> So far we have seen that the deontologist inevitably must support doing things that are worse for people in expectation. When acting under uncertainty, the deontologist must support doing things that make one person worse off in expectation and benefit no one. That is a bad result.
I am very confused. This is just the core claim of deontology restated. I am aware that deontology makes people "worse off in expectation" as evaluated from a utilitarian standpoint. That's the whole bit! I just don't particularly care about that or think it requires me to abandon deontological principles.
The core claim of deontology is that there are constraints, not that you should make everyone worse off in expectation--or take sequences of acts that are good for some and bad for none.
The constraints apply *even when* they make everyone "worse off" in expectation. Not sure what you mean by "sequences of acts that are good for some but bad for none."
That is decidedly NOT the deontological core claim! There is a difference between choosing A over B, when B has higher expected value, and choosing A over B, when B does nothing, and A is worse for some and better for no one. Obviously deontology sometimes supports the former, no one is disputing that. But the latter is certainly not the core claim of deontology. Thinking they are the same amounts to flattening our evaluations of actions into their expected values, ignoring the structure of harms and benefits to different people--which is a little rich for a deontologist to hold, I must say.
I always thought of the core claim of deontology as being that there are constraints against harming *one* person as a means to helping *others*. We might generally expect this to be better at least for the one!
I'm surprised that you've always thought about deontology that way, but after spending an hour reading your blog posts and subscribing to your Substack, I can't find any hard evidence to the contrary.
I did find a post where you defended a (hypothetical) Nazi concentration camp guard foiling a Jew's attempt to escape because he knew staying in the camp would ultimately make the Jew wiser. It seems like a deontological prohibition against holding people in concentration camps against their will would apply here...
It just seems objectionably fetishistic to care about rules that don't help *anyone*! You might as well have a deontic constraint against non-procreative sex. (Many historically have, of course, but I'd expect contemporary deontologists to be a little more principled, humanistic, or "patient-centered", in the sense of trying to develop a theory that actually makes some kind of sense rather than just being transparently arbitrary rule-worship.)
Edited to add: if anyone reading along is curious about the context of the concentration camp case, see the section "Respecting Moe's Normative Perspective" in this post:
(It involves the time-travelling future version of the prisoner begging the beneficent guard to *not* let their present self escape.)
The upshot (also relevant to the present discussion):
> "There’s something very strange about preferring an outcome, for S's sake, that diverges from what S would (rationally) prefer or choose for themselves. It seems disrespectful of S’s normative perspective: like an objectionable form of paternalism, only worse because you’re not even serving their interests. When beneficence and respect for autonomy both point in the same direction, it’s hard to see how any competing reason could stand against their combined force. Other reasons just can’t possibly be so significant."
It just seems objectionably fetishistic to "maximize individual autonomy" by keeping someone in a concentration camp they are actively trying to escape. You might as well think that soil nematode qualia are more important than all of mankind, past and present. (Many currently do, of course, but I'd expect intelligent utilitarians to be a little more principled, humanistic, or "patient-centered", in the sense of trying to develop a theory that actually makes some kind of sense rather than just being transparently arbitrary "hypothetical rational impartial observer" soil nematode worship.)
If you seriously think that word salad rationally (and not just syntactically) parallels my argument then I won't waste any further time on you. Muting.
i recommend this paper all the time when people ask me why i love analytic philosophy. the precision to discover a non-obvious implication of deontology is just beautiful.
Nice writeup! Finally, some love for Caspar Hare, who I should mention likes numbers.
My own take (which is far from unique) is that this literature, as well as the post-Frick literature on contractualism, is valuable but not dispositive. The examples are high noise, low signal, because they don’t specify whether the relevant parties actually consent. (“People in Suitcases” never even uses the word “consent.”)
If everyone does consent to your pushing the opaque suitcase…great! Push away. If they don’t consent…what’s the problem with saying you shouldn’t push? Consent matters! That’s the whole point of deontology.
The only real issue arises in action-sequences where it’s foreseeable that some parties will withdraw consent. But this is just the familiar problem of inalienable rights, which everybody already knew conflicted with Pareto. (See eg Kamm on agreements in MM2.) And you can also be a deontologist without believing in inalienable rights. (See eg Thomson in the Realm of Rights.)
This is far from the final word on the subject. (Often we don’t know whether people would consent and must choose anyway. Perhaps the cases should be recast explicitly along those lines.) But I thought I’d mention why I’m not yet feeling devastated :)
I don't think that this helps at all. In each of the cases, we can imagine that you can't ask for their consent because they're in a suitcase (maybe they're also unconscious). In such a case, I can't see what the deontologist can say that's plausible at all. Don't see why any of this depends on any claim about inalienable rights.
Suppose rights are alienable. Then people in the opaque suitcases might alienate their rights against being killed—and indeed it would be rational for them to do so, since ex ante this maximizes their chance of survival. Then you should kill one to save five *even if* it’s revealed to you who’s in the one suitcase.
