A good cause

Abstract

I explore the question of how to ground the responsibility of agents in some tricky cases involving multiple agents who act in a non-coordinated fashion. These are scenarios where no single agent has the individual ability to make a difference to a harmful outcome, but where the outcome would have been avoided if they had all acted as they should have (thus, the agents collectively made a difference to the outcome’s occurrence). I argue that an important source of the problem is that it’s hard to motivate a concept of cause that can be behind the agents’ responsibility in these cases. I illustrate the problem with a particular example: Yablo’s proportionality criterion on causation. I then sketch a possible solution.

1 Introduction

Causation grounds responsibility. That is to say: our moral responsibility for what happens depends, in some significant way, on our causal contributions to what happens. Still, the project of explaining how it is, exactly, that causation grounds responsibility faces important obstacles. In this paper I discuss one such obstacle arising from an intriguing set of cases: scenarios involving non-coordinated agents who make a collective difference to an outcome without making an individual difference.

As I intend to show here, these cases give rise to interesting puzzles because, although the agents involved in them seem to be clearly responsible for the ensuing outcomes, it’s hard to see how their responsibility could be grounded in their causal contributions. In particular, given that those agents are not acting in a coordinated fashion (in the examples I’ll offer, they don’t even know about each other), more familiar problems of collective responsibility or intentionality don’t even come up.¹ Instead, the puzzles are more basic puzzles about the nature of causation itself. Reflecting on these cases is important, then, to improve upon our understanding of the concept of cause that grounds responsibility. By this I have in mind a natural concept of cause, one that captures objective features of the world, and thus one in which our concept of responsibility can be objectively anchored. What is involved in that concept of cause, and how does it ground responsibility?²

A caveat: some of the cases discussed in this paper involve omissions by agents. In the causation literature, there is a debate about whether omissions, and absences in general, have causal powers. Some believe that they do (as in the absence of rain caused the drought, or the failure to administer a medication caused a patient’s death), but others believe that they don’t, on the grounds that only positive events have causal powers and omissions are not positive events.³ Here I’d like to remain neutral on this debate. For simplicity’s sake, I start by assuming that omissions have causal powers. However, as we will see, the puzzles arise even if omissions don’t have any causal powers. For, if they don’t have causal powers, they can still contribute to the explanation of outcomes in other kinds of ways, and this is arguably enough to ground the moral responsibility of agents.⁴ This feature of omissions is all that’s needed for the puzzles to arise.

Note that the type of responsibility that is our focus here is moral responsibility for outcomes—such as events or states of affairs that result from our actions or choices. This of course assumes that we can be responsible, not just for what we do, but also for the consequences of what we do (an assumption that aligns with commonsense and with most non-skeptical views on responsibility). In the problem cases we’ll discuss, agents seem to be responsible for the ensuing outcomes, not just for their actions or choices. However, it’s hard to see how that responsibility could be grounded in a causal contribution, and this is what gives rise to the puzzles.

The main aims of this paper are to, first, explain how the puzzles arise and, second, to provide a solution. My proposed solution consists in a combination of two strategies: one is an argument for the possibility of disjunctive causes/explainers (here I build on the discussion in Sartorio, 2006), and the other is a revamping of the relation between causation/explanation and proportionality considerations (considerations of the kind proposed by, e.g., Yablo, 1992a). As we will see, both strategies are important, as they must work in tandem to result in a successful solution to the puzzles.

The structure of the paper is the following. In Sect. 2 I introduce the problem cases, and in Sects. 3 and 4 I explain how they give rise to the puzzles. In Sect. 5 I illustrate the challenges one faces when trying to find a solution by focusing on, specifically, Yablo’s proportionality criterion. Then, in Sect. 6 I argue for my own solution—again, one that combines disjunctive causes/explainers with a certain understanding of the role played by proportionality considerations. Finally, I discuss the main takeaways in Sect. 7.

2 The problem cases

In this section I present two problem cases. Again, these are cases involving non-coordinated agents who make a collective difference to an outcome without making an individual difference. And they are “problem” cases in that, although the agents involved in those scenarios seem to be clearly responsible for the outcome, it’s hard to motivate a concept of cause that can ground their responsibility. One of the cases involves only positive actions, and the other one involves an omission. Although the examples are otherwise structurally similar, the difference between action and omission is significant enough that it’s important to see how the puzzles arise in each case (as we will see, omissions are special in ways that will be relevant to our discussion later).

