronging.

   To see this in action, let’s take a version of the Slice and Patch case where only Slice has bad attitudes – Slice doesn’t care about saving Patient, but Patch would do his part if he thought that Slice was going to do hers. The question we’re asking is: who wrongs Patient here? Who warrants Patient’s resentment? The answer I want to give is that only Slice warrants Patient’s resentment. But what can Patient resent her for? Not, on my view, for violating a moral requirement: there is no requirement obligating her to cut (while there is one obligating her to not-cut).

   In essence, my position is that Slice commits an attitudinal wrong. She warrants Patient’s resentment because of an attitude she has towards him, rather than because she has acted in violation of a moral requirement to which she is subject. But we need to be careful here. As David succinctly explains, I reject the view (which I ascribe to Niko Kolodny) that Slice warrants Patient’s resentment merely because she lacks due concern for his interests. After all, Slice could lack such concern even if Patient had no need of an operation – but that would seem like a different, much less serious type of wrong.

   What is different about the original Slice/Patch case – where Patient needs the operation – is that, here, Slice’s bad attitude is sufficient to cause Patient’s moral claim to the life-saving operation to go unmet. This is where joint requirement comes into play. On my view, Patient’s moral claim corresponds to a joint requirement on Slice and Patch to deliver the operation; in order to say that Patient has this claim, we have to posit this joint requirement. Slice warrants Patient’s resentment, and thus wrongs him, not just because she lacks concern for his interests, but moreover because this lack of concern suffices to ensure that this joint requirement is not complied with, and so that Patient’s moral claim goes unmet.

   This raises a question (which I think David is pressing in his precis): if Slice wrongs Patient by having an attitude which suffices for Slice and Patch not to comply with their joint requirement to operate, then does this joint requirement imply an individual moral requirement on Slice not to have such an attitude? More generally, if A and B are jointly required to X, then are A and B each individually required to have attitudes conducive to their X-ing, and not to have attitudes which suffice for their not-X-ing? If the answer is ‘yes’, then joint requirements will turn out to imply individual requirements – although this would still not imply that if A and B are jointly required to X then they’re each required to do their respective parts of X (which I reject).

   At the risk of waffling on, let me briefly address this question. My gut instinct is to say, no, a joint requirement on Slice and Patch to operate does not entail an individual requirement on Slice not to have attitudes which suffice for Slice and Patch not to operate. To motivate this, consider an individual case. Suppose that I (morally) ought to X. My believing that I ought not to X is not conducive to my X-ing, and may even suffice for my not-X-ing. But does my moral requirement to X imply a further moral requirement not to believe that I ought not to X? That sounds implausible to me. Similarly, I find it implausible to say that because Slice and Patch jointly ought to operate, Slice is morally required not to have any attitude which suffices for Slice and Patch not to operate. Slice may wrong Patient by having such an attitude (e.g. if she doesn’t care about Slice’s interests), but she doesn’t thereby violate a moral requirement.

   Of course, there are further things to be said about the relationship between volitional moral requirements and the fittingness-conditions of attitudes; and a fully worked-out theory of joint requirement will have to take part in that discussion. I’m hoping to think more about that in the near future, and I welcome any suggestions.

3. This is a fascinating paper addressing one of the most recalcitrant topics in recent social ontology. I share Holly’s self-reported bafflement at the claim that a duty can be held by an ‘aggregate’ – something that is not an agent. I think Mellor’s paper might help demystify this claim. I want to thank not only Rowan Mellor, but David Estlund for the helpful summary and criticism.

   The paper is quite sophisticated; I suspect that my criticism does more to reveal my own ignorance and misunderstandings than it does any shortcomings in the paper. Basically, I worry there might be a possible equivocation in the basis of resentment toward Slice and Patch in cases where a) one of them has a bad a motivations, versus cases where b) they are both badly motivated. And I worry that resolving the equivocation results in a dilemma.

   Take a case where Slice but not Patch has a bad motivation. Call this Case 1. According to Mellor, Patient is wronged here and so has grounds for resentment – but specifically toward Slice and not Patch. Why? Mellor writes that it is “fitting for him to resent Slice for preventing his claim to the operation from being met, if she does so out of disregard for his interests.” Notice here what Slice is being resented for: viz., preventing a claim from being met. Likewise, later on, Mellor says that “Patch’s lack of due concern explains why the surgeons fail to act as they jointly ought and meet Patient’s claim; Patch’s declining to do his part of the operation is sufficient for this joint failure, and he declines because he would rather golf than save Patient”.

   So far, so good. Now turn to a case in which both Slice and Patch have bad motivations. Call this Case 2. In such a case, Mellor says that Patient can reasonably resent each of them. “This is because both surgeons disregard the interest underwriting Patient’s moral claim to the operation; both of them prefer golfing to saving Patient’s life. Moreover, their disregard explains why they fail to meet Patient’s claim and act on their joint requirement; both surgeons decline to do their parts of operation because they would rather golf.”

   It seems at first glance that the basis for resentment is the same in the two cases. But notice that in Case 2, Patient’s death is overdetermined. If Slice acted differently, Patient would have been saved. Not so in Case 1. If any one of them acted differently, Patient still would have died. It seems, then, that the basis for Patient’s resentment must be different in these two cases. In Case 1, Slice’s disregard makes a difference – it explains why they both fail to meet Patient’s claim. But in Case 2, Slice’s disregard doesn’t make a difference and Patch’s disregard doesn’t make a difference; on what basis, then, can Patient resent either of them?

   There are some moves Mellor might make at this point. He might say that the basis for resentment relies solely on the attitude of (dis)regard the individual shows, rather than on the effects of his actions. But he explicitly denies this view.

   Alternatively, Mellor might say that Slice’s and Patch’s respective disregard jointly explains why they fail to meet Patient’s claim. But this leaves us with a basis for joint resentment rather than individual resentment. My worry, then, is that Mellor hasn’t provided an adequate basis for individual resentment given his commitment to an “attitude+” model. But as I suggested at the outset, I suspect I’ve gone wrong somewhere; maybe Mellor or another participant can set me right!

4. Thanks for the great paper Rowan, to David for the terrific intro, and excellent discussion so far.

   The challenge for Bad Motives, everyone agrees, is that the bad motives of Slice and Patch do not, on their own, ground attitudes about the failure to perform the operation, and the resulting harm. But Patient has a complaint about the failure to perform, and the resulting harm. As Slice and Patch are each individual required not to act, there are no failures of individual duties to connect the bad motives to the failure to perform. So to get an intuitively satisfying solution, we need something to do this connecting job. Rowan suggests joint duties, as another possibility over those rejected, such as group agents with duties, plural duties, and so on. As there is no plausible group agent, as Holly notes, if there are duties of these kinds, they are held by non-agents, which seems like a price.

   Another way of connecting motives to outcomes, though, is just by causation. The bad motives of Slice and Patch together cause the failure to perform, and the resulting harm. Slice has a had motive that contributes to this cause. Call this view ‘Causation’. One attraction is that it does without dubious talk about duties not held by agents.

   Such a view needs careful elaboration to ensure that it doesn’t overextend. One reason is that Causation is only plausible if we restrict the kind of causal relation – funky causes won’t do. But that’s true of causal views of many other moral phenomenon. We also need a view about what it means to contribute to a cause, but there is plenty of work on that in the philosophy of causation.

   A second reason that might be offered is that it is perhaps insufficient for a complaint that X has a bad motive that causes X not to prevent harm to Y. Suppose it would be very costly, and so supererogatory, for X to rescue Y from harm. X is disposed to bear the costs, but not for the sake of Y, who X inexplicably dislikes. Does Y have a complaint, and is resentment warranted. I’m not sure about this. Perhaps Y has a complaint – they have no complaint against not being rescued simpliciter, but they have a complaint at not being rescued due to Y’s dislike. Similarly, it might be argued that Patient lacks a complaint against Slice that the operation is not performed simpliciter, but they have a complaint against Slice that the operation is not performed due to the combination of the motives of Slice and Patch, given Slice’s contribution to the cause.

   Suppose this is rejected. We might nevertheless argue as follows. The bad motives of Slice and Patch cause Patient not to be rescued. The only reason why Slice and Patch lack an obligation to do their bit is the fact that they cannot get the other to act – they are not required to perform the operation because they can’t get it performed successfully. This is just a failure of Ought Implies Can (OIC). We might then argue as follows. Suppose X and Y have bad motives that both 1) result in Z being harmed; and 2) make it true that they lack individual obligations to avert harm to Z just because their motives ensure OIC cannot be satisfied; then resentment is warranted just as it would be if X and Y had individual duties to Z.

   This view seems plausible in the individual context too. Consider:

   Racist Bomb Disposal: X is a racist expert in bomb disposal. Y is an Arab who has been strapped to a bomb that needs disarming. X could normally dispose of the bomb. However, as they hate Arabs they lack the ability to focus on disarming adequately. So they can’t disarm the bomb.

