Suppose you are an ordinary virtuous agent. Up until now, you haven’t lived the kind of life that involves making any huge life or death decisions, but you are kind to your friends and acquaintances, charitable to distant strangers in need, honest to everyone except those you can’t trust etc. Now an evil demon comes to you and tells you that, for the rest of your life, he will kill 5 people (randomly selected strangers in another country) every time you act virtuously, and won’t kill any people every time you act non-virtuously. He makes it impossible for you to commit suicide (if this were an option, you might think you should immediately take it – although doing so would, quite plausibly, lead to the death of 5 people, the killing would stop there). (Alternatively, one could simply specify that the demon tells you that whenever you happen to die, but not before, he will move on to present some other ordinary virtuous person with the same problem.) On the assumption that the evil demon will keep his word, what are you to do? Given that you start off virtuous, what might we predict you do in this situation? And what might we predict the demon does in response to your subsequent acts?
Let us first consider one way of hearing the demon’s pledge. If we take acting virtuously to entail be extensionally equivalent to acting rightly, as many are inclined to do, then it follows that the evil demon will kill 5 people every time you act rightly. Suppose you think, “I will now commit an act of violence against a stranger. That will be wrong, so the demon won’t kill 5 people at this time.” It seems that an act that is normally wrong (violently attacking a stranger) is now right, because it prevents the death of 5 people. But if it is right, then it will actually lead to the demon killing 5 people. So it is wrong. But if it is wrong… You see the problem.
We have a paradox on our hands. Or do we? It might be thought the paradox is generated by a questionable consequentialist type of assumption. Perhaps your original thought was correct: committing an act of violence as a means to preventing the killing of 5 is wrong. I think many of us would resist this thought (whether or not we are committed to consequentialism as a general theory). Suppose the demon had said he would kill 50 people, or 500 people, or 5000 people. How could committing an act of violence against one person as a means to preventing the killing of many people not be permissible? However, suppose the non-consequentialist thought here is right. Then, since you are a virtuous agent, we can expect you to try to think through what you should do before you act. We are now assuming your thinking reveals your non-consequentialist commitment (if it revealed the consequentialist commitment, instead, then you would be in the paradoxical situation, at least so far as you could determine [you might not be able to determine whether the consequentialist commitment is correct]). So you would think committing the act of violence would be wrong. And it doesn’t matter that it would lead to a much better agent-neutral outcome than if you didn’t commit the act of violence. So you shouldn’t commit the act of violence. So it wouldn’t be an option for you at all, as a moral response to the evil demon’s pledge.
Perhaps some less significant act, normally considered wrong, would seem an appropriate means (perhaps you would think it not wrong to tell a lie to someone as a means for preventing the killing of five). But then the paradox is generated again. So we might think that nobody can avoid the paradox. I suppose you could avoid the paradox if the non-consequentialist thought is correct and you simply help an elderly person across the road; i.e. you do something that would be right whether or not it is a means to preventing killing, and you don’t think it becomes wrong because it foreseeably leads to the killing of five people (most of us would find this very implausible) – you avoid the paradox, but only in a way that leads to the demon actually killing people. Better to confuse the demon with a paradox than let him simply have his way!
In fact, despite my initial reasoning, the consequentialist alone might avoid this first paradox (thank you to Stephen Kearns for pointing this out to me): the consequentialist tells us it is right to choose the option (or one of the options) that has the most value, even when each of the available options is very bad. From this point of view, it would be right of me to just continue on making the decisions I was (virtuously) going to make anyway, before the demon arrived on the scene; sure, every one of my actions will lead to 5 people dying, but the actions will still be non-paradoxically right, since they will still be actions that correspond to the best options (the best option every time just includes 5 people dying, so that fact can be ignored when deliberating)! If consequentialism is correct then the demon gets to kill many people, rather than being caught in a paradox.
So far I have been considering what happens if we take acting virtuously to entail be extensionally equivalent to acting rightly. Suppose we reject this entailment (I happen to think we should). A virtuous person might reason thus: “Anything I do will lead to the killing of at least 5 people. However, suppose I work to develop a disposition that is not a virtue (perhaps, to be safe, I will develop a vice), and act on that disposition as much as I can, then I will certainly minimize killings. Let me go to the television and start the process of developing apathy and sloth.” In the situation we are imagining, the virtuous person should develop vices, or at least dispositions that guide their action that are not virtues, and act on those dispositions. They should plan not to be virtuous in the future.
If the virtuous person were to take the alternative view, that they should protect their own virtue even at the expense of many people being killed by another, they would be extremely self-indulgent, in a morally blameworthy fashion. Perhaps they realize this (they are virtuous, after all!). In that case, it seems they could continue to act in whatever “virtuous” ways they were planning to act before approached by the demon, and reason thusly: “my ways of acting are now self-indulgent, because they show insufficient concern for others, therefore I am not acting virtuously at all now, even though I am behaving the same way I would have behaved if I hadn’t been approached by the demon. My self-indulgence contaminates my virtues, and either turns them into vices, or crowds them out and makes my actions now vicious, rather than virtuous. Fortunately, that will save lives.” Now we are back in a paradoxical situation. If no one gets killed because one is acting “self-indulgently” in the way just described, then this type of self-indulgence is not a vice, so people will get killed, but then one is acting self-indulgently…
So, let’s back up: it looked like a virtuous person could avoid the paradox by simply developing sloth, and acting on it (watching television all the time, etc.). But we might also worry about this option: perhaps sloth would be a virtue in a situation where one only develops sloth, and acts on it, in order to prevent killings. However, I don’t think this thought is right. If it is right, our thought experiment has serious consequences for the way we understand virtue. It has an ultra-Humean consequence. Humeans about virtue (not ultra-Humeans), like Julia Driver, think that something can be a virtue in one social world and a vice in another, since a virtue is simply a character state that brings about generally good consequences, while a vice is a state that brings about generally bad consequences. We are meant to consider the long term and widespread effects, not just the more limited effects. An ultra-Humean would say a virtue can quickly become a vice for an individual (rather than across society in general) if it can be expected to generally bring about bad consequences from now on in the environment surrounding the individual, and a vice might quickly become a virtue for an individual if it can be expected to generally bring about good consequences from now on in that individual’s environment. The ultra-Humean view is implausible. Therefore, no paradox is generated in the case where one develops sloth in order to prevent the demon killing people.
If I am right, there is an interesting asymmetry here. Being self-indulgent about one’s own virtue in the way just imagined is enough to contaminate one’s virtues and make one act non-virtuously, but developing, then acting on sloth, simply because one wants to develop the vice for strong moral reasons is not enough to contaminate the vice and make ones laziness virtuous.
[Highlighted changes occurred after the first comment by Campbell Brown, below.]