Monday, March 22, 2021

Doing, refraining and Double Effect

The Principle of Double Effect seems to imply that either there are real dilemmas—cases where an action is both forbidden and required—even for agents who have always been virtuous and well-informed, or else there is a morally significant distinction between doing and refraining.

Here is the argument. Consider two cases. In both cases, you know that teenage and now innocent Adolf will kill tens of millions of innocents unless he dies now.

  1. Adolf is drowning. You can throw him a life-preserver.

  2. Adolf is on top of a cliff. You can give him a push.

Double Effect prohibits throwing Adolf a life-preserver. For Double Effect says that an action that has good and bad foreseen consequences is only permissible when the bad effects are proportionate to the good effects. But the deaths of tens of millions of innocents are disproportionate to the life of one innocent teenager.

Now, I take it that in case 2, it is wrong to push Adolf over the precipice. Double Effect certainly agrees: pushing him over the precipice is intentionally doing an evil as a means to a good.

If there is no morally significant distinction between doing and refraining, then it seems that refusal to throw a life-preserver in the drowning case is just like pushing in the cliff case: both are done in order that Adolf might die before he kills tens of millions. If in the cliff case we are forbidden from pushing, then in the drowning case we are forbidden from not throwing the life-preserver. But at the same time, Double Effect forbids throwing the life-preserver. So we must throw and not throw. Thus, the drowning case becomes a real dilemma—and it remains one even if the agent has always been virtuous and well-informed.

I find it very plausible that there are no moral dilemmas for agents who have always been virtuous and well-informed. (Vicious agents might face dilemmas due to accepting incompatible commitments. And agents with mistaken conscience might be in dilemmas unawares, because their duties to conscience might conflict with “objective” duties.) I also think the Principle of Double Effect is basically correct.

This seems to push me to accept a morally significant distinction between action and abstention: it is not permissible to push teenage Adolf off the cliff, but it is permissible—and required—not to throw a life-preserver to him when he is drowning.

But perhaps there is a distinction to be drawn between the two cases that is other than a simple doing/refraining distinction. In the cliff case, presumably one’s purpose in pushing Adolf is that he should die. If he survives, one has failed. But in the drowning case, it is not so clear that one’s purpose in not throwing the life-preserver is that Adolf should drown. Rather, the purpose in not throwing the life-preserver is to refrain from violating Double Effect. Suppose that Adolf survives despite the lack of a life-preserver. Then one has still been successful: one has refrained from violating Double Effect.

Nonetheless, this is still basically a doing/refraining distinction, just a more subtle one. Double Effect requires one to refrain from disproportionate actions—ones whose foreseen evil effects are disproportionate to their foreseen good effects. But Double Effect does not require one to refrain from disproportionate refrainings. For if Double Effect were to require one to refrain from disproportionate refrainings, then in the cliff case, it would require one to refrain from refraining from pushing—i.e., it would require one to push. And it would require one not to push, thereby implying a real dilemma. But in the cliff case, classical Double Effect straightforwardly says not to push. (Things are a little different in threshold deontology, but given threshold deontology we can modify the case to reduce the number of deaths of innocents resulting from Adolf’s survival and the point should still go through.)

In fact, this last point shows that embracing real dilemmas probably will not help a friend of Double Effect avoid a doing/refraining distinction. For even if there are real dilemmas, the cliff case is not one of them: pushing is straightforwardly impermissible.

It is tempting to conclude from this that Double Effect only applies to doings and not refrainings. But that might miss something of importance, too. Double Effect gives necessary conditions for the permissibility of a doing that has foreseen evil effects and an intended good effect:

  1. the evil is not a means to the intended good

  2. the action is intrinsically neutral or good

  3. the evil is not disproportionate to the intended good.

The argument above shows that (5) is not a necessary condition for the permissibility of a refraining. It seems that all refrainings are intrinsically neutral. So, (4) may be vacuous for refrainings. But it is still possible that (3) is true both for doings and refrainings. Thus, while it is permissible to refrain from throwing the life-preserver, perhaps one’s aim in refraining should not be the death of Adolf, but rather the avoidance of doing something disproportionate. And even if (5) is not a necessary condition for the permissibility of a refraining, there may be some weaker proportionately condition on refrainings. Indeed, that has to be right, since it’s wrong to refrain to pull out a drowning child simply to save one’s clothes, as Singer has pointed out. I don’t know how to formulate the proportionateness constraint correctly in the refraining case.

We thus have two Double Effect positions available on doing and refraining. One position says that Double Effect puts constraints on doings but not on refrainings. The subtler position says that Double Effect puts more constraints on doings.


Burke Rea said...
This comment has been removed by the author.
Burke Rea said...