I’ll build up to the problem cases from simpler scenarios involving just one agent. Consider, first, the following scenario:

Switcher

A runaway trolley is bearing down a track where a person (Victim) is trapped, up ahead. There is a switch and a side track. A person standing by the switch (Switcher) flips the switch and, as a result, the train turns onto the side track. However, the tracks reconverge after a while, before the location where Victim is trapped, and thus Victim still dies (in pretty much the same way, at around the same time).

In this case Switcher doesn’t seem to be responsible for Victim’s death. Intuitively, this is because he didn’t cause the death by flipping the switch: he slightly altered the causal route to the death, but he didn’t cause the death. In recent years, scenarios of this kind (switching scenarios, or just “switches”) have been discussed quite extensively by causation theorists. Many agree with this causal judgment and try to capture it with their accounts of causation.⁵

Given that causation is a natural relation, Switcher’s intentions (what he was trying to do by flipping the switch, or what motivated him to do that) are irrelevant to the causal powers of his behavior. This means that Switcher wouldn’t have caused Victim’s death even if he had acted with a bad intention. Imagine, for example, that he falsely believed that part of the main track was disconnected or broken, and the reason he flipped the switch is that he thought that turning the trolley onto the side track was the only way to guarantee Victim’s death (otherwise the trolley would have derailed before reaching Victim, after traveling on the broken track). In that case Switcher still wouldn’t have been responsible for the death because he wouldn’t have caused it. He would have been responsible for trying to cause the death, but not for the death itself. In what follows, I’ll focus on versions of the case where Switcher acts with a bad intention of this kind.

Now add a second agent who acts independently. Consider, for example, the following scenario (based on a case presented in Sartorio, 2006):

Switcher and Reconnecter

Switcher flips the switch, knowing that part of the main track was broken, and thinking that turning the trolley onto the side track was the only way to guarantee Victim’s death. At the same time, another agent (Reconnecter), who also wants Victim to die and who is unaware of what Switcher is doing, reconnects the part of the main track that was broken.

As we will see, this case raises special problems.⁶

Before we get to this, let me introduce the second problem case. Again, start with a single-agent case:

Lazy While walking along the beach, a perfectly capable swimmer (Lazy) sees a child starting to drown and crying for help, but he’s too lazy to attempt a rescue. The child drowns. Unbeknownst to Lazy, some sharks were swimming in the area and they would have attacked Lazy and thwarted the rescue attempt if he had decided to jump in. (Based on a case presented in Fischer and Ravizza 1998: 125.)

As many responsibility theorists have pointed out, including Fischer and Ravizza themselves, in this type of case the agent (Lazy) doesn’t seem to be responsible for the child’s death. Intuitively, this is because, unbeknownst to Lazy, the presence of the sharks made it the case that Lazy’s omission was causally irrelevant to the child’s death (and this is so even under the assumption that omissions in general have causal powers).⁷

Now add a second agent who acts independently, as in the following scenario (based on a case presented in Sartorio, 2017):

Lazy and Sharky

The same as in Lazy, except that, just as Lazy was deciding not to attempt a rescue, another agent (Sharky) was independently releasing the sharks in the water, hoping that this would prevent any drowning victims from being rescued. Lazy and Sharky are unaware of each other’s existence and intentions.

Again, this case will raise special problems.

These are, then, the two problem cases that will be the focus of our attention: Switcher and Reconnecter and Lazy and Sharky. Switcher and Reconnecter involves two positive actions but Lazy and Sharky involves an omission: Lazy’s failure to attempt to a rescue. Note that, in both scenarios, and unlike what happens in the corresponding single-agent versions, where the agent seems to be off the hook for the outcome, somebody—either one of the agents or both—is clearly to blame for what happened. For the outcome is the result of two agents acting independently in clearly blameworthy ways. If both agents had acted as any morally decent agent would have acted in the circumstances, the outcome would have been prevented. In Switcher and Reconnecter, if Switcher hadn’t flipped the switch and Reconnected hadn’t reconnected the main track, the trolley would have derailed while traveling on the main track, and thus Victim would have survived. And, in Lazy and Sharky, if Lazy had jumped in to attempt a rescue and Sharky hadn’t released the sharks, the child would have survived. The outcome isn’t the result of an “act of God” or some unlucky coincidence; clearly, somebody is to blame for it.⁸

On the other hand, as we will see, it’s hard to explain how these agents can be responsible for the outcomes, for it’s hard to make sense of how they could have made a causal contribution to those outcomes. I discuss this in the following sections. The next two sections draw on the arguments presented in Sartorio, 2006: section I. My discussion summarizes the main arguments there and builds on that discussion until we reach the special challenges presented in Sect. 5 and my final solution to the puzzles.

