ABSTRACT: Some philosophers who defend the claim that there is a morally significant difference between killing and letting die (doing and allowing harm) rest their arguments on a controversial claim in the metaphysics of causation: that omissions cannot be causes. Not wanting to let our moral theory be thus hostage to metaphysics, Tomkow and I defended an account of "the Dif" which does not assume this. We did not deny that when Baker stands by, twiddling his thumbs while a child drowns, his failure to save the child may be an action, an event, and a cause of the drowning. But in so doing we left ourselves open to the criticism that our account requires an equally controversial metaphysical claim: that omissions can be causes. I think that this assumption can be defended but it would be nice to have an account of the Dif which does not take a stand on the question of whether omissions can be causes. In this post, I offer such an account.
Tomkow and I have argued that the difference between killing and letting die is an instance of a more general distinction between an agent causing an outcome and an agent allowing that outcome. Suzy broke the window (by throwing the ball, which hit the window). Billy allowed the window to be broken (by not reaching up with his catcher's mitt to catch the ball).
In saying this, we were not invoking any mysteries of agent causation; the difference between an agent's causing and allowing an outcome is, we said, an instance of the difference between an object's causing and allowing an outcome. The rock broke the window; the acid corroded the metal; the dog ate my homework. By failing to destroy the rhinovirus, the immune system allows us to catch cold but the common cold is not an autoimmune disease; by failing to sound, the defective fire alarm allowed but did not cause the fire.
In claiming that the difference between killing and letting die is an instance of the difference between agent (or object) causing and allowing, we were not invoking the claim, made by some defenders of the moral significance of "the Dif" that omissions (or failures or not-doings) cannot be causes. On the contrary, we said that an agent or object allows an outcome only if the agent or object's failure to do something is a cause of that outcome. (If you don’t think that failures can be actions or events, feel free to understand this as a claim about fact causation.) The defective fire alarm allowed the fire only if its failure to sound was a cause of the fire; Billy allowed the breaking of the window only if his failure to catch the ball was a cause of the breaking. The Dif between causing and allowing an outcome is not the difference between causing and not causing the outcome; the Dif concerns how the agent or object causes the outcome.
Setting aside some complications, the account we endorsed can be stated as follows:
An agent S caused an event E iff E happened and the causes of E include one of S's basic actions.
An agent S allowed an event E iff E happened and something that S did is among the causes of E but S didn't do that thing by doing any basic action that caused E.
'Basic action' is a term of art. We do some things by doing other things. A person's basic action, on a particular occasion, is something that she does, but not by doing anything else. To use Davidson's well-known example, someone might alert the prowler by turning on the light, turn on the light by flicking the switch, and flick the switch by moving his finger. In this example, the person's basic action -- his moving his finger -- is a voluntary and intentional body movement. But we also do things by accident, without intention, and without acting voluntarily. A person might turn on the light by bumping into the switch, or by being pushed into the switch. Our use of 'basic action' is intended to cover these cases as well. Our basic actions are the body movements whereby we do whatever else we do.
Our account of the Dif allows us to agree that Baker, who stood by, twiddling his thumbs while a child drowned, did something -- he intentionally failed to jump in to save the child; he refrained from rescuing her; he willfully omitted to perform any life-saving action. We can agree that what he did - his omission, refraining, or failure -- counts as one of his actions in the usual philosophical sense of action; it was something over which he had the relevant kind of control. And we can agree that his omission was among the causes of the child's death.
But Baker's omission -- his intentional failure to jump -- isn't one of his basic actions. He failed to jump by standing by, twiddling his thumbs. Neither the twiddling of the thumbs nor the standing by are among the causes of the child's death, so our account says that Baker has only allowed, not caused, the death of the child. He let her die.
By contrast, Able's push - a basic action - caused the child to fall into the water and drown. So our account says that Able caused the child's death. He killed her.
Our account has some nice features. It explains the ubiquity of our talk of object and agent causation without invoking controversial metaphysical assumptions; agent causation is an instance of object causation, and object causation is just a special case of event causation. It does not attempt to provide an analysis of causation and is thus compatible with a variety of different theories. It is compatible with a wide range of different accounts of events (and causal relata more generally), including a liberal theory of events that counts omissions as events.
Despite these virtues, you may worry that our account of the Dif is suspect because it takes a stand on a controversial matter in the philosophy of causation.