   Plausibly, X is not required to disarm the bomb, because they can’t – OIC is not satisfied. But Y surely has grounds to resent X.

   Finally, here is a variation that this view seems well placed to explain, but that joint duties might struggle with.

   Slice and Patch (Opposites). Slice and Patch are influenced by each other to have the opposite attitudes from the other. If Slice likes Patient, Patch will hate them and won’t Patch. If Patch is likes Patient, Slice will hate them and won’t slice. Slice likes Patient.

   The fact that Slice and Patch are influenced in the way they are also ensures that neither Slice nor Patch can individually ensure the operation goes ahead. But I am at least tempted by the view that only Patch is the appropriate target of resentment. I also find it implausible that Slice violates a joint duty to Patient, given that Slice is well disposed. Patient can resent Patch, though, because Patch’s bad attitudes contribute to the cause of the operation not going ahead.

   Anyway, the general idea is that perhaps we can just avoid all this talk of duties, and just rely on causation to explain the connection between bad motives and outcomes that grounds complaints and resentment in cases where individual duties, for various reasons, fail.

5. I really enjoyed reading the paper and the precis. I think the paper is a great contribution to an exciting debate, and Mellor makes many interesting points. But there are a few crucial features that are not really discussed in detail, and I think these features matter a lot for understanding the relevant individual moral requirements in these scenarios and whether the joint ought must be ontically reducible or irreducible (whether claims about what they ought to do is made true by what they individually ought to do, see Wringe 2016).

   First, the case does not make clear whether communication is possible. In some cases, it can be as simple as signaling that one is willing to perform one’s part of the optimal collective pattern in order to create the relevant sort of salience (e.g., see Bjornsson 2020, Collins 2019, Schwenkenbecher 2021). If communication is possible, it should be clear that we can easily explain what is morally wrong about the actions of Slice and Patch: they fail to try to establish any sort of collective action (note that this is not necessarily the same as Collins’ collectivization duties). And if say Patch refuses when prompted to help by Slice, it should be clear that Slice discharged his duties (and some of these duties may depend on mediated joint abilities, see Pinkert 2014). As this is not discussed (aside from briefly mentioning common knowledge when discussing group agency), I’ll assume that communication is not possible (or overly costly).

   Second, it is unclear to me whether Mellor is concerned with objective, belief-relative, or evidence-relative moral requirements, but perhaps I overlooked something. This matters a great deal for understanding the relevant moral requirements of the agents. Beliefs briefly come up in Mellor’s discussion of Kolodny. For example, in the good-motive version, each agent declines to help the patient because they reasonably, but falsely, believe that the other would not do their part if they were to do their own. But these are beliefs about facts about the psychological state of the other agent and such facts could still be relevant for objective duties. Of course, as soon as we focus on objective oughts, concerns about the action-guidingness of moral requirements can be thrown out the window anyway (e.g., the familiar Mine Collapse or Doctor Jill cases raised by Regan, Jackson, Zimmerman, etc.). I think this (among others) is reason for many to reject objectivism about reasons and oughts, and a more plausible position is some version of perspectivism or contextualism (though I’m sure not everyone agrees with this).

   Third, even if we’re concerned with objective duties, the case should not be described in a manner where the wrongdoing has already occurred. At least, not if we’re concerned with figuring out the relevant moral requirements, for backward-looking moral responsibility it’s obviously a different story. This may affect our intuitions in a manner that is favorable to joint/plural oughts. In light of all the facts, at the time of decision-making, we can accept that the psychological facts about Slice and Patch are such that they intend to go golfing. But it should still be left open whether they actually do. A person could still change their mind. Now, clearly, for any belief-relative or evidence-relative view of moral oughts, this will be more difficult to ascertain for each agent unless they can communicate, but I’m sure we can think of some cases. But even for objective duties, this I think must not be left out. We shouldn’t treat the future as fixed/determined for determining our moral requirements right now (for example, it would make little sense to talk about diachronic duties if we hold the future as fixed). But in the case as described, the wrongdoing indeed occurs, which only reaffirms the intuition that it is certain or inevitable that the other agent will go golfing. But this is not a certainty at the time of decision-making, and it should not be treated as such. This excludes the possibility that the person comes to their senses. Of course, uncertainty becomes even more a factor on evidence-relative or belief-relative views. This is as far as I can tell not considered in the discussion between actualism and possibilism in the paper, but it seems very relevant to me.

   This may all seem like nitpicking, but these three features become very relevant when considering another feature: the stakes involved. This is not yet in print, but I think the stakes matter greatly for the individual duties in collective action problems where communication is not possible. Let’s say that a high-stakes case involves two competing collective options of high moral value. For example, suppose Patch and Slice’s alternative option is not to go golfing, but to perform a non-life threatening surgery for Patient 2 that is in extreme pain (but it needn’t be a collective action, it suffices if their actions both have considerable moral value). If they cannot communicate, and there is great uncertainty that the other will do their part, then this is indeed what they both ought to do given the stakes involved. But in such cases with high stakes it is no longer clear that there is a joint ought on part of them both to save Patient 1 with a life-threatening disease nor is it clear that Patient 1 is wronged by them nor that Patient 1 can appropriately resent them. However, if the case involves low-stakes, meaning one collective option has a high moral value, but the other option does not, then I think given the uncertainty involved, and the negligible value of the second option (going golfing), this may simply point to them indeed having violated a moral requirement to do their part of saving the patient: They each have an atc moral requirement to do their part. In that case, there is indeed a joint ought, but it is reducible to their individual oughts. I’m sure more is to say about this, but this strikes me as a plausible explanation of these types of cases, and it would show that a joint ought is not ontically irreducible, and that these oughts can be action-guiding (if they’re evidence-relative, which is the view I favor).

   I hope this does not sound too critical, I really enjoyed the paper and I look forward to reading the rest of the discussion.

6. Thanks to everyone for a great discussion, and especially to Rowan and David.

   My first comment is related to Niels’. I think that the case of Slice and Patch is under-described in all kinds of ways, and that we should worry that any intuitions we might have about the case are based on how we fill in the details. The description of the case suggests that Slice and Patch each has only two options: (1) perform his or her role in the surgery, or (2) go golfing. This is highly artificial. Why, for example, is showing up for the surgery, and only performing one’s part if the other shows up ready and willing to perform the surgery, not an option? Why is not possible for each doctor to do something to influence the other doctor to perform his or her part? I strongly suspect that the reason why it seems plausible that each doctor warrants resentment is that there was something that each of them could have done to influence the other to do his or her part, and each doctor failed to do it. If, instead, each doctor is so hell bent on going golfing that nothing would get him or her to help with the surgery, then the doctors’ motives seem so pathological that I don’t really know what to say about whether either one of them warrants resent. Maybe the appropriate response is simply abject horror at such inhuman callousness.

   My second comment is about Mellor’s views on the transmission principle. He rejects plural transmission but accepts individual transmission. Yet Slice and Patch is so structurally similar to Professor Procrastinate that it seems to me that both principles stand or fall together. Mellor writes that possibilists reject treating Procrastinate’s future behavior as fixed, because the behavior “is entirely up to *him*.” But possibilists typically mean that his future behavior is up to him in the sense that if he accepts, then reviewing the book will be up to him in due course. At the time at which he must accept or decline, he has no more influence on his future behavior than Slice has on the behavior of Patch. If, nevertheless, the “diachronic” ought that says that he ought to accept and review entails that he ought to accept, then it seems to me that the “joint” ought that says that Slice and Patch ought to perform the surgery entails that each ought to do his or her part.

7. Thanks Dave, Rowan, and all for a very interesting discussion. As Dave mentioned in one of his comments, Rowan and I have a forthcoming paper which explores the semantics of joint ‘ought’ claims. We try to make sense of cases in which ‘A and B ought to φ and ψ respectively’ is true, while ‘A ought to φ’ and ‘B ought to ψ’ are both false.

   Drawing on this work, and in response to Holly and Saba’s helpfully pressing comments, I’d like to say something about how I understand the claim that a plurality of agents jointly has an obligation. Rowan may or may not endorse the following way of thinking.

   As I see things, the claim that Slice and Patch together ought to slice and patch is nothing more than a claim about what the individual Slice ought to do made relative to a specific context – namely, a context which leaves open the possibility of Slice’s slicing and Patch’s patching. Equally, it is a claim about what the individual Patch ought to do made relative to this same context. So ‘Slice and Patch ought to slice and patch’ is an action-guiding ‘ought’ claim which relative to a specific context attaches to Slice the individual (just as it does to Patch).

   The claim that Slice ought not to slice is made relative to a different context – namely, a context which takes it as given that Patch is going to fail to patch.

   Both claims pick out ‘oughts’ which pertain to the agent Slice.