3 No individual causation

In this section I explain why I think that the agents in our problem cases don’t make an individual causal contribution to the outcome. The argument rests on a parallel with the corresponding single-agent cases.

Start with Switcher and Reconnecter. Here is the argument that Switcher doesn’t cause Victim’s death in that case:

• 1. Switcher doesn’t cause Victim’s death in Switcher.

• 2. If Switcher doesn’t cause Victim’s death in Switcher, then Switcher also doesn’t cause the death in Switcher and Reconnecter.

• 3. Therefore, Switcher doesn’t cause Victim’s death in Switcher and Reconnecter.

The justification for Premise 1 is something I mentioned in Sect. 2: Switcher isn’t responsible for Victim’s death in Switcher (the single-agent case), even if he wanted to cause it. Intuitively, this is because he didn’t cause Victim’s death—even if, again, he wanted to cause it. And the justification for Premise 2 is this. Given that Reconnecter reconnected the main track in Switcher and Reconnecter, the two cases (Switcher and Switcher and Reconnecter) are arguably on a par with respect to Switcher’s own contribution. For it seems irrelevant that the track used to be disconnected, if it is or will be reconnected at the relevant time—when the trolley was supposed to travel on it.

In turn, the argument that Reconnecter didn’t cause Victim’s death is this. Given that Switcher flipped the switch, the trolley never travelled on that part of the main track (the segment that used to be disconnected and that Reconnecter reconnected). But, if the trolley never travelled on that segment of track, Reconnecter’s act was causally irrelevant to the harm inflicted by the trolley. Therefore, Reconnecter’s act didn’t cause Victim’s death either.

In other words, Switcher and Reconnecter don’t make an individual contribution because their contributions cancel each other out. Each would have made an individual contribution in the absence of the other: if Reconnecter hadn’t reconnected the main track, Switcher’s flipping the switch would have causally contributed to the death; conversely, if Switcher hadn’t flipped the switch, Reconnecter’s reconnecting the main track would have causally contributed to the death. But they don’t make an individual contribution when they act together; when they act together, their causal powers cancel each other out.

The argument for Lazy and Sharky is similar. First, the argument that Lazy doesn’t make an individual causal contribution is this:

• 1. Lazy doesn’t cause the child’s death in Lazy.

• 2. If Lazy doesn’t cause the child’s death in Lazy, then Lazy also doesn’t cause the child’s death in Lazy and Sharky.

• 3. Therefore, Lazy doesn’t cause the child’s death in Lazy and Sharky.

Here too, Premise 1 is motivated by some comments from Sect. 2: in Lazy (the single-agent case) Lazy isn’t responsible for the child’s death. Intuitively, this is because he didn’t contribute to it, although he thought he did. And the argument for Premise 2 is, again, that the two cases seem to be on a par with respect to Lazy’s causal contribution. For it cannot plausibly matter to whether Lazy contributed to the child’s death if the sharks were in the water naturally, or if Sharky put them there.

In turn, the argument that Sharky didn’t cause the child’s death is this. Given that Lazy never attempted a rescue, the sharks never did anything to thwart a potential rescue. But, if the sharks never did anything, then Sharky’s having released them in the water was causally irrelevant to the child’s death. Therefore, Sharky’s act didn’t cause the child’s death either.

Thus, Lazy and Sharky is another example of causal powers that cancel each other out when the agents act simultaneously. Each would have made an individual contribution in the absence of the other: if Lazy had made a rescue attempt, Sharky’s releasing the sharks would have contributed to the child’s death; conversely, if Sharky hadn’t released the sharks, Lazy’s failing to make a rescue attempt would have contributed to the child’s death. But they don’t make an individual contribution when they act together: when they act together, their causal powers cancel each other out.