Our account assumes that an agent's omission or not-doing of an action can be a cause and the current literature on the metaphysics of causation is divided on this point. On the one hand, Schaffer has provided some compelling arguments that causation by disconnection is ubiquitous, and that it would be a disaster for a theory of causation to deny that this counts as genuine causation. Cases of causation by disconnection are cases in which it appears that the non-occurrence or absence of an event (in the technical terminology of the literature, an omission) is both an effect of a prior event and a cause of a subsequent event. Strangling a person (an event) causes the absence of the flow of oxygenated blood to the brain (an omission), and this in turn causes death (another event). On the other hand, there are serious worries concerning omissions as causal relata: worries concerning the ontological status of omissions, worries concerning the spatio-temporal location of omissions, and worries about how to draw a principled distinction between the omissions that are causes and those that are not.
There might be a middle ground. Perhaps we can agree that cases of causation by disconnection are cases of genuine causation without acknowledging omissions as causal relata, for in these cases there are two wholly distinct events (the strangling and the death) and the occurrence of the second event counterfactually depends on the occurrence of the first. Counterfactual dependence of wholly distinct events ordinarily suffices for causation. So if we are willing to say that the strangling caused the death without invoking the absence of oxygenated blood flow as a causal intermediary, we can count this a case of genuine causation without admitting omissions as causes or effects. If all cases of causation by disconnection can be handled in this way, then -- perhaps -- the correct theory of causation is one that does not acknowledge omissions as causes.
I've got doubts about whether this middle ground can succeed; my hunch is that an adequate theory of causation must acknowledge omissions and absences as causes. (I will explore these questions in future blog posts.) But it would be nice to have an account of the Dif that doesn't force us to take a stand on this matter. So in the remainder of this blog post, I will try to formulate an alternate account of allowing -- an account that neither affirms nor denies that omissions can be causes. (So far as agent causing is concerned, the account is unchanged from our original analysis.)
Analysis 1:
S caused E iff E happened and E was caused by a basic action of S.
S allowed E iff E happened and S could (in some relevant sense, e.g. had the ability and the opportunity) have prevented E. For short, S could have prevented E.
Thus, Baker allowed the child's death because he's able to swim and he was nearby; he could easily have saved her. The toddler playing in the sand near the water did not; he doesn't know how to swim (or how to do anything else that might have prevented the death). Queen Elizabeth did not allow the child's death; regardless of her swimming and other abilities, she was too far away at the time; there is no relevant sense in which she could have prevented the child's death.
Counterexample: Bad Guy shot Vic and Vic died immediately. Bad Guy caused Vic's death. But Analysis 1 says, falsely, that Bad Guy allowed Vic's death. For it is also true that Bad Guy could, in some relevant sense, have prevented Vic's death. Bad Guy could have refrained from shooting Vic, and had he done so, Vic's death would not have occurred.
If this counterexample seems fishy it is because we are reluctant to say that a person prevents death simply by not shooting a potential victim. This makes preventing too easy; it means that anyone with a gun and easy access to a potential victim is preventing that person's death. But we want to draw a distinction between the deaths we "prevent" by not shooting (not strangling, not drowning, etc.) and the deaths prevented by bodyguards, lifeguards, doctors, and others who intervene to save lives. When a bodyguard takes the bullet intended for his client he prevents death by his basic action; his leap intersects the bullet's path, stopping it from getting further. When a lifeguard saves the life of a drowning child, she prevents death by her basic action of pulling the child out of the water. When a doctor performs CPR, she prevents death by the basic actions which restart the heart of the person who would otherwise have died.
We want our analysis of the Dif to distinguish true preventings from the kind of "preventing" that Bad Guy would have done, had he not pulled the trigger. This suggests the following revision:
Analysis 2:
S caused E iff E happened and E was caused by a basic action of S.
S allowed E iff E happened and S could (in some relevant sense) have done some action A such that:
i) if S had done A, S would have done A by doing some basic action B; and
ii) B would have prevented E.
Analysis 2 delivers the right verdict about Able and Baker. Able caused the child's death because one of his basic actions -- his push -- caused the child's death. Baker allowed the child's death because Baker (who knows how to swim and is nearby) could have done something (jumped into the water in an attempt to save the child) such that if he had done it, he would have done it by performing some basic action (pulling the child out of the water) which would have prevented the child's death.