   Holly writes: “The question is [not?] whether properties can be attributed to pluralities, it is whether pluralities of individuals are the kinds of things that can have obligations.” On my deflationary way of thinking, to say that a plurality has an obligation to φ is just to say that a bunch of individuals, reckoned together, have an obligation to φ. We reckon individuals A and B together in making and evaluating claims of the form ‘A and B ought to φ,’ when we leave open A’s φ-ing and B’s φ-ing. In other words, we do not take it as given that A will fail to φ or that B will fail to φ, even if there is decisive evidence to that effect. By contrast, when determining what A ‘reckoned alone’ ought to do, we do not perforce leave it open that B is going to φ; if it is foreseeable that B will fail to φ, we take this into account.

   There of course remain many difficult questions about exactly what conditions must be satisfied in order for a plurality of agents to be subject to an obligation (e.g. whether they must be able to communicate; how large the pluralities subject to joint obligations can be; etc.); about how blame and resentment are aptly allotted in cases where multiple individuals together have most reason to act in some way which differs from the conjunction of how each of these individuals, reckoned individually, has most reason to act (this is probably the most pressing question); and so on. My remarks go no way toward addressing these questions.

   However, I think we should tease them apart from the more fundamental question of what it is for a plurality to have an obligation. On my deflationary way of thinking, for a plurality to have an obligation is for a bunch of individuals, reckoned together (as briefly described above), to have obligations. It is not for something that is not an agent to be subject to an obligation.

8. Thanks, everyone! A question for Rowan and Victor about Saba’s comment:

   It seemed to me that a causal account (with or without joint requirements) ideally addresses Saba’s query about Saba’s Case 2. A causal account should explain how someone contributes to causing a claim to go unmet when that person’s contribution to the outcome is overdetermined.

   Rowan – I read your first comment as seeking to argue that in Saba’s Case 2, because each person’s bad attitude is sufficient to make it true that a claim will go unmet, each person’s bad attitude is sufficient to cause that outcome. Saba’s comment might be trying to highlight the fact of overdetermination as a counterargument. Are overdetermined contributions to an outcome insufficient to cause it, even if sufficient to make it true?

   Victor – Maybe the work you mentioned on the philosophy of causation can help here? (You’d like to think so!)

9. Thanks to everyone who has commented so far – this is a great discussion! I’m going to respond to some comments together, and others separately.

   Holly & Maggie: Holly raises an important issue – if Slice and Patch don’t comprise a group agent, then it seems like saying that they bear a joint obligation to operate implies that things which aren’t agents can have obligations, and that’s a big cost. The response which I – along with e.g. Bjornsson, Pinkert, Smith, Schwenkenbecher – try to press is that there’s no one thing which has this obligation; rather, it’s held jointly by many agents. But Holly replies: sure, some properties can be possessed jointly by pluralities; but you haven’t done enough to convince me that ‘ought to φ’ is such a property. One possible thing to say here is that if positing joint oughts gives us a powerful solution to the Slice/Patch problem, then that’s at least some reason to think that ‘ought to φ’ can be possessed jointly.

   Maggie raises another possible reply (which comes out of a forthcoming paper we co-authored – David mentions it in a footnote). We could just drop talk of properties all together. The idea that ‘ought to φ’ is a property fits best with a semantic theory (defended by Mark Schroeder) according to which the word ‘ought’ (or at least, the action-guiding disambiguation of ‘ought’) is to be analysed as a two-place predicate, relating an agent to an action. But this is not the standard semantic analysis of ‘ought’. A more mainstream analysis has it that ‘ought’ functions like a necessity operator, and says that ‘A ought to φ’ is true iff A φ-s at the best possible worlds. Angelika Kratzer’s contextualist theory (the best-known representative of this view) tweaks this, adding that ought-sentences are always evaluated relative to a context which holds fixed some set of propositions and standards, in turn fixing the set of ‘live’ worlds and the ranking of those worlds.

   On a Kratzerian view, ‘Slice and Patch ought to slice and patch’ can be glossed as something like ‘it would be best if Slice and Patch were to slice and patch’. Maggie and I argue that this sentence has to be evaluated relative to a context which leaves it open that the surgeons will slice and patch. Why? Because, we claim, there’s a pragmatic rule which tells speakers/hearers not to rule it out that the subject/s of an ought-sentence will do the thing which that sentence says they ought to do (unless it’s clear at the context that they cannot do this thing). In the case of ‘Slice and Patch ought to slice and patch’, Slice and Patch both occupy the subject position, and so we leave it open that they’ll do what the sentence says that ought to do viz. slice and patch. So, at this context, the sentence is true – the worlds where the surgeons slice and patch are best.

   However, the sentences ‘Slice ought to slice’ and ‘Patch ought to patch’ are evaluated relative to difference contexts. For the first sentence, we leave is open that Slice will slice (since she occupies the subject position) but we hold it fixed the Patch won’t patch (since this is clear at the context); and vice versa for the second sentence. Given this, both sentences come out false; given that Patch won’t patch, the world where Slice slices is not best, etc.

   What this shows, I think, is that we can make sense of the thought that ‘Slice and Patch ought to slice and patch’ is true, but that ‘Slice ought to slice’ and ‘Patch ought to patch’ are both false, without leaning on the idea that ‘ought’ is a property that can be possessed jointly.

   Maggie writes: “the claim that Slice and Patch together ought to slice and patch is nothing more than a claim about what the individual Slice ought to do made relative to a specific context – namely, a context which leaves open the possibility of Slice’s slicing and Patch’s patching.” Maybe I’m missing something, but I wouldn’t quite say that. As I see it, this is *not* a claim about what Slice ought to do. Rather, it’s a claim about what Slice *and* Patch, taken as a plurality, ought to do. That is why the claim gets evaluated relative to a context which leaves it open that they’ll slice and patch; both individuals occupy the subject position. The broader point is that you can’t reduce claims about what many agents ought to do to claims about what they each ought to do. (It’s likely, however, that I’ve misread Maggie here).

10. I’m a big fan of Rowan’s work, and I think this paper is an all-round excellent contribution to a very difficult and important debate. Thanks also to David for the extremely helpful précis.

    I find Rowan’s arguments against alternative views extremely helpful. In particular, I thought that his argument against the group agency view, along with David’s summary, really spell out perfectly why such views cannot do the necessary work.

    However, like others, I find the proferred “joint ought” account unsatisfactory.

    Rowan wants to say that Slice and Patch “jointly ought” to save Patient’s life, but Slice ought not to cut and Patch ought not to stitch.

    But *why* is in the case that Slice and Patch *in particular* have this obligation, rather than some other people having the obligation? This is not a standard rescue case, where any two bystanders could save a drowning stranger. Slice and Patch *in particular* are obliged to save Patient’s life because Slice is an expert in cutting and Patch is an expert in stitching.

    And their obligation in this case is not to save Patient’s life *in general* – they have no special obligation to defer golfing in order to prevent Patient from being hit by a runaway trolley. Their obligation as doctors is specifically to save Patient from dying from this specific illness, by each playing their appointed role – by Slice cutting and Patch stitching.

    Given this, I find Rowan’s rejection of the transmission principle for such joint oughts surprising – if the joint obligation is [to save Patient’s life, from this particular illness, by *Slice cutting* and *Patch stitching*] and this fact is in part explained by Slice’s expertise in cutting and Patch’s expertise in stitching, then it seems just weird to deny, as Rowan appears to, that Slice has no reason to cut and Patch has no reason to Slice!

    By contrast, a “reasons pluralist” view (like Woodard’s pattern-based view) seems to get the right result. Patch and Slice each have a reason to play their parts in bringing about the following result [save Patient’s life, from this particular illness, by *Slice cutting* and *Patch stitching*], and each playing their part means that Slice cutting and Patch stitching.

    Now, *if* one of them has decided to ignore this pattern-based reason, and go golfing, then the other will have an individual reason *not* to play their part, which plausibly outweighs the pattern-based reason. But in deciding to go golfing whatever the other surgeon does, either agent would be doing something that contravenes their pattern-based reasons (to play their part in bringing about the best possible outcome) and does something they have no individual reason to do either (since there is only individual reason for either to refrain from surgey *once* the other has resolved not to cooperate).

    In other words, *if* both Slice and Patch are determined to go golfing by time T, then it will be true that for both of them the pattern-based reason to play their part in the surgery is outweighed by the individual reason not to cause useless suffering. But it will also be true that, whichever of them first decided to go golfing come what may (or both of them, if they decided at the same time), did something at time T-1 that they had most reason not to do. So, over the entire span of time, at least one of them acted wrongly.

    On this view, we may have both we-reasons to play our parts in bringing about the best outcome open to us collectively, and I-reasons to bring about the best outcome open to each agent individually. Sometimes, these will clash, such that an agent has most reason not to play her part in bringing about the best outcome. However, we usually also have *both* we-reasons and I-reasons to ensure that, in future, our own and others’ we-reasons and I-reasons align rather than clashing – to ensure that the best thing each agent can do, given what the others will do, is to play their part in bringing about the best result that all the agents together can bring about.