Although the agents don’t make an individual contribution in our problem cases, they could still have made a collective contribution. And I think it’s clear that they did. Otherwise, our agents would be off the hook for the outcome, which they are obviously not. But, the problem is, it’s not at all clear how we can say this. I turn to this in the next section.

4 Collective contributions

First of all, let me explain what kind of collective causal contribution those agents must have made. Recall that by “collective” contribution I don’t have in mind anything that involves collective action or collective intentionality, but simply a type of contribution that combines the contributions made by two individuals who are acting independently from each other.

On the face of it, the collective contribution that our agents make cannot be a “conjunctive” kind of contribution. For example, we cannot say that a conjunctive fact such as the fact that Switcher flipped the switch and Reconnected reconnected the main track caused Victim’s death, without either of the individual facts making a causal contribution. For, how could the conjunctive fact be causally relevant if none of its parts (the individual conjuncts) made a causal contribution? The contribution of the whole must be grounded, at least partly, in the contribution made by the parts (at least, it must be grounded in the contribution made by some of the parts). The same goes for other understandings of a conjunctive contribution, such as an interpretation in terms of mereological sums of events: arguably, the sum that is the mereological sum of different events cannot be causally efficacious if none of the parts (the individual events) made a causal contribution.⁹

But that leaves a “disjunctive” type of contribution, such as the contribution of a disjunctive fact, as the most obvious candidate. In contrast with conjunctive facts, I think we can, at least in principle, understand how a disjunctive fact could make a contribution that none of the disjuncts make. For the disjuncts are not parts of the disjunctive fact in the same way the conjuncts are part of a conjunctive fact: a disjunctive fact cannot be understood as the sum of its disjuncts; correspondingly, it has a weaker essence, not a stronger essence, than its disjuncts. As a result, it’s possible to conceive of scenarios where a disjunctive fact makes a contribution without any of the individual disjuncts making one.

To illustrate: imagine, for the sake of the argument, that being red is a matter of either being scarlet or crimson or … (some other shade of red), and that you were really hoping to get a red shirt for your birthday. Someone gifts you a scarlet shirt and this makes you happy. It might be argued that your happiness from getting the gift is the result of your getting a red shirt—a shirt that is either scarlet or crimson or… (some other shade of red), not of your getting a scarlet shirt specifically. In contrast, if you’re happy after getting a shirt that is both red and thick, arguably, your happiness must be the result of the shirt being red, or of the shirt being thick, or of the shirt having both properties (a conjunctive contribution must be grounded in a contribution made by the conjuncts).

Returning to Switcher and Reconnecter, then: if we could motivate the claim that Victim died because either Switcher flipped the switch or Reconnecter reconnected the main track, that wouldn’t be obviously in tension with the causal inefficacy of the disjuncts. But, how could one argue that Switcher and Reconnecter in fact made such a disjunctive collective contribution?

The easiest way is to note that Victim wouldn’t have died in the absence of that disjunctive fact. For, if that disjunctive fact hadn’t obtained (which is to say: if Switcher hadn’t flipped the switched and if Reconnected hadn’t reconnected the main track), then Victim’s death wouldn’t have occurred. In other words, Victim’s death counterfactually depends on the disjunctive fact. Given that counterfactual dependence is standardly considered to be a sufficient condition for causation, this is a pretty strong reason to believe that (if disjunctive facts are eligible candidates for being causes) the disjunctive fact is causally relevant to the death. In contrast, notice that, arguably, Victim’s death doesn’t counterfactually depend on the conjunctive fact. For, if the conjunctive fact hadn’t obtained, then, given that the agents acted independently from each other, one of the conjuncts might still have obtained, and thus Victim’s death might still have occurred.¹⁰

Similarly for the other case, Lazy and Sharky. Here the relevant disjunctive fact is the fact that either Lazy didn’t attempt a rescue or Sharky released the sharks. The child’s death counterfactually depends on that disjunctive fact (whereas it doesn’t clearly depend on the conjunctive fact). For, had that disjunctive fact not obtained (had Lazy attempted a rescue and had Sharky not released the sharks), the child wouldn’t have died. Motivated by these considerations, one could argue that the disjunctive fact is part of the causal explanation of the child’s death.