Analysis 2 also gives the correct verdict about Bad Guy. Bad Guy could have refrained from shooting Vic, and if he had refrained, Vic's death would not have occurred. But if Bad Guy had refrained, and Vic remained alive, it still wouldn't be true that Vic's death was prevented by one of Bad Guy's basic actions. There are many different ways in which Bad Guy might have refrained from shooting Vic. He might have turned on his heel and walked away; he might have stayed where he was and smoked a cigarette; he might have continued to point his gun at Vic for a long time before finally shooting into the air. In all these scenarios Vic remains alive. But neither the firing into the air, nor the smoking, nor the walking away are among the causes of Vic's survival.
But Analysis 2 fails as well.
Counterexample: A. Sassan shoots and kills Victor while the back-up assassin, Baxter, watches, prepared to do the killing himself should Sassan fail. Sassan caused Victor's death. But Sassan could have shot Baxter instead of shooting Victor. And if Sassan had shot Baxter instead, he would have done so by a basic action (squeezing the trigger of his gun) which would have caused Baxter's death and thereby prevented Victor's death. So according to Analysis 2, Sassan lets Victor die. But that's wrong. Sassan could have let Victor die -- by standing by, letting Baxter kill him instead. And Sassan could have prevented Victor's death -- by shooting Baxter instead. But, as a matter of fact, Sassan did neither of these two things. What he actually did was shoot Victor, causing his death.
We can rule out cases of the Sassan variety by adding a clause to our analysis.
Analysis 3:
S caused E iff E happened and E was caused by a basic action of S.
S allowed E iff E happened and S could (in some relevant sense) have done something A such that:
i) if S had done A, S would have done A by doing some basic action B; and
ii) B would have prevented E; and
iii) S did not actually cause E. (That is, E was not caused by any of S's basic actions.)
But Analysis 3 doesn't work either, for it is possible for a person to both cause and allow the same event. Indeed, many cases of causing are also cases of allowing. Smith causes and allows Victim's death by poison if he gives him the poisoned drink and also refrains from giving him the antidote. Able lets the child drown, in addition to drowning him, if he first pushes the child into the water and then stands by, refusing to save him.
In these cases a person first does something that would (if not prevented) cause an event and later has, but does not exercise, the ability to do something that would prevent that event. If all cases of causing and allowing are sequential in this way, we can revise our analysis by adding a temporal constraint:
Analysis 4:
S caused E iff E happened and E was caused by a basic action of S.
S allowed E iff E happened and there was a time t at which S could (in some relevant sense) have done something A such that:
i) if S had done A, S would have done A by doing some basic action B; and
ii) B would have prevented E; and
iii) no basic action of S at time t was a cause of E.
Is Analysis 4 correct? That depends on whether it might be true, of someone, that he is, at the very same time, both causing and allowing an event.
You might think this impossible on the ground that a person allows an event only if he can, but doesn't, intervene to prevent it from happening, and the basic actions that are, or would be, preventings of an event must happen after the basic actions that are, or would be, causings of that event. You can't pull a child out of the water before the child has been pushed into the water.
It may often be true that the basic action that is, or would be, a preventing happens after the basic action that is, or would be, a causing. But must it always be so? Consider the following case.
TWO BUTTONS: Two buttons are connected to a light in the following way. Pushing the first button causes the light to go on a second later; pushing the second button disables the light so it won't go on, even if the first button is pushed, and even if the first button is pushed at exactly the same time as the second button is pushed. Given this set-up, one person, A, might push the first button at time t1, while another person B pushes the second button at t1; the second-button pushing cancels the effects of the first, so the light doesn't go on. In this story, B has prevented the light's going on, so A doesn't succeed in causing the light to go on. Now change the facts just a bit so B doesn't (though he could) push the second button at t1. Now it is true that A caused (at t1) the light's going on and B allowed (also at t1) the light's going on. Now change the facts once more so only one person, S, is involved, and S has easy access to both buttons. Then S can perform any of the following combinations of actions at the same time:
i) S doesn't push either button (light doesn't go on);
ii) S pushes both buttons (light doesn't go on);
iii) S pushes the first button and doesn't push the second (light goes on);
iv) S doesn't push the first button and pushes the second (light doesn't go on).
In scenario i) the light doesn't go on, but the causes of the light's not going on don't include any of S's basic actions. S doesn't prevent the light's going on.
In scenario ii) S prevents the light's going on. By pushing the second (disabling) button with his right hand he prevents his left-handed button-pushing from causing the light to go on.
In scenario iii), S both causes and allows the light's going on. He causes because he pushes the first button with his left hand. He allows because he refrains from using his right hand to push the second button.