    Maggie Shea’s interpretation of the view seems different to me from Rowan’s. However, I’m not sure it helps in determining the action-guiding content of the obligations in question. On this view, relative to the context in which Slice takes Patch’s actions to be open, she has an obligation to cut, whereas relative to the context in which Slice takes Patch’s actions to be determined, she has an obligation not to cut.

    But in determining the action-guiding question of what *she* has most reason to do, which context is relevant to Slice? After all, Patch *could* decide to play his role and stitch, even if he doesn’t intend to. If each agent does what they ought to do, relative to the open context, they will bring about the optimific result and perform the surgery. If each does what they ought to do, relative to the closed context, then they will bring about the second-best result, and leave Patient to die without inflicting unnecessary suffering. If one does what they ought to do relative to the closed context, and one does what they ought to do relative to the open context, then they will bring about the worst result, and both leave Patient to die and cause unnecessary suffering.

    In other words, I think the question of which context is relevant in determining our action-guiding reasons is subject to the same kind of collective-action problem that we started with.

    Ultimately, I think there’s never going to be a totally satisfactory solution to these problems – every answer is *some* version of the claim that, in some sense Slice and Patch are doing something wrong, and in some sense Slice and Patch are doing something right. There just is a bump in the rug, and we’re all moving it around. The question is where it’s best to move it to.

11. Hi Rowan,

    I don’t think you’ve misread me, but I also suspect there is no deep disagreement between us.

    To clarify, I believe (in line with our shared paper you summarize) that there is a pragmatic rule to the effect that *in general* when we talk about what some individual agent A ought to do, we treat B’s foreseeable future actions as fixed. So it would be highly misleading to say ‘Slice ought to slice.’ Again, a claim of this form (with only the name Slice in the subject position) is naturally made and interpreted relative to a context which holds fixed Patch’s foreseeable failure to patch.

    However, ‘Slice and Patch ought to slice and patch’ is still a claim about what Slice ought to do. Relative to one and the same context, ‘Slice and Patch ought to slice and patch’ entails ‘Slice ought to slice.’

    You say: “you can’t reduce claims about what many agents ought to do to claims about what they each ought to do.”

    I agree with this claim, IF it is interpreted in the following way. Claims of the form ‘A ought to PHI’ are naturally made and evaluated relative to one context, while claims of the form ‘A and B ought to PHI’ are naturally made and evaluated relative to a specifiably different context. In general, in ordinary thought and talk, we can’t felicitously infer ‘A ought to PHI’ from ‘A and B ought to PHI.’ For the latter claim is true relative to a context relative to which the former claim may well be false (as with ‘Slice ought to slice’ and ‘Slice and Patch ought to slice and patch’).

    Still, relative to a single context you CAN reduce claims about what a plurality of individuals ought to do to claims about what each individual in this plurality ought to do.

    The relevant “individual obligations” will NOT be what most of us have in mind when we talk, ordinarily, about what an individual agent ought to do. But there is still a very good (if somewhat technical) sense in which claims about what a plurality of individuals ought to do are simply claims about what those individuals ought to do (relative to a specific context).

12. Saba, Victor, and Kartik: Saba raises a great question, and Kartik anticipates my answer. In Case 2, where both surgeons have bad attitudes, and so Patient’s death is overdetermined, it’s still true – I think – that each surgeon’s bad attitude is *sufficient* for Patient’s death. It’s just that neither is necessary.

    Victor, your causation view is interesting. (I tried to address something like it in Section II.A. of the paper, but maybe this didn’t go far enough). It might help to compare my view and the causation view side-by-side.

    Causation view: Slice and/or Patch warrant Patient’s resentment if either of them has a bad attitude which causally contributes to harm to Patient.

    My view: Slice and/or Patch warrant Patient’s resentment if either of them has a bad attitude which suffices for a moral claim of Patient’s to go unmet.

    The main difference is that my view says that Patient is not merely harmed, but has a moral claim (corresponding to a (joint) obligation) violated. My thinking is that this is a more intuitive description of the case. Patient not only stands to have something bad happen to him (as a result of the surgeons’ attitudes); he has a claim on the surgeons to save him, and they have a corresponding obligation to do so. Moreover, either surgeon warrants Patient’s resentment when their bad attitude suffices for them to fail to meet this obligation.

13. Hi Max,

    Thanks for your very interesting comment.

    You say: “In other words, I think the question of which context is relevant in determining our action-guiding reasons is subject to the same kind of collective-action problem that we started with.”

    I reply: I cautiously agree, depending on what exactly you mean by “action-guiding.” In line with what Rowan and I argue in our shared paper, I would want to say that both ‘Slice ought not to slice’ and ‘Slice and Patch ought to slice and patch’ are action-guiding since they pick out courses of action which the implicated agents have most reason to intend to perform.

    Now turning to my current view. I think that what we are really hungry to understand is which context is relevant in determining individual agents’ blameworthiness. So if what you mean by the ‘action-guiding ought’ is the ‘ought’ that tracks blameworthiness-for-not-performing-the-relevant-act, I agree with your quoted remark. That is: I think what we really want to know is which context is the relevant one to consider in determining individual agents’ moral responsibility for actions and outcomes. I’m working on this issue at the moment.

    My precious comments don’t at all address this key normative question. I simply wanted to elucidate the good sense in which joint ‘oughts’ attach to individual agents.

14. *previous, not precious! (LOL)

15. Niels and Sam: Niels raises tons of great questions. For the moment, I’ll address the one which Sam picks up on – can Slice and Patch communicate, and how does an assumption either way affect our judgements about the case? I’m not sure I agree with Niels that an assumption that Slice and Patch *can* communicate necessarily yields a simple solution involving individual obligations. Suppose that Slice knows that Patch won’t patch. Why should Slice communicate with Patch? After all, he knows this will be futile. At best, Slice is under a conditional obligation: if communication will (or at least) might change Patch’s mind, then Slice ought to communicate. The problem is that we can just come up with a case where the antecedent is not satisfied, and the obligation embedded in the consequent is not triggered. And such a case, to my mind, looks like one where Patient stands to suffer a serious moral wrong.

    Niels writes: “we can accept that the psychological facts about Slice and Patch are such that they intend to go golfing. But it should still be left open whether they actually do. A person could still change their mind.” I think I disagree with you here. It seems to me that lots of facts about what I should do take for granted what others will do – I should drive on the right (here in Chicago) because I take it for granted that others will do so. Of course people can change their minds (or fail to act as they intend); and I should remain sensitive to evidence about what others will do – I should look out for people driving on the left. But I act on a presumption about what others will do.

    Sam points out that this is a highly artificial case. I agree; usually, people can be persuaded to change how they intend to act. But that doesn’t matter. The point is that this (artificial) set of assumptions raises a conceptual puzzle – Patient stands to be seriously wronged by the surgeon’s behaviour, and yet neither surgeon (individually) ought to act otherwise.

16. Great paper, precis and discussion. Without taking due care about the details in any of them, however, may I take up the idea that Niels introduces that ‘the case does not make clear whether communication is possible. In some cases, it can be as simple as signaling that one is willing to perform one’s part of the optimal collective pattern’. If Slice and Patch communicate with one another, but without forming a group agent, that will certainly make them jointly fit to be held responsible for whatever they jointly do as a result: fit to be blamed if they let down the patient, fit to be praised if they get to work. But such communication may require just an acquiescence in the assumptions they manifestly make (i.e. make as a matter of common awareness) about one another, not any very explicit exchange. And if even such a simple form of communication is within their joint reach, that surely makes them fit to be held jointly responsible too; there is an option available for them that would enable joint action. As Sam says, the case is under-described. I suspect that the indignation we may feel about how Slice and Patch behave is due to our reading it as a case where they are neglecting something like that option. If they each have good grounds for holding that the other will not turn up, and if there is nothing they can do to sponsor a joint action, then I don’t think that they are jointly fit to be held responsible.

17. I very much like Rowan’s paper, and I think the discussion here is excellent.

    Since some of my ideas are mentioned in the paper and the precis, I hope it is in order to comment briefly on them here. None of what follows amounts to an objection to the paper, or the precis, or any of the discussion above.

    There are two versions of my view about pattern-based and act-based reasons. The old version (before my 2019 book) was that pattern-based reasons can exist even in cases where the other agents required to realise a pattern are unwilling to play their parts in it. I am grateful that Rowan discussed this idea in his paper — but I would point out that I would not have proposed it, even when I held it, as an explanation of the idea that Slice and/or Patch acts wrongly, or wrongs the patient, by going golfing. I would have said that each had a reason to play their part in the operation, despite the fact that the other was unwilling. But I would have quickly added that they should not act on this reason, since it is outweighed by a conflicting (‘act-based’) reason not to play their part, given the harm this would do in the circumstances. I would not have claimed that, in acting as they ought, they nevertheless still somehow wronged the patient.

