Philosopher extraordinaire has a paper pressing a pretty forceful objection to deontology: it seems deontologists are on the hook for thinking that third parties should hope that people do the wrong thing, in cases where utilitarianism and deontology recommend different actions. The details are complex, so I won’t try to summarize them. But it occurred to me recently that there’s a version of this that’s more straightforward and perhaps even more forceful.
Consider some prototypical case of an action deontologists think is wrong: killing one to save five by killing the one and harvesting his organs to save five. DISCLAIMER: utilitarians should agree that this would ordinarily be very bad to do, because it’s reckless and likely to have horrible consequences—to get around this we need to add the stipulation that you know with certainty that the action will be secret, it will have no other negative effects, etc. Deontologists think this action is wrong and you shouldn’t do it. So consider the following three possibilities:
The doctor does nothing. The one doesn’t die, the five do.
The doctor has an odd involuntary muscle twitch that lasts unusually long. This twitch leads to them involuntarily killing the one and saving two people, in a way that is relatively painful.
The doctor voluntarily kills the one painlessly to save five.
What should you prefer? What should be your ranking across these? I’m going to use > to denote preferability. A>B means “you should prefer A to B.”
Obviously 3>2. 3 is better for four people and worse for no one: it leads to three extra people’s lives being saved plus the victim dying painlessly instead of painfully. Similarly, it’s clear that 2>1. In 2, the doctor doesn’t perform any voluntary actions. Instead, the killing is merely an involuntary muscle spasm. So just as you’d rather one person die than two, you’d rather an involuntary muscle spasm that leads to one person dying rather than two.
Here’s another way of making the point more vivid. Imagine swapping out 2 with:
2*. By chance, the one person who it appears wasn’t going to die—the one with the healthy organs—dies painfully, but two of the five who were on track to die instead survive.
In this case, obviously 3>2* and 2*>1. Now, the preferability relation is transitive. If you ought to prefer A to B and B to C, then you ought to prefer A to C. So then from this it follows that 3>1. This means that you should prefer that the doctor kills the patient to save five to them not doing so if you’re a third party.
But here’s where this argument has even more force. Consider the perspective of the doctor. What should they prefer? It seems that likewise their preference ordering should be 3>2>1 and 3>2*>1. So this seems to deliver the result that deontologists should think that you should prefer that *you *take the deontologically-prohibited action that has good consequences even at the time when you take it. At the time you refuse to push the man off the bridge or kill the man to harvest his organs, you should hope that you push him or kill him.
But this seems incoherent. How can it be that you ought to hope that you do P, it be in your power to do P, yet you shouldn’t do P? Shouldn’t you do what you should hope you should do? This is a very paradoxical set of attitudes. It also intuitively just seems crazy for this to be your preference ordering! Surely if you should hope you P—and if God is sitting up in heaven hoping you P—then you should P. Come on, in such a case, just go P!
I can think of a few objections.
First, you can deny that there are preferences you’re supposed to have. An objector might say: your preferences just depict what you want, but there are no rational constraints on what you *should *want. Here we should distinguish between two senses of preference. One sense of a preference is what you feel some psychological pull towards. This, arguably, is beyond rational evaluation.
But there’s another sense of preference which is what you’d all things considered rather happen. This is subject to rational evaluation. You should all things considered rather, say, that the world not end tomorrow, that your mother not drop dead, and that fewer people go to hell rather than more. If a person would rather the world be full of endless suffering, slavery be reinstituted, or mallets fall on the heads of every baby in the world, they’d be making an error with respect to which things are more choiceworthy than which other things.
Second, you might deny that it makes sense to have preferences between actions and states of affairs. But this obviously makes sense. It makes sense to say “I hope I exercise tomorrow, but I’d rather the war in Ukraine end than I exercise tomorrow.”
Third, you might deny that the doctor should prefer 3>2. This can be explained in the following way: they should rather 2 than 3 because in 3 they do something immoral. But this involves too much concern for one’s hands being dirtied. On this picture, one prefers three extra people die and one person experience intense pain so long as it spares them from acting wrongly. Obviously that is immoral.
We can also replace 2 with:
The doctor has an odd involuntary muscle twitch of unusually longlasting magnitude. This twitch leads to them involuntarily killing the one and saving two people, in a way that is relatively painful. It also leads to them fiddling with their brain so that they turn evil.
In this case, the doctor is a worse person in 4 than 3. Thus, clearly the doctor’s preference ranking should be 3>4>1. Because 3>1, the problem repeats.
In addition, this still implies that third parties should hope people do the wrong thing. This is problematic because what one should prefer seems tightly bound up with all sorts of other things including:
What one should prevent. If you want X to happen, you shouldn’t prevent it! So then on this picture, if you can prevent the doctor from killing one person and harvesting her organs to save five, you shouldn’t! Similarly, if a legal system banned killing one to save five, and this prevented some instances of this occurring, that would be
unfortunate.What one should regret. If you prefer that something happen, then you shouldn’t regret it. But it seems weird to think that you shouldn’t kill one person to save five, but then you should immediately regret it.
What you should want to
convince people of. It seems odd that you should want people not to be convinced of deontology, even though deontology is right, even if the only effect of them being convinced of deontology is acting in the way deontology recommends. It similarly seems odd that you shouldn’t talk your friend out of killing one to save five, even if you could, because you’d rather he do it than not.
The deontologist is then left with a very weird picture of the world. You shouldn’t kill one to save five but you should:
Recommend others kill one to save five.
Hope you’ll kill one to save five in the moment when you’re making the decision not to do it.
Hope that efforts to kill one to save five succeed.
Not prevent actions that would kill one to save five.
Hope you involuntarily kill one to save five.
All these attitudes are counterintuitive, but they’re forced upon us by strong arguments. But once you’ve adopted all these counterintuitive attitudes, then you should just go all the way and think you should kill one to save five. You’ve already onboarded most of the counterintuitiveness; taking on this final judgment has the nice upside of making your theory elegant and unified, rather than a hideous mess where you should hope you do what’s wrong.