(It doesn't matter, for present purposes, how we classify scenario iv) I will be arguing, in later blog posts, that iv) is also a case of preventing.)
TWO BUTTONS shows that Analysis 4 is false. Someone may cause an event and at the very same time allow that event. We'll need to revise our analysis one more time to accommodate this possibility. But first, a few remarks about the structure of causation that makes cases like TWO BUTTONS (and, more generally, other cases of causing and allowing) possible.
Regardless of what the correct theory of causation turns out to be, the events we usually call 'causes' (or 'the cause') rarely bring about their effects all by themselves. The right background conditions must also be present. Striking a match causes it to light only if the match is and remains dry and in the presence of oxygen. Turning the key in the ignition causes the car to start only if the battery is working. A gun fired point blank at the victim's heart causes death only if the bullet isn't stopped by a bulletproof vest or the body of another person. Cutting the rope that is supporting the dangling mountain climber causes death only if the fall of the climber isn't stopped or slowed by a back-up rope, safety net, or parachute.
On some theories of causation, these favorable background conditions count as causes. Other theories of causation deny the status of causes to these background conditions, claiming that they are merely necessary or enabling conditions. No matter. The key point is that the events that everyone calls 'causes' -- match-strikings, key-turnings, gun-firings, and so on-- cause their effects only given the co-operation of the environment during the relevant time. To put it another way, the events we usually call 'causes' do not nomologically necessitate their effects. It is possible, given the laws, that the striking of the match (the turning of the car key, the firing of the gun, the cutting of the climber's rope) happens but the match doesn't light, the car doesn't start, the victim's death doesn't happen. So there is a sense in which it is nearly always possible for the cause to happen without the effect also happening.
In many cases, this is a mere possibility. (But note that it's a nomological possibility, not just a logical or metaphysical possibility.) If the climber is unprotected -- by a backup rope, parachute, safety net, or something soft to cushion his fall -- then there is no real possibility that cutting the rope won't result in the climber's death. In these cases there is no relevant sense in which anyone can intervene to prevent the cause (the cutting of the rope) from causing its effect (the death). Perhaps something could have been done earlier -- the back-up rope or parachute. But by the time of the causing it is too late.
But in other cases, the event we call 'the cause' doesn't guarantee its effect in this kind of way, and in these cases it may be true that a suitably positioned person both causes and allows the effect. I accidentally knock a vase off the shelf, but quickly grab it before it reaches the ground, thereby preventing the breaking I would otherwise have caused. If I could have, but didn't, perform this preventing act ("I never liked that vase") then I both caused and allowed the breaking. Struck by sudden remorse immediately after handing his victim the poisoned drink, Smith shouts "Don't drink! It's poison!" If he could have, but didn't, shout out the warning, then he both caused and allowed the death. In these cases the basic action that is, or would be, the cause happens before the basic action that is, or would be, the preventing. But the time of the preventing action doesn't matter. What's relevant, so far as causing and allowing are concerned, is only that the person whose basic action caused an event was able (in the relevant sense) to do some other basic action that would have prevented that event by making it the case that one of the (other) necessary conditions of the event is missing. A dropped vase breaks only if it falls all the way to the ground. A person dies of poison only if he drinks the poison (and only if he isn't given the antidote). A struck match lights only if it is and remains dry. A domino falls over only if it isn't glued to the surface or otherwise securely held in place. Once we see this we'll see that cases of causing while also allowing are common. I strike a (dry, well-made) match with one hand while pouring water on it with the other, thereby preventing the lighting I would otherwise have caused. If I've got the water but don't pour it, then I have caused and at the same time allowed the lighting. A child stacks up a row of dominos, then pushes the first one over while holding her hand on top of the fifth domino, thereby preventing it and the rest from falling over. If she could have, but didn't, secure the fifth domino, then she caused and at the same time allowed the fall of the dominos. Smith offers his victim the poisoned drink while whispering in his ear "Don't drink! It's poison.", thereby preventing the death he would otherwise have caused. If he could have, but didn't, deliver the warning, he has simultaneously caused and allowed the death.
Objection: You may be right about the structure of causation, but doesn't this show that something has gone wrong with your attempt to provide an analysis of the Dif? After all, the hope was to discover a non-moral difference between causings and allowings that would ground the moral difference.