    According to the more recent version of my view (from my 2019 book), neither of them has a pattern-based reason to play their parts associated with the pattern , because the other is unwilling to play their part. However, it is possible that they have some relevant pattern-based reason associated with a different pattern. For example, they each might have a reason to seek to persuade the other to cooperate, associated with the pattern , in which most medics willingly do cooperate. (I now think that we should not cite patterns that would not be realised — but we might want to cite patterns that extend more widely than the immediate protagonists in this case.)

    So, like others, I am inclined to think that the case is underspecified. If the options are just ‘operate’, or ‘go golfing’, I think each should go golfing, given the other’s dispositions, and neither does anything wrong, or even has a reason, to operate. But if the options include something like ‘seek to persuade the other to operate, rather than to go golfing’, then I think they have pattern-based — and act-based! — reasons to do that, and act wrongly if they do not, and wrong the patient if they do not.

18. My angled brackets have not appeared. I will replace them with exclamation marks. The first sentence of the third paragraph should read:

    “According to the more recent version of my view (from my 2019 book), neither of them has a pattern-based reason to play their parts associated with the pattern !Slice cuts, Patch stitches!, because the other is unwilling to play their part.”

    The third sentence of the fourth paragraph should read:

    “For example, they each might have a reason to seek to persuade the other to cooperate, associated with the pattern !medics treat patients in need!, in which most medics willingly do cooperate.”

19. Rowan, to clarify, I don’t think that the artificiality of a case is necessarily a problem. Perhaps a better way to put my point would be this. You maintain that this case raises a conceptual puzzle. The puzzle is supposed to be that each doctor does as he or she ought, yet the patient also has reason to resent Slice and Patch for declining to do their respective parts in the surgery. My concern is that the intuitive response that Slice and Patch each ought to go golfing relies on the artificial construal of the case (where each doc has only two options), whereas the intuitive response that the patient has reason to resent each of them relies on us filling in the details of the case based on reasonable assumptions about other options that each doctor has. There clearly is some clash between these intuitive responses, but if each is based on a different way of understanding the case, then there’s no genuine puzzle. In the artificial version, each doctor does what he or she ought to do and resentment is unwarranted (except perhaps resentment based on the doctors’ motives). In more realistic versions, resentment is warranted, because each doctor fails to do something that he or she ought to do.

    I hope that this at least clarifies what I was getting at in saying that the case is artificial.

20. Hi Rowan!

    I have a (hopefully) short clarification question (having read the precis but not the paper, and this not being my area): what is it that makes the joint ought “stick” (so to speak) to Slice and Patch, rather than some other set of surgeons or agents? What is it that puts the requirement on *them*? In this case, I suspect the answer is “because they’re the surgeons assigned to the case” or “they’re the only ones that can do it”. Is that what’s doing the work?

    Follow-up question: I can think of parallel cases in which features that I suggested above aren’t present. Some of these cases make the joint ought diagnosis look a bit puzzling to me, in part because it’s not clear to me which agents have the joint obligation (if any do):

    Say (and this might be a bad case) i’m passed out and drowning. It would take two persons acting together to save me and there are plenty of competent swimmers around. If only one person comes to tries to come the rescue, i’ll wake up and experience significant additional trauma and distress before drowning anyway. As it happens, no one will come to help regardless of what the others do.

    This look relevantly analogous (to me anyway) to the Slice and Patch case. It looks like a moral wrong has happened here, but it also doesn’t look like anyone has acted impermissibly (perhaps everyone acted as required). But on the other hand, it’s not clear to me what appealing to a joint ought would look like as a solution to this more disparate case or who the joint ought would apply to.

    Is this analogous? What’s your analysis in this kind of more disparate case?

    Thanks!

21. Hi Max, thanks for your comments. Suppose that Slice would do her part if she thought Patch would do his, but Patch intends to golf regardless of what Slice does. (Assuming Chris’s earlier view of we-reasons – which you seem to be adopting here) it seems plausible to say here that, when Patch formed this intention, he contravened both a we-reason and an I-reason: intending to patch is both part of the best pattern, and Patch’s best option (since it will foreseeably trigger Slice to intend to slice, thus triggering the operation).

    But now suppose that neither surgeons intends anything regarding whether to golf or slice/patch. If either surgeon were to form the intention to golf in this scenario, they might contravene a we-reason. But it’s not clear to me that they would contravene an I-reason. If Patch has no intention, is intending to slice Slice’s best option? Not obviously – if he ends up patching, then slicing is good, but if he doesn’t then slicing is very bad. As such, it’s not clear to me that a surgeon must always have failed to do what they had most reason to do at the time at which they form an intention to golf.

    Another point, which I make briefly in Section II.C of the paper. The view you suggest says that Slice and Patch have we-reasons to play their parts in good patterns (e.g. slicing and patching, thereby saving Patient’s life). But if we say this, then why not also say that Slice and Patch have we-reasons not to play their parts in bad patterns? – e.g. Slice and Patch have we-reasons not to play their parts in Slice slicing while Patch golfs, thereby causing agonising death for Patient. But if so, then Slice and Patch each have pattern-based reason both to play their parts and not to play their parts in the operation. The we-reasons view would thus seem to yield to determinate result about what the surgeons should do. Whatever they do, they contravene a we-reason.

22. further to max hayward’s comment. he wrote:

    “But *why* is in the case that Slice and Patch *in particular* have this obligation, rather than some other people having the obligation? This is not a standard rescue case, where any two bystanders could save a drowning stranger. Slice and Patch *in particular* are obliged to save Patient’s life because Slice is an expert in cutting and Patch is an expert in stitching. And their obligation in this case is not to save Patient’s life *in general* – they have no special obligation to defer golfing in order to prevent Patient from being hit by a runaway trolley. Their obligation as doctors is specifically to save Patient from dying from this specific illness, by each playing their appointed role – by Slice cutting and Patch stitching.”

    — i teach carolina sartorio’s two-buttons case (from her ‘how to be responsible for something without causing it’), and the students tend to be very resistant to the idea that two employees of the same factory, people with specific dedicated safety roles! could not have stronger obligations speaking to their employment roles and their position in an organizational structure. of course we could argue about whether the ‘high-risk chemical plant’ sartorio imagines is itself a corporate agent, and whether that’s the lowest level of collective agency relative to the two-buttons explosion, and how obligations distribute from those collective agents (nested or otherwise) down to the individuals in the respective rooms. but i cannot recall any student agreeing that in such a context, we should proceed *as though* this case has a similar structure to the standard rescue case with complete strangers that max mentions. this pushes in the direction philip notes as well, that even without group agency, we sometimes still think there are conditions that require (and maybe structure) individuals to be more responsive to each other, or to act in ways that are more conditional / interdependent on each other.

    maybe it would be worth running a version of slice/patch where they are complete strangers with relevant expertise not structured by institutional roles or working for the same hospital, and to see if the intuitions about appropriate resentment change…?

23. & further to margaret shea’s first comment. she wrote:

    “As I see things, the claim that Slice and Patch together ought to slice and patch is nothing more than a claim about what the individual Slice ought to do made relative to a specific context – namely, a context which leaves open the possibility of Slice’s slicing and Patch’s patching. Equally, it is a claim about what the individual Patch ought to do made relative to this same context. So ‘Slice and Patch ought to slice and patch’ is an action-guiding ‘ought’ claim which relative to a specific context attaches to Slice the individual (just as it does to Patch).”

    i agree with this analysis. in fact, although i did not put this in terms of ‘contexts’, i think the proposal i defended in my 2013 paper ‘non-ideal accessibility’ about the professor procrastinate case is more or less this solution. mine went something like: there’s one non-conditional obligation which is relative to the best possible world, and then there is an ordering of successively-less-ideal obligations phrased with ‘given that’ operators in front corresponding to successively less ideal worlds. so the professor ought to do his reviewing, but given that he won’t, he ought to refuse the assignment. and so on. (side note: this paper is funny in light of this discussion thread. i was mad at david estlund for pulling out of examining my phd thesis after an already-very-long-wait, so i called the procrastinating professor “david”. hah! what a minor revenge).

    but/so, even though i like this proposal, what i don’t get it why you’d call the top (or any high-up) obligation a ‘joint ought’. the whole story works for individual obligations, it’s just that the better contexts permit individual obligations to interact with what others do in a way that produces the most desirable outcomes. best-context-relative oughts are still individual oughts. what am i missing?