I'm not convinced that "violating Double Effect" is an object of intention. Perhaps its constitutive of the good of the action (or refraining), and that's what makes it a reason to act (or refrain), but it's hard for me to see it as a characterization of a means or an end. In your example, either there's no further object (so that an evaluation of acting/refraining doesn't even arise) or it's "to preserve millions of lives" (*thereby* acting/refraining in accord with DE). What makes the former implausible is that we have certain duties to members of our community (which may include the entire human community) such that our not throwing a life preserver must be characterized as act of refraining (or perhaps that, normally, we feel the weight of needing to act such that it takes an act to willfully ignore).

Similarly, "refraining" may be too indeterminate to say whether it is neutral or good. Maybe it's a *mode* of exercising agency rather than an act type: I can kill, support, lift, spray (act types) by doing something or refraining from something (either mode). The relevant act type here would be failing to save or letting drown or something more determinate (even though it still may be neutral).

When talking about doing/refraining, a student brought up the end of Batman Begins, where Batman and Ra's al Ghul are fighting on the train that is about to be destroyed. Ra's taunts Batman, saying he's finally learned that killing is necessary. Batman responds, before jumping out of the train, "I'm not going to kill you. But I don't have to save you." Maybe something similar is going on in your case. The relevant action to be done or not is saving. Usually, we have to save even strangers (barring excessive difficulty/risk/other obligations). But in the case of Adolf, we don't have to save him because we have to preserve the common good, or the many lives that we know he would end.

If we had to save Adolf from drowning (it's much easier to see that dictator-Adolf doesn't have to be saved than teenager-Adolf, though), then not throwing the preserver would be an evil for the sake of some good. A parent shouldn't let their child starve (not feeding them) for the sake of just about any further good (even "not violating Double Effect," but I'm sure there are plausible counterexamples). That would violate (3) and (4), since failing to save someone you ought to save is evil. But maybe we don't have as many universally applicable requirements as we do universally applicable prohibitions.

Alexander R Pruss said...


Suppose I am a doctor and a virtuous and reputable hospital ethicist tells me that a long and complex analysis has shown that a procedure which I proposed would violate the Principle of Double Effect. There is no time for explanations: the procedure needs to be done now or never, and if it is not to be done, then other steps must be immediately taken. I thus refrain from the procedure in order not to violate the Principle of Double Effect. It's hard to give any other end to my refraining, since I do not know any of the details in the reasoning.

There are multiple reasons why a parent might be obligated to let their child starve. For instance, there might be food enough for only one of their children. Or the food might belong to someone else who would starve if one took it from them. (One is permitted to take food for survival, but not when doing so deprives the putative owner of what is needed for *their* survival.) Or one might be Jewish, and one's persecutors deliberately gave one only non-kosher food. (Judaism holds that it is permissible to eat non-kosher when this is necessary for survival, but NOT in cases of persecution; cf. 2 Maccabees 7.)

Disproportionateness could be one reason why it is wrong to feed one's child. Suppose the food is rigged to a bomb that will kill thousands. Then it's wrong to feed the food to one's child, precisely because of Double Effect disproportionateness considerations. But now imagine that you don't know about the bomb, but I do, and you completely and rationally trust me. I tell you: "Feeding the food to your child is morally wrong for reasons having to do with Double Effect. Unfortunately, I have been informed that thousands will die if I tell you anything more. You need to trust me." If you still trust me, then you need to refrain from feeding your child, in order to avoid doing what is wrong by violating Double Effect. But you don't know any other specifics, so you can't say: "I am refraining to save thousands of lives." You trust there is good moral reason, but have little idea what it is.

Alexander R Pruss said...


Are there *any* positive duties that one should fulfill no matter what consequences? I can't think of any cases. It seems that for any apparent positive duty, if I were to learn of large enough bad consequences, I wouldn't be permitted to act. Supposing all earth life is about to be annihilated by a giant asteroid. I could blow up the asteroid. Normally, I should. But I find out that if I blow up the asteroid trillions of intelligent aliens will die. Now I see that I shouldn't save earth. (If you think that we have duties only to humans and not other intelligent species, imagine that I find out that there are human outposts outsides the solar system with trillions of humans, and they will all die if I blow up the asteroid.)

Burke Rea said...

Could an "I know not what" fill the place of an end? Maybe that's just biting the bullet, but testimony seems to be exactly the kind of place where an unknown end would show up (Equally, an unknown [in a non-technical sense] object of belief). "Why (in the sense of "for what?") aren't you doing the procedure?" "I'm not sure - but the ethicist tells me that, whatever it is I'm preserving, it's good, or whatever harm I would otherwise cause, it's bad." "Why do you think it's good or bad?" "Because the ethicist said it conforms to DE." And again, "Why aren't you feeding your starving child?" "Because there's some end I'm required to pursue that's incompatible with my feeding my child. I don't know what it is, but Alex tells me that I would otherwise violate DE." The reason some end is good/bad is not itself the end. (Something similar could be said for positively acting on the basis of testimony.)