No. Things haven't gone wrong. First, not every causing is also an allowing. Second, even in those cases where a person causes and allows at the same time, the causing is distinct from the allowing because the relevant basic actions are distinct. What makes it true that a person is both causing and allowing an event are facts about, on the one hand, the basic action whereby he is causing, and, on the other hand, facts about a different possible but non-actual basic action whereby he would be preventing. Finally, it remains true that most allowings are not also causings. And that's what's crucial, so far as the claim that ceteris paribus, causing harm is morally worse than allowing harm. What we mean, when we say this, is that causing (with or without also allowing) harm is morally worse than only allowing harm.
Let's return, then, to the attempt to provide an analysis of the Dif. We want our analysis to accommodate the possibility of causing and also allowing without collapsing the distinction so that every case of causing is also a case of allowing. Here's an analysis that might do the trick.
Analysis 5:
S caused E iff E happened and E was caused by a basic action of S.
S allowed E iff E happened and E's causes included some event C and S could (in some relevant sense) have done some action A such that:
i) If S had done A, S would have done A by doing some basic action B; and
ii) C would still have occurred; and
iii) B would have prevented E; and
iv) B would have prevented E by preventing C from causing E.
Analysis 5 gets the right results about the cases we just discussed. It says that the child allowed, as well as caused, the falling over of the fifth domino because his left-handed holding of the domino would have prevented his right-handed push from knocking the domino over. It says that I allowed, as well as caused, the lighting of the match because my right-handed pouring of water would have prevented my left-handed striking of the match from lighting it. It says that Smith allowed, as well as caused, the death of his victim because his warning would have prevented his offer of the poisoned drink from causing death. On the other hand, Analysis 5 say that the person who cuts the rope of the dangling mountain climber causes, but does not also allow, death. For there was nothing he could have done (then and there, in the relevant sense) to prevent his rope-cutting from causing death.
What about Sassan? Sassan is a more complicated case, but I think Analysis 5 gets the right result. Granted, there are ways of telling the story such that Sassan allowed the death he also called. He might have shouted out a warning to Victor as he fired, and then very quickly turned his gun on Baxter before Baxter had a chance to fire. If that sequence of basic actions would have prevented Victor's death, and if Sassan could have done these things, then he both caused and allowed Victor's death because he allowed his own basic action to cause the death. But let's suppose that no such details are present; there was nothing that Sassan could have done, either at the time of his shooting or later, to prevent his shooting from causing Victor's death. (Either Victor wouldn't have heard the warning or wouldn't have been able to duck quickly enough.) If those are the facts, then Analysis 5 correctly says that Sassan caused, but didn't allow Victor's death. For though it is true that Victor's death happened and Sassan could have prevented it (by the basic action of shooting Baxter instead), it isn't true that Sassan would have prevented Victor's death by preventing an event that was one of the actual causes of Victor's death from causing his death. For Baxter's shooting wasn't one of the actual causes of Victor's death. So Analysis 5, unlike Analysis 2, is able to say that while Sassan could have prevented and could have allowed Victor's death, what he actually did was cause (and only cause) the death.
I conclude that Analysis 5 succeeds as an alternative account of the Dif. Even if it turns out that omissions (absences, failures, non-occurrences) are never causes, there is a nonmoral difference between causing and allowing an event.
I am more sure of the claim that there is a Dif between causing and allowing an event than I am of the claim that omissions are causes, and that is why I have pursued this investigation. But there may be a bonus.
If Analysis 5 -- or, for that matter, any of the analyses I have been considering here -- succeeds as an alternative account of the Dif, we may have discovered something about causation. One of the biggest problems for acknowledging omissions (or failures to act, or non-occurrences of the relevant actions) as causal relata is the problem of spurious causes. On what principled grounds do we distinguish the omissions that are genuine causes from the ones that are not? The problem is that it is too easy for the relevant counterfactuals to be true: If Queen Elizabeth (or the toddler playing in the sand, or the old man in a wheelchair) had pulled the drowning child out of the water, the child's death would not have occurred. But the absence of the relevant action by Queen Elizabeth, the toddler, and the old man is not a cause of the child's death.
But note this. If any of the analyses I have been considering here succeeds, then the cases in which a person allows an event are all cases in which the person could, in some relevant sense, have prevented that event. This rules out -- as it should -- Queen Elizabeth, the toddler, and the old man as allowers. But these are also the cases where an omission is a spurious cause.
This suggests that the key to solving the spurious causes problem for omissions is to be found in our commonsense talk of allowing. If we aren't prepared to say that a person allowed an event, we should not count that person's failure to perform some relevant action as a cause of that event, even if the event would not have happened had the person done the relevant thing.
I will explore this thought further in a future blog post.