24. On Philip’s post (and some others): I’d like to come back to what difference it makes if Slice and Patch could communicate. I think these thoughts are in the spirit of what Rowan said about communication just moments before Philip’s post (and which Philip might not yet have seen). Philip, partly following Niels, seems right to suggest that the kind of communication that would make joint action possible might be pretty undemanding–maybe even tacit. And suppose that the surgery we hope for would necessarily be a joint action, which communication can make possible. Philip then suggests that if they were to be in at least minimal communication this would, “certainly make them jointly fit to be held responsible for whatever they jointly do as a result: fit to be blamed if they let down the patient, fit to be praised if they get to work.”

    First, whatever the circumstances about communication, I don’t know if neither’s doing their part would count them as jointly declining to do the surgery in the intended sense, rather than simply as a failure to act together at all. So, it’s hardly clear that they thereby jointly violate any requirement. And if not (and actually even if so!), what agent does violate any requirement?

    Second, if “they” and “them” signal that the references is to a joint requirement to perform the joint action, the scenario leaves in place the central questions about that idea such as: requirement on what agent? I’m not sure how the introduction of communication and joint action are meant to help with those.

    If instead it is the individuals who are “fit to be blamed,” it’s hard to see how. As Rowan points out, given the example-defining non-cooperative motives of each doctor, Slice’s communicating with Patch would have no tendency to bring about the surgery. Add to that (responding to Niels’ proposed requirement to propose cooperation; and Holly also seems to express a version of this (in her post tagged 1:01am)) the fact that either doctor’s—say Slice’s—attempting to coax the other into doing their part would itself seem to be quite wrong given that Slice will not do her part of the proposed cooperation in any case. Patch’s doing his part would only be worse all around.

    Even though, for several reasons, the Slice/Patch case is not a Prisoner Dilemma* it’s nevertheless structurally like one in certain ways. For example, here we see, I think, that in both contexts the possibility of communication turns out not to accomplish much and for a similar reason. It’s because, given the motives that are defining features of the examples, each doctor and each prisoner will defect whatever the other doctor or prisoner does. That leaves in place the evident rightness of each doctor not making things even worse by alone doing their own part.

    I’m concerned that any temptation to say the possibility of communication would somehow expose a wrongdoer or a requirement to arrange cooperation involves illegitimately thinking they each ought to be differently motivated and to communicate those better motives to each other. That wouldn’t address the Slice/Patch example, but would switch to an easier one. The bad motives are defining features of the case.

    * For an example of how Slice/Patch is not a Prisoner’s Dilemma, unlike the Prisoners Dilemma, it is no part of the Slice/Patch structure that the case in which both doctors cooperate is better from each doctor’s point of view than the case in which neither of them does. We know nothing about each doctor’s utility function over the outcomes except, perhaps, that each prefers not to cooperate whatever the other doctor does. By contrast, in the Prisoner’s Dilemma, each prisoner’s preferring mutual cooperation to either of them defecting is central to what is thought puzzling about the Prisoners Dilemma.

25. Thanks for the helpful responses to everyone. This is really interesting. To say more about why I think communication makes it relatively easy to explain the relevant individual requirements:

    Rowan writes: ‘Suppose that Slice knows that Patch won’t patch.’ What I’m trying to say is that even if this is something that Slice can know, we should not treat such knowledge ascriptions as if this implies an absolute certainty about how the future unfolds. Even if all Slice’s evidence points to that Patch will go golfing, it still makes sense to double check his evidence given the stakes involved. It seems not unreasonable to expect someone to double check this crucial piece of information when lives are at stake. This may be considered part of their duty to establish some joint action. So following Philip’s helpful comment, if any simple form of communication is within reach, it seems clear to me that both of them have failed to discharge their relevant moral requirements, therefore commit moral wrongdoing, and are fit targets for blame.

    If this signaling or communication (whether it is for double-checking or establishing joint action) comes at a significant cost to the agent, as David points out as a possibility, this is an important feature of the scenario too. And this may certainly affect our intuitions and judgments. But Patch and Slice as described doesn’t make this clear.

    And to respond to David’s other question: what requirement on what agent? It seems very plausible to me that when the optimal collective pattern is not salient, that agents can have duties to establish such salience in addition to duties to perform their part of that collective pattern where this pattern may be a joint action or just a set of individual actions. Felix Pinkert discusses this in his 2014 paper on joint duties. We can say ‘they ought to save the patient’ but perhaps this is just shorthand for this conjunction of individual duties. And they will have such individual duties even if the other agent is non-compliant (see above comment for one reason thinking why). This is why I think it is very important to make clear whether communication is possible/impossible or overly costly. And note that we can retain the bad motives as a feature of the case (in response to David’s concern).

    And Rowan writes: “It seems to me that lots of facts about what I should do take for granted what others will do – I should drive on the right (here in Chicago) because I take it for granted that others will do so.”
    I fully agree that there are cases where it makes sense to rely on such assumptions, but if that’s the kind of case you’re interested in, that’s the kind of case that should be used for making the argument. Patch and Slice as described is not such a case I think. Note that in the Chicago case you are plausibly both morally and epistemically justified in assuming this, I don’t think this is the case in Patch and Slice when communication is possible (and not too costly). For this reason, I don’t think we should extrapolate moral judgments from Patch/Slice to other cases that are disanalogous in significant ways. Our intuitions may be quite different when these features are not present.

    Consider the following case discussed by Olle Blomberg and Bjorn Petersson (2024), which they adapted from Colman et al 2014:
    Burning Building: Three children are trapped in a burning building. One of them is in one room, and the other two are in a second room some distance away. The neighbours Agnetha and Benny see each other approaching the building from opposite sides. Agnetha breaks in and has enough time to do her part of rescuing either the child in the first room or the two children in the second room. The rescue can succeed only if Benny heads straight for the same room with his fire extinguisher. If both go to the first room, they will only rescue the first child. If both go to the second room, they will only rescue the two other children. If each goes to a different room, no child will be rescued. Suppose that Agnetha and Benny can make these choices without any significant risk to their own or each other’s life or health. All this is common knowledge between them, but they do not have any opportunity to communicate with each other—each must choose which room to head for independently of the other.

    We could add the bad motives again as a feature of this Hi-Lo case. We can imagine Agnetha and Benny somehow both know that the other will opt for saving the first child in the first room. (An assumption of the case is that this option is morally worse than saving two children). The point is just that in this case, communication is clearly not possible. And this might matter for our intuitions. To me at least, it seems plausible to me that Agnetha and Benny don’t do anything wrong in this case given the stakes involved and the ‘bad’ motive of the other agent.

    I’m not sure if this captures what David exactly has in mind. But it’s not a prisoner’s dilemma and the same bad motives are in place. And if an essential part of Patch and Slice is the golfing aspect, then we could here replace the single child in the first room with something of less moral value (e.g., a family heirloom or whatever). But this would change the stakes, and therefore what the agents ought to do in my view.

26. Thank you very much for your paper, Rowen! I want to preface my comment by saying that I did not have enough time to fully absorb all the intricacies in your paper. Nor am I well acquainted with the literature on this issue. So, my questions may well reveal my own ignorance, or misunderstanding, or both.

    (1) My first question has to do with your discussion of the Bad Motives View (BMV), which is attributed to Niko Kolodny. I was quite drawn to the view when I first read it in your paper. If I understood it correctly, your objection to BMV is that, in S&P, Patient is wronged not only by the objectionable attitudes of each of S and P, but also by each of their action/omission. And this later kind of wronging is unable to be captured by BMV.

    But is not there a simple extension of BMV that obviates this line of objection? We may note that, in many cases, individuals have claims against not only objectionable attitudes that are held towards them, but also actions that are motivated by such attitudes. For example, while an employer who habours private racist attitudes towards an employee wrongs the employee, an employer who acts on such attitudes in, say, promotion selections commits a further wrong against the employee. Importantly, in this case, the employee does not have an unqualified claim to being promoted by the employer; rather, her claim is that the employer not deny her promotion on the basis of the employer’s objectionable attitudes.

    Similarly, it seems, we may say in the case of S&P, Patient has a claim, directed individually against each of S and P, that they show due concern for her well-being (an attitudinal claim). But Patient also has a further claim, again directed individually against S and P, that they not refrain from performing life-saving surgery based on their lack of due concern (an act claim). Once again, Patient does not have an unqualified claim against S that S slices and against P that P patches. But she does have a claim that S not refrain from slicing on the basis of S’s lack of due concern for Patient’s well-being (mutatis mutandis for P).

    And by violating the individual claim directed against them, each of S and P wrongs the Patient. Notice, however, on this suggestion, though S wrongs Patient by S’s omission, S need not be behaving impermissible all-things-considered (that is, wrongdoing is not necessary for wronging). Granted, some philosophers might be quite allergic to the idea that we can have wronging without wrongdoing (Johann Frick has an extended defense of this view in his paper “Two Faces of Morality”). But if I read Rowen’s reply to David above correctly, then he has already committed himself to this claim, so there shouldn’t be a problem here.

    In this way, we can capture both of the following claims, without resorting to joint oughts:
    (a) S and P each behave permissibly in the prescribed circumstances.
    (b) Patient is wronged by each of S and P.