Now, all of this is assuming that it’s legitimate to think that disjunctive facts are ever causally relevant to the occurrence of outcomes. Some would balk at the suggestion that something as metaphysically unnatural as a disjunctive fact can be a cause (see, e.g., Armstrong, 1978: 19–20; Lewis, 1986b: section VIII; Yablo, 1992a, 2003). But recall what we said about omissions: if omissions are not causally relevant, at least they are explanatorily relevant in some other kind of way. Arguably, the same could be said about disjunctive facts, if they were not eligible candidates for being causes: disjunctive facts could still be explanatorily relevant in some other kind of way (and that is how we should understand the relevant collective contributions).

Relatedly (and of noteworthy importance, given what will come next), another thing one could do to motivate the relevance of the disjunctive fact in our problem cases is to note that the disjunctive fact is more proportional to the outcome than the individual facts themselves (in the sense of proportionality discussed by Yablo, 1992a or List and Menzies, 2009). For the agents in our problem cases didn’t both have to behave in the way they did in order for the outcomes to occur: it was enough that one of them behaved in the way they did. As a result, appealing to the individual behaviors might seem like an overshoot, and one might then be led to think that a weaker and more proportional fact (the disjunctive fact) does the work.¹¹

We’ll revisit proportionality in the next section. For now, imagine that disjunctive facts are the way to go. How could one ground the responsibility of the agents on that basis? One could do this by revising (if only slightly) the standard way of grounding responsibility in causation (see, for example, Sartorio, 2022). The standard way consists in claiming that responsibility for an outcome requires having caused it, in the sense of making an individual contribution. I have explained why I think that the agents in our problem cases don’t meet that condition. In contrast, the revision would consist in claiming that responsibility for an outcome requires being responsible for one of the outcome’s causes. The thought is that the agents in our problem cases can meet this condition without meeting the standard condition. For one of the outcome’s causes in our cases is the disjunctive fact, and the agents (arguably, both of them) are presumably responsible for that fact. For one thing, notice that they each behaved in a way that guaranteed it.¹²

Still, there are some important obstacles that remain. I turn to this in the next section.

5 Proportionality and causation

As we have seen, appealing to disjunctive causes or explainers seems to be a promising way to solve our puzzles. Plus, this move can be plausibly motivated by appeal to, for example, proportionality considerations. However, a potentially serious problem arises at this point. The problem is that this strategy seems to overshoot, for the most proportional facts or events aren’t always, or even often, the causally or explanatorily relevant ones. Moreover, the reason they aren’t causally or explanatorily relevant seems to be, precisely, that they are too “disjunctive” in nature. This results in an important tension at the core of the solution I have proposed.

To see how the problem arises, consider, first, the following scenario:

Suzy and Billy

Suzy and Billy throw rocks at a window. Suzy’s rock is a little faster and makes the window shatter before Billy’s (Billy’s rock sails through empty space).

What caused the window to break in this case? Clearly, it was Suzy’s throwing her rock. Correspondingly, Suzy is morally responsible for the window shattering and Billy is not (Billy may still be responsible for trying to make the window shatter, but not for the shattering itself). But notice that, if disjunctive facts are eligible causes, then Suzy’s throwing her rock isn’t the most proportional candidate cause. For consider, instead, the fact that either Suzy or Billy threw a rock at the window. This fact is more proportional. For the window wouldn’t have shattered in its absence, whereas the window would still have shattered if only Suzy hadn’t thrown her rock.

The problem, or some version of the problem, generalizes to more ordinary cases too. Imagine now that only Suzy throws a rock at the window and the window shatters. Again, Suzy’s throwing her rock caused the shattering. However, here too, there is a fact that seems to be even more proportional to the shattering than this fact about Suzy, namely, the fact that somebody threw a rock at the window (with a certain momentum, etc.). And, again, despite this, the more proportional fact is not what’s causally relevant to the window shattering. It’s certainly not what grounds Suzy’s moral responsibility for the shattering: what grounds her responsibility, it seems, is the fact that she made the window shatter by throwing her rock at it.

In his influential work on causation and proportionality, Yablo has drawn attention to this kind of problem (Yablo, 1992a, b, 2003).¹³ Let me briefly explain his reasoning, and his suggested solution; we’ll then return to the implications for our discussion here.