Maybe I just have idiosyncratic views about the structure of action. I don't think "trying" is an action type or an end either. But "trying" and "according with permissibility principles" function differently than ends in explaining action (the former indicates an action's teleological aspects, without itself being ordered teleologically, the latter its desirability aspects, while the end is the [further] object of pursuit). This is relevant because refraining from saving Adolf doesn't escape the means-end structure, or similar considerations of requirements or obligations, just because what one consciously wants to do is equivalent to "doing something permissible." If you can’t intend only the good aspects of an action in deriving a means-end structure, it seems you can’t intend only the good aspects of refraining either. (And I suppose I have to say that "in order to do something permissible" is also not an end.)

Counter: Suppose a teacher assigns "Do something permissible" (or "Do something in accordance with DE") as homework. Then, doing something permissible would be an end.

Reply: Sure, but not in the same way. Suppose a teacher assigns "Write something in red ink on the board," and the student takes a black pen and writes the words "something in red ink" on the board. The student gets points for cleverness (or maybe gets docked for cheek), but not for completing the assignment correctly. Similarly, a student might complete the permissibility assignment by generating a list of everything *the teacher* thinks is permissible, with no regard for its real moral status, and then uses a random number generator to tell him what action to do. So, whatever action he does to fulfill the assignment will in fact be something permissible (because the teacher is correct), and it will in fact be his end, but he will not do it *because* it is permissible, at least not "because" in the sense of not saving Adolf "because" one wants to act permissibly according to DE.

Burke Rea said...

I concede that the wildly disproportionate consequences override familial duties. But don't familial duties override at least some disproportionate effects? Suppose my neighbor has more children than I do (which is trivially true, I guess). Even still, I owe more to my own children than my neighbor's, even if what I would provide for all of mine would aid all of my neighbor's children. (That's not absolute, though.) I take it an important aspect of Aquinas's defense of self-defense is that I have a stronger duty to protect my own life than a stranger's. Similarly, I may have a stronger duty to protect my family's lives than others. (Aquinas may be wrong, but I find that intuition plausible.)
Of course, we might say that “disproportionate” is more than just expected consequences of lives lost or saved, and can include considerations of familial duties, even if those duties aren’t absolute.

I also think we can explain the obligation to let the child starve either by denying the obligation to save (in that circumstance) or by appealing to further ends which bear on whether feeding one’s child was required or not, rather than disproportionate expected consequences.
(a) Food for one. Ought implies can, so while one is required to feed *as many* of their children as they can, they don't have the duty to feed *all* of them, because all can't be fed. The duty to feed one child is ordered to the higher good of feeding all of them, but when that cannot be done, one ought to feed as many as one can. At this point, it becomes triage, and ordinary requirements not to let someone die don’t apply.
(b) Someone else's food. The prohibition against theft of someone else's means of survival entails that there is no food *for* one's child. Ought implies can, but one cannot feed their child (for there is no available food).
(c) Persecution. The duty of fidelity to God overrides duties even to one's own health and life, and even those of one's family, because health and family are themselves ordered our fidelity to God as a higher good.
(d) Disproportion. The duty to the common good may override the duties to one's family, because individuals are ordered to the common good as a higher good. That gets tricky because individuals are also *constitutive* to the common good, which I take it is part of the reason why direct homicide is unjustifiable, even for the sake of the common good. So while the common good cannot be pursued *by* killing my child, it might be pursued despite not feeding him/her, when feeding him/her is ordinarily ordered toward the common good.

I'm not sure if there are any absolute positive duties - except for an extremely general duty like "Do good and avoid evil" or "do what God commands." But there's no more specific action that must be done at all times that would be required to fulfill those duties (even God's commands allow for fulfilling them at the right times). Particular ends might have absolute positive duties *for the sake of those ends*, but we could come up with a counterexample where the particular end isn't itself required for some further end.

Alexander R Pruss said...


Perhaps we can say that we are aiming at good and avoidance of evil, but not at any specific good. This happens in other cases. If I buy a ticket for a lottery whose prizes are (a) me getting a friend, (b) a hungry stranger getting fed, (c) me becoming wiser and (d) my best friend being cured of cancer, then I am aiming at good and avoidance of evil, but it seems I am not aiming at a specific good or to avoid a specific evil.

BTW, it seems clear to me that proportionality had better take into account relational factors.