    (2) I have a concern about the formulation of your own view, regarding the proper explanation of why Patient is wronged. On your view (if I understood it correctly), each of S and P individually wrongs Patient because each of their omission *suffices* to ensure that Patient’s joint claim goes unmet.

    I am worried about the following kind of case:

    S&P500: Patient is sick, and requires a joint surgery from both S and P. There are only two available time slots for the surgery (surgical schedule is otherwise full): 5:00pm and 7:00pm. If S and P both perform the surgery at the same time, Patient will survive unscathed. But if S and P perform their part of the surgery at two different times, then Patient will die an agonizing death. As it happens, S is only prepared to do the surgery at 5:00, regardless of whether P shows up then; P is only prepared to do the surgery at 7:00pm, regardless of whether S shows up then.

    Now, insofar as we have the intuition that Patient is wronged by each of S and P in the original case, I think we should have the same intuition in S&P500. But Rowen’s view seems to have trouble here. This is because S’s omission to show up at 7:00 is not *sufficient* to ensure that Patient’s joint claim is unmet (after all, P could have showed up at 5:00). Similarly, P’s omission to show up at 5:00 is also not *sufficient* to ensure that Patient’s join claim goes unmet (after all, S could have showed up at 7:00).

    My more general worry is that Rowen’s view runs into trouble when we have a disjunctive joint claim on our hands, claim of the form: Either S and P jointly Φ at t1 or they jointly Ψ at t2.

27. I am so sorry that I misspelled your name, Rowan! I typed the comments in haste…

28. Hi Rowan

    I agree that there is something initially intuitive about the idea that the surgeons ought to save patient. But when we examine what that idea involves, it all gets a bit murky, as some of the discussion has suggested.

    Here’s my way of putting the objection. When I think of claims or rights that people have that correspond to duties, I have in mind moral properties of those people with claims or rights that feature in the practical reasoning of those who hold the corresponding duties. But Slice and Patch are not a reasoner together.

    Perhaps, if they could communicate, you could argue that they ought to become a reasoner together. But, in response,

    1) the case seems no different if they can’t communicate; and

    2) to say they ought to communicate together implies that there is a practical reasoner who should generate an entity which can practically reason together, for ‘ought’ is a feature of practical reasoning. But there is no practical reasoner who is able to achieve that if Slice and Patch are each unwilling to reason together and can’t influence each other. So the most we can say is that there are two badly motivated people who, due to their bad motivations, together cause such a reasoner not to exist. But then we are just back with the causal view at an earlier stage.

    So if we go down the line of joint duties etc, we need to believe that there are oughts, rights, duties, etc that float free from the practical reasoning of any entity that gives these deontic properties a home. And that’s just hard to believe.

    The better view, perhaps, is that certain responses that are mainly warranted for wrongdoing, such as resentment etc, might also be warranted in cases where some properties needed for wrongdoing, such as satisfaction of OIC, are absent because the bad motivations of agents ensure that they can’t be satisfied, but the remaining properties are present. In other words, Slice and Patch don’t escape these responses because their bad motivations ensure that some properties needed for wrongdoing are missing.

    We could even call these things ‘wronging’ if we like, though the distinction between wrongdoing and wronging is just philosopher’s artifice of course, and I’m not sure it’s the most perspicuous way to put things.

29. I’m travelling and writing this on my mobile phone, so apologies for the brevity: Thanks for an interesting paper and the great discussion!
    Like David, I’m worried that making Slice and Patch merely severally blameworthy (or resentment-worthy, at any rate) will undermine the point of joint requirements/obligations.
    Rowan, you don’t rule out the view that Slice and Patch jointly warrant resentment, but seem reluctant to go this route (a route I would take). In a footnote, you say: “One obstacle facing a view of this sort would be to show how someone could resent
    several individuals jointly, without thereby resenting any one of them.” I don’t quite feel the intuitive pull of this. I can express anger at several individuals jointly without being angry with any one of them (say, angry with my the players in my home soccer team without being angry at any one player). I also think that Slice and Patch could each feel guilt for their (joint) failure to save patient, even if they were not individually at fault. Admittedly, it seem unfair to resent Slice-and-Patch if only one of had a bad motive. Is that the reason you don’t want to take this route?
    Btw, an alternative that is open here is to say that Slice and Patch are jointly blameworthy but that only those with bad motives are fitting targets of resentment in particular, while others might be fitting targets of other negative reactive attitudes (something like this is Gunnar Björnsson’s view, I think – see his “Being Implicated” paper).

30. Kartik,

    In response to Saba’s case 2, Saba seems to suggest that X is warranted in resenting Y for some outcome o only if o counterfactually depended on what Y did. I doubt that’s right; as you suggest the paper I wrote on causal contributions in Ethics a little while back suggested otherwise. Suppose 10 people together push a boulder off a cliff onto X breaking X’s leg. Y is one of the 10. It would only take 9 to push the boulder off the cliff, so X’s injury does not counterfactually depend on the conduct of any of the 10. If resentment is warranted for anything (I’m actually a sceptic, but leave that aside), X is warranted in resenting Y, and each of the other 10, aren’t they?

31. Hi Victor,

    I’m traveling in Slovakia at the moment and pecking this into my phone, so please excuse the brevity! Putting aside that your boulder case involves potentially inculpatory cooperation, I don’t want to suggest that there’s *no* basis for resentment in cases of overdetermination such as that one. Intuitively, resentment is warranted. But it’s not obvious what justifies this intuition. We need an account one way or another. Case 1, though, is much easier, because there’s straightforward counterfactual dependence there. It seems then that the basis for resentment in Case 1 is going to be different from and, I think, stronger than, the basis of resentment in Case 2, whereas Rowan seems to suggest that they are the same in both respects. Maybe though it doesn’t matter. What matters is that there is a basis for resentment in each case one way or another; they needn’t be the same; nor does resentment need to be just as strong. Rowan can just pass the buck to accounts of resentment in the context of over determination.

32. The discussion so far is great (as of course is the paper), but getting a bit sophisticated for me to jump into. So, I’m going to contribute something which I think hasn’t been offered yet: a defence of the group agent solution.

    Rowan’s worry is that Slice and Patch might not contribute a group agent, and even if they do, there are other similar cases in which there is no group agent. (David extends this generally in his precis.) The full paper makes this case for the Pettit/List account of group agents, and briefly mentions that we could use more permissive accounts to get the result that Slice and Patch form a group agent.

    The more permissive group agency account I have in mind holds that if a set of agents could become a Pettit/List-style (or maybe Collins-style?) corporate agent, then it counts as a group agent. In Rowan’s example in the paper, Slice and Patch have common knowledge and can communicate, so could come to form a decision-making entity, so count as a group agent. Even more plausibly, I’d say the hospital where they work is a group agent (probably even on the Pettit/List view) and this agent has failed in its duty to Patient – if Slice and Patch can’t communicate, the hospital’s failure to facilitate communication is part of its failing. In David’s case, the people in the park ‘could have acted together’ which suggests that they could have communicated and made a decision together, hence they meet my criterion.

    I acknowledge this criterion is not currently well worked out (what do I mean by ‘could’?), but, as Rowan and David acknowledge, it gives us a neat solution in such cases. There are two objections from both of them:

    (1) Rowan says “this response seems a little ad hoc. What reason do we have to believe in such a permissive theory of group agency, beyond the fact that it helps the group-agent solution to avoid the above objection?”

    It strikes me that the same could be said of Rowan’s own proposal. As this discussion has shown, the joint ought view requires revision to certain other moral principles, like that that all obligations are action-guiding, and that wronging requires wrongdoing. So any solution to the Slice and Patch case looks like it will revise some intuitions to solve the case. At least group agents seem to figure more in ordinary language than do joint oughts. I think of this as the group agent view having a more expansive ontology, and the joint ought view a more revisionary ideology (in Quine’s sense).

    (2) David suggests that there will always be some cases in which there is no plausible group agent and yet you get Slice and Patch-esque failures of collective action. This is true, at least unless we go with Jackson’s ultra-permissive view, which David elsewhere calls ‘easy agency’. But, I’m not sure how much of a cost this is. In the cases where my criterion for group agency is not met – i.e. no set of agents could form a decision-making group that could deliver the good outcome, I don’t think it’s so intuitive that there is a wrong. Of course, we can still say that the outcome is bad, etc. But, very sadly, there are some bad things such that there is no agent who can do anything about them – as I think we all recognise.

33. Hi Libby, thanks for your questions. To your first question, I think there’s such a thing a joint ability – what we’re together able to do. This is what makes the joint ought stick to the surgeons.

    To your second question, I think the case you describe is complicated. To see this, imagine a similar case where one swimmer alone can save you, but more than one would be worst (maybe it would be dangerous). Here, I think each is conditionally obligated to save you, on the condition that others don’t. Back to your case. I suspect that what we have here is a complex structure of many, conditional, joint oughts. Each pair is conditionally, jointly obligated to save you, on the condition that no other pair does.