On the one hand, Yablo argues for the idea that causes are proportional or commensurate with their effects. Roughly, this is the thought that causes have enough detail built into their essential properties to make the effects happen (they are specific enough), but without having too much detail. For example, the proportionality criterion would rule out Suzy’s throwing a rock while wearing a red hat as a cause of the window shattering. For that event is too specific and is screened off by a less specific one: Suzy’s throwing a rock.

On the other hand, however, Yablo argues that proportionality shouldn’t be pursued at any cost, lest we end up with highly unnatural causes and an unintegrated causal order. The way to avoid this, Yablo thinks, is to restrict proportionality by appeal to the right choice of causal ontology (the events that are in principle eligible as causes and effects). In particular, according to Yablo, the right choice of causal ontology must avoid events that are too unnatural, too disjunctive, and too “dedicated” to their effects (the kinds of events that result in an unintegrated causal order).

To illustrate with the Suzy and Billy case again, consider the following series of potential causes of the window shattering:

C1 = Suzy threw a rock at the window.

C2 = Somebody (Suzy or Billy) threw a rock at the window.

C3 = A suitably large and heavy object was propelled at an adequate velocity towards the window in the presence of appropriate gravitational forces.

Yablo argues that events like C3 shouldn’t be included as part of the causal ontology (Yablo, 1992a). They are highly disjunctive in a way that, if allowed as eligible causes, they would only end up causing their “dedicated” effects. For what these events do is, basically, capture the different ways in which the effects in question could potentially be caused, which results in a highly disjunctive condition.

Presumably, Yablo would also rule out C2 as an eligible cause, on similar grounds. For, even if C3 is more disjunctive and more dedicated than C2, C2 is already quite disjunctive and quite dedicated. Thus, Yablo’s suggestion is that the right ontology, for the purposes of causal theorizing, is one that strikes the best overall compromise between proportionality, on the one hand, and naturalness and the integrity of the causal order, on the other (Yablo, 1992a, 2003). In turn, naturalness and the integrity of the causal order are preserved by not letting in events that are too disjunctive and too dedicated to their effects. The upshot is that proportionality has its limits. In our example, we need to stop at C1. C1 is an ordinary, natural, non-disjunctive event.

Importantly, note that, as an added plus, this move has the right kinds of implications about moral responsibility. Again, in Suzy and Billy, Suzy is clearly to blame for the shattering and Billy isn’t. Intuitively, this is because she caused the shattering and he didn’t (even if he also wanted to cause it, and even if he would have caused it if Suzy hadn’t beaten him to it). But that means that we need to stop at C1 (and not C2 or C3) in order to ground Suzy’s responsibility in causation in the most natural way.

However, this solution only works thanks to a blanket ban on disjunctive causes. And such a blanket ban is problematic.

For one thing, it would appear to rule out causation by omission altogether. For omissions are highly disjunctive, in a similar kind of way. They are also quite dedicated to their effects, for similar reasons. For example, when Lazy fails to attempt the child’s rescue in Lazy (the simple case without the sharks) and the child dies, the child dies partly because of Lazy’s omission. But Lazy’s failure to attempt the rescue is necessarily equivalent to a disjunction of all the different things that Lazy could have been doing instead of rescuing the child (he could have been eating an ice cream on the shore, or reading a book, or dancing, etc.). Lazy could have been doing anything but attempting the rescue, and he would have still caused the child’s death by omission. Thus, if omissions are highly disjunctive causes, this is in tension with the restriction on proportionality.

Of course, one could try and address this problem by also banning causation by omission. If only positive events are causes, proportionality could be restricted by appeal to a causal ontology of positive non-disjunctive events. But this would just push the problem back one step. For the problem would reappear, under a different guise, at the level of the explanatory powers of omissions. This is because, as noted above, even if omissions didn’t have any causal powers, they would still be part of explanations. And, if proportionality applies to causation, we should probably expect it to apply to explanations too—that is to say, we should probably think that, other things being equal, a more proportional explanation is a better explanation. For example, compare Lazy’s failing to attempt a rescue (in the Lazy case) with Lazy’s failing to attempt a rescue while wearing red clothes. Isn’t the former a better explanation of the child’s death than the latter?¹⁴ In sum: the basic problem persists even assuming that omissions don’t have any causal powers. To the extent that omissions can help explain things, the problem is here to stay.

We have seen that a blanket ban on disjunctive causes, or disjunctive explanations, is in tension with the assumption that agents can be responsible for outcomes by omission. And then, of course, to return to our main discussion, there are our problem cases. In those cases, as we have seen, if the agents make a contribution to an outcome (in particular, one that can ground their responsibility for the outcome), it is arguably a collective type of contribution that is best captured by a disjunctive cause or explanation. Again, a blanket ban on disjunctive causes or explanations isn’t compatible with this.

What to do, then? I turn to my proposal in the next section.

6 Putting proportionality in its place

I think the key to solving the problem lies in realizing that proportionality plays a more secondary role than it may have initially seemed. We shouldn’t always look for more proportional causes or explanations. Rather, proportionality should only be pursued when needed (as in our problem cases.)

Start by reflecting about how proportionality was motivated. I mentioned the example of Suzy making the window shatter by throwing a rock while wearing a red hat. In this case, we correctly think that her wearing the red hat isn’t part of the causal history or explanation of the window shattering, and we appeal to proportionality considerations to capture this fact. Yablo does this in terms of counterfactuals, and this is indeed a very natural way to do it. For example, we may note that the window would still have shattered (as a result of Suzy’s throwing the rock) if she had been wearing a blue hat instead of a red hat. This suggests that the color of hat that Suzy was wearing at the time—or her wearing a hat, for that matter—is causally irrelevant detail.¹⁵

However, focusing only on counterfactuals in this way discounts the importance, for causation, of actual causal processes. Take the Suzy and Billy case, for example. Why do we think that Suzy is a cause of the shattering, even though the disjunctive fact involving Suzy and Billy is more proportional to the shattering? Arguably, this is because there is a physical process of the right kind linking the event of Suzy’s throw with the event of the window shattering. Given the existence of this process, we rightly feel that we don’t need to appeal to anything more complex to explain why the window shattered: there is an ordinary explanation in terms of what just Suzy did.

Sometimes the physical processes that do the relevant kind of work are not actual but counterfactual (or, in more complex cases, a combination of actual and counterfactual). For example, take the Lazy case. In that case we think that Lazy is a cause of the child’s death (or at least part of the explanation of the child’s death, if omissions aren’t causes), not because of the existence of an actual physical process linking Lazy with the death, but because of a counterfactual physical process: in this case, the process linking Lazy’s attempted rescue with the child’s survival. This is a process that obtains in nearby counterfactual worlds where Lazy overcomes his laziness and attempts a rescue; in those worlds, the child survives. In cases of omission like this, the existence of a causal or explanatory relation between the omission and the outcome can only be grounded in a counterfactual physical process, not in an actual one, for omissions are absences (absences of actions of certain kinds) and absences are not part of physical processes. Still, the existence of the relevant counterfactual physical process allows us to link the agent to the outcome’s occurrence.

In contrast, our problem cases are importantly different in that in those cases we cannot link the agents to the outcomes by means of similar individual physical processes (actual or counterfactual). For example, the process that Lazy would have started by attempting a rescue in Lazy and Sharky is not one that leads to the child’s survival (Lazy gets eaten by the sharks and the child still dies). That kind of counterfactual process only leads to the child’s surviving if we also imagine that Sharky doesn’t release the sharks. Thus, the only such processes that do the relevant work in our problem cases are processes involving both agents. They are “collective” processes, so to speak, not individual processes, which represent collective contributions.

The same goes for Switcher and Reconnecter. In that case too, both agents must act differently in order for Victim to be spared (Switcher by not flipping the switch, and Reconnecter by not reconnecting the track). Again, their contributions are collective contributions, not individual contributions.