34. Hi Niels and David – thanks for these follow up comments. To the first point Niels raises, let’s suppose that you’re right that the high stakes in this case obligate each surgeon to check the other’s intentions. Suppose they’ve done this, and they’re each certain that the other is set on golfing. What now? It looks like each ought not to do their part. But, to me, it still seems like Patient stands to suffer a serious moral wrong.

35. (I’m struggling a little to keep up with the comments, so apologies if I miss anyone. Please feel free to email me if you want to speak more after comments close later today. I’m enjoying thinking through everyone’s questions!)

36. *thumbing this into my phone, so apologies for any typos*

    Hi Erik, two great questions, thanks! On your first question, I think there’s some cost to saying that the surgeons each wrong Patient by acting on their bad motives (while acting as they ought). I agree that in cases of discriminatory hiring, what makes the act wrong is acting on a particular motive (e.g. rejecting a candidate because they’re a woman, rather than because of their experience). But I tend to think this is a special case (Sophia Moreau’s ‘deliberative freedom’ account of the wrong of discrimination, for example, implies that the wrong depends on motives. But that doesn’t mean that we can say in general that when someone acts on a bad motive, they commit a further wrong, beyond just having that motive). Of course, there will be theoretical costs with my view too. It’s just a matter of which costs we’re most willing to accept.

    On your second question, this case is really interesting. I think that whether we see the surgeons’ intentions as sufficient for Patient’s death depends on what we treat as fixed (going back to Maggie’s previous points). If we treat it as fixed that Patch will only do his part at 7, then Slice’s being set on slicing at 5 looks sufficient for Patient’s death. But if we treat what Patch will do as open, then Slice’s slicing at 5 doesn’t look sufficient for this – Patch could patch at 5. What I’ve been trying to do is to describe the case such that, when we assess what either surgeon ought to do, the appropriate thing to do is to hold fixed the other’s behaviour. (I think there are plenty of cases where this is the right way to deliberate about what to do. Hence the opening sentence of my paper “Sometimes what you ought to do depends on what other will do”).

37. Niels, briefly on your post about communication tagged “7:42am”: A thought about non-negligible chances, and one about (that along with) the Burning Building.

    Before turning to the building example: if communication between the doctors is possible, you suggest that each has an individual duty to reach out to the other doctor just in case they might resolve to do their part after all. I see two problems, which are cumulative. First, since you’re rightly ruling out absolute certainty, let the chance that this would succeed be extremely small, almost negligible, which is perfectly realistic. Maybe you want to say that not reaching out is still blameworthy, but only slightly. But, both because this would be about the failure to reach out, and also because the blameworthiness would be so slight, it then doesn’t seem to me to answer to the temptation to indignation about Slice and Patch that gives us the puzzle.

    Second, is it even slightly required for Slice to reach out in case maybe Patch would cooperate after all, given that she will not herself cooperate in any case? Or should we add that there’s at least a non-zero chance (which she should realize) that she will? But that tiny chance compounds the tiny chance of Patch changing his mind. If not reaching out is still blameworthy, it would be but only very, very (“very” twice this time) slightly. It (again and more-so) doesn’t seem to provide an explanation of the puzzles tendency to tempt indignation.

    As for the Burning Building case: As you grant at the end, the case is only analogous if there is no person to save in the first room but, perhaps, a keg of beer that either neighbor alone could to rescue and tap. (Even your “heirloom” might include some distracting hint of noble motives, so we should wash that out if we want it to be like the disgusting doctors.) And, for the same reason, we should keep it clear and salient that (as you want us to assume) neither neighbor takes any (distractingly noble) personal risk at all, so it’s probably best if it’s not a burning building. Perhaps instead, then: Room 1: keg. Room 2: child locked in as the sealed room runs out of oxygen. We add that each neighbor will almost certainly head to Room 1 for the keg whatever the other neighbor will do. Now that it’s analogous to Slice/Patch, it doesn’t seem to me here, either, that our sense of indignation on behalf of the child goes missing, and would be present only if they could communicate.

38. Hi Olle, thanks for your question. This doesn’t really come across in the paper, but I’m totally open to the idea of joint blame/blameworthiness. In the footnote you mention, I try to very briefly outline an obstacle for that approach. But I don’t think it’s insurmountable. And I like your football example. (I think Thomas Smith has a similar one – an actor can fear their audience, without fearing any one of them).

    You write “Admittedly, it seem unfair to resent Slice-and-Patch if only one of had a bad motive. Is that the reason you don’t want to take this route?” Yes, that’s what I was trying to explore. But maybe there’s a way to hold onto this, while be open to the idea of joint blame.

39. To Victor’s second comment, you write: “When I think of claims or rights that people have that correspond to duties, I have in mind moral properties of those people with claims or rights that feature in the practical reasoning of those who hold the corresponding duties. But Slice and Patch are not a reasoner together.” Let me foreshadow something I’m working on at the moment. I agree that duties and claims are things which feature in practical reasoning. So, whose reasoning do joint oughts feature in? Not that of a group agent (for reasons you’ve explained), and not that of any one natural person, since joint oughts don’t require anything of any one individual.

    The view I’m developing in a paper I’m working on right now is that agents engage with joint oughts by participating in joint reasoning. This is a form of reasoning which is necessarily done together by several agents, and it typically aims to settle the question of what ‘we’ are to do (or believe – I think there’s doxastic joint reasoning too). My view is that several agents jointly ought to do something only if they can reason jointly about whether to do it. Slice and Patch, I think, *can* reason jointly about whether to operate – they together have the ability to reason with each other, it’s just that neither wants to. (For those who are interested, there are some really interesting recent papers about joint reasoning – I really like the stuff co-authored by Berislav Marusic and Stephen White, as well as Johann Frick’s paper ‘What We Owe to Hypocrites’, and Brendan de Kenessey’s work on promising).

    Of course, this brings back Niels’s question about communication – does it change things if Slice and Patch can’t communicate? In light of the position I’ve just described, I’d have to say yes, it does. If Slice and Patch can’t communicate, then it’s hard to see how they could engage in reasoning together, and if they can’t reason together then they can’t have a joint ought. But I think this gets things right. I totally in agreement with Niels that it seems like there wouldn’t be a wrong in a no-communication case. (Probably should have clarified this earlier).

40. Hi Nikhil, a quick reply to your question. I accept that I probably haven’t done enough to sway those attracted to a more permissive theory of group agency. But I do think it’s worthwhile to try to articulate a view of joint obligation which isn’t committed to such a view, and I’ve tried to do that. In the end, whether you’re more attracted to a view like mine or a permissive theory of group agency will come down to which theoretic costs you’re more willing to accept.

41. Before comments close, I’d like to say thank you to the PEA Soup team for organising this discussion, and to all the participants for your great questions and comments. I’ve enjoyed it, and learnt a lot. I realise that I haven’t responded to everything – please feel free to reach out over email if you’d like to continue the conversation.

    Some of you may be attending the Social Ontology Conference at Duke next week. If so, I’ll see you there!

42. Thank you for the response, Rowan! On your reply to my first point, I agree that BMV should be not be advanced as a general theory. But it seems to me that S&P is indeed a *special case* (just like the discrimination case is special case), as evidenced by Kolodny Good- vs. Bad-motive versions of the case. I also agree that, at some point, the debate becomes one of measuring the price. But I suppose my thought was that BMV captures all the correct verdicts in S&P without resorting to any new and contentious theoretical apparatus, which is a huge point in its favor. The only price I see is the admission that we can have cases of wronging without wrongdoing (personally, I am very hesitant to sign on to the idea) , but this price is equally borne by your view. So, BMV seems to come out ahead.

    On your reply to my second point, I’d like a little more clarification. Granted, it may well be that “when we assess what either surgeon *individually* ought to do, the appropriate thing to do is to hold fixed the other’s behaviour.” But is the same true when we assess what both surgeons joinly ought to do?

    I have in mind a case of the following structure (a modified version of my S&P500 above):
    Best option: S&P jointly operate at 5.
    Worse option: S&P jointly operate at 7 (Patient will survive, but left arm will be permanently paralyzed).
    Worst option: S&P operate at different times (Patient will die).

    Suppose as it happens, P will show up at 7 regardless of whether S will show up then. Do we say, in ascertaining the joint ought, that since it is fixed that P will show up at 7, what S and P jointly ought to do is to operate at 7? Or do we say, S and P jointly ought to operate at 5, and if S shows up at 7, S’s conduct has ensured that Patient’s joint claim goes unmet?

43. Sorry Rowan, didn’t see your previous comment about the discussion being closed. No need to reply; I will follow up with you through email! Thank you once again for your great paper; I learned a lot from it!

Comments are closed.