Thus, my suggestion is this. The reason we need to posit complex (collective and disjunctive) causes or explanations in our problem cases is that in those cases the outcomes are not adequately accounted for by simpler (individual) processes or contributions; hence, we need to appeal to more proportional collective contributions that are disjunctive in nature. But this doesn’t mean that we should look for such explanations in all cases. For in most cases, and perhaps in all ordinary cases, there are run-of-the-mill explanations in terms of individual contributions or processes. Those explanations are perfectly good explanations in such cases, and there is no need to appeal to collective disjunctive contributions.¹⁶

In other words, my suggested solution is that we should take the role of proportionality to be quite limited. Proportionality considerations can certainly be useful, say, at the time of choosing among alternative non-disjunctive explanations. For example, they can favor the event of Suzy’s throwing a rock over the event of her throwing a rock while wearing a red hat as a cause of the shattering. (The event of Suzy’s throwing a rock is not a disjunctive event, even if it’s less specific and more proportional than her throwing a rock while wearing a red hat.) But we cannot overdo the importance of proportionality, or else we risk ending up with highly complex disjunctive explanations in cases where there are simpler explanations that work just fine. In particular, the only times where the scope of proportionality should be taken to recommend a collective disjunctive explanation are those cases where there isn’t a simpler explanation in terms of the individual contributions of the agents involved.

Note that this presupposes that we have some antecedent grasp of when there is a perfectly good simpler individual explanation. When is there such a thing? I don’t have anything particularly useful or novel to say at this point, except that we need to look at the facts, or at the true theory of causation or explanation. A good theory of causation/explanation will entail that Suzy’s throwing a rock caused the window shattering, and that Lazy’s omission caused or explained the child’s death when there are no sharks. So, in those cases, we won’t need to appeal to collective disjunctive explanations.

In other words, my suggestion is that proportionality should be restricted by causation, or explanation, itself. Highly proportional facts such as collective disjunctive facts are explanatory only in special cases, namely, when there is a failure of individual causation or explanation. And this means that, instead of using proportionality considerations to determine what causation or explanation is, we must use causation or explanation to limit the reach of proportionality considerations.

7 Conclusion

Let me end by summarizing the discussion and what I take to be some of the most important takeaways.

First, I introduced some problem cases and I explained how they give rise to a puzzle: the puzzle of accounting for the agents’ responsibility in those cases. Building on the discussion in Sartorio (2006), I argued that we need to posit some disjunctive causes, or at least some disjunctive explainers, to ground the moral responsibility of agents in those cases, and that this seems to be independently motivated by, notably, proportionality considerations. This, in turn, means that we can’t have a blanket ban on disjunctive causes or explainers. However, such a blanket ban is what’s typically used to restrict proportionality in a way that allows us to avoid the implication that the only true causes or explanations are always disjunctive (which is obviously false). As a result, this gives rise to an important challenge of its own. In other words: solving the first puzzle results in a second puzzle.

I then argued for a solution to that new challenge. In a nutshell, the solution I proposed consists in minimizing the importance and thus the reach of proportionality considerations, in a way that departs from Yablo’s own views. On my view, proportionality shouldn’t be treated as a universal criterion or condition on causation, but it should be relegated to a more secondary role. That secondary role includes identifying some collective disjunctive causes or explainers in cases with explanatory gaps. I’ve also argued that filling in those gaps in that way provides the missing grounds for responsibility. This is not an ad hoc move, I’ve suggested, but it’s motivated by the absence of the relevant individual processes or contributions, which calls for disjunctive collective contributions.

So, one important takeaway of the paper is this: the puzzles can only be solved by balancing the right kinds of considerations in the right way—this includes the postulation of collective disjunctive causes or explainers and a reexamination of the significance of proportionality considerations. I argued that, although there is an initial tension between them, that tension disappears when the right balance is struck.

Another important (albeit less initially obvious) takeaway of the paper concerns the significance of the debate over the causal powers of omissions. I’ve suggested that it doesn’t quite matter, for the purposes of grounding responsibility, if omissions are causes or if they are explanatorily relevant in other kinds of ways. In particular, the main puzzles I’ve discussed seem to arise either way, in a similar form: in particular, if disjunctive causes are a threat, so are disjunctive explainers. Thus, this is a problem that doesn’t go away simply by working out the underlying metaphysics.

On reflection, this is as it should be. For what we’re after, when we’re looking for the grounds of moral responsibility, is not necessarily a concept of a “good cause,” but a potentially broader and more encompassing concept of what’s explanatorily relevant, in a sense that matters for moral responsibility. After all, it’s clear that agents would still be morally responsible, by omission, if omissions didn’t have any causal powers. Thus, this potentially broader concept is the concept we need, at the end of the day, if we are to accommodate the different ways in which responsibility could be grounded in the contributions of moral agents.