It’s underspecified what deontologists should say about cases where people can’t consent. But suppose you take the simple view that in such you should defer to what people would rationally consent to. Then, if rights are alienable, people would be rational to alienate their rights. So you should kill one to save five in the opaque suitcase case, and you can even kill them if the suitcase becomes transparent after you made your choice. (You also don’t need resolute choice to get this; it’s enough that the right is alienated, i.e. permanently waived.)
That doesn't help with most of the problems. It doesn't help e.g. with the Hare worry or the incentive for ignorance worry or the footbridge case where your eyes are closed and you push the people while their backs are turned.
It seems that it runs all the same objections as the proposal for resolute choice--it's in no one's ex-ante interest for *their* right specifically to be killed to save to be waived. So it will have to be something like "if people are rational to prefer that some class of act is allowed, then they waive their right against that class of act." But then you just get the same issues: everyone now is made better off in expectation by waiving their right to not have killings to save.
The entire thought experiment just smuggles in the repugnant conclusion as if it's a ground truth and then proceeds from there (AKA, every consequentialist argument ever made). Literally don't even need to read past the third paragraph to dismiss the entire piece. Nice try though!
"Morality, if it means anything, means doing what’s best for everyone."
simply begs the question. That is a consequentialist assumption in the first place -- especially when you go on to claim that such a thing can and should be measured.
You're misreading the claim. It's not that morality is doing what's best in the aggregate, it's that if some action is better for every single person, then you should do that! Perhaps you should read the post before snarking about it.
Even if it were a misread, which it's not, because again, "some action" being better for "every single person," and the idea that this is an actually meaningful, measurable thing that we should be trying determine ahead of time, is still just a restatement of consequentialism -- but regardless, let's pretend it was a misread: how would "reading the post" address a misread?
The notion that actions can benefit people is not the same thing as consequentialism. If you read the post rather than sounding off after a few paragraphs, you might be able to see why your reading was a misreading! Not a guarantee, it might be that you are just incurably confused.
The notion that one both can and should work out the benefits of each action, whether on a per-person basis or in aggregate, ahead of time, and then act accordingly, is quite literally the founding principle of consequentialism. It's practically a tautology, given the name.
That this notion is patently absurd, meaningless, and impossible (and, in many cases, either disastrous or outright evil), is the general rebuttal from all other schools of meta-ethical thought. Sorry, I know you probably usually find success by asserting your position as the null hypothesis and then pretending to proceed in good faith from there, but I'll call that out when I see it.
It seems to me that "whoever is in the suitcase you are about to push off the bridge" is a person (even if we don't know many facts about them, such as what they look like), and so it's not crazy to say that killing that person violates a deontological rule.
In Section 5 it seems that the crucial step is the assumption that "If an action lowers a person’s risk of death and affects no one else, then you ought to take it." But that means we need to deal with the definition of what "a person" is. My proposal at the start of this thread was that "the person in the suitcase you are about to push" is indeed a legitimate definition of a person, and that that person is indeed negatively affected by you pushing the suitcase off the bridge (or whatever weird manipulations you do in Section 5).
In Section 6 you write "d>c because it’s better for A and worse for no one." But d is a prohibited action so it can't be better than a non-prohibited action. Similarly, in Section 8 (1) can't be better than (2) since (1) is prohibited and (2) isn't. In both cases you point out that the consequences are better for the prohibited action than for the non-prohibited action. But for the deontologist, this is not a decisive consideration because deontologists are not pure consequentialists.
//My proposal at the start of this thread was that "the person in the suitcase you are about to push" is indeed a legitimate definition of a person, and that that person is indeed negatively affected by you pushing the suitcase off the bridge (or whatever weird manipulations you do in Section 5).//
Each time you push a button you don't know that there is a harmed person. There is one person you know--A, say--and their risk of death goes down. Of course, you can designate people as "guy in the top suitcase," and he is harmed by the sequence of acts, but you can do that in the case of any action that ever has downsides. Nevertheless, it makes sense to talk about ex-ante improvements.
You can say that you're simply never allowed to kill, but again, as I explain in section 6 that involves doing what is worse for everyone--including the victim of your act. And as I explain in section 8, this will struggle to vindicate the normal permissibility of ex-ante improving medical treatments.
-"You can say that you're simply never allowed to kill, but again, as I explain in section 6 that involves doing what is worse for everyone--including the victim of your act."
Right. I think a helpful analogy would be the question of whether it is ethical to force people onto suicide missions, assuming that if the suicide mission fails then your entire country will die. It is not obvious to me that this is ethical, even though it makes everyone less likely to die including the person you are coercing. I guess this is just an irreconcilable difference of intuitions between deontology and utilitarianism.
-"And as I explain in section 8, this will struggle to vindicate the normal permissibility of ex-ante improving medical treatments."
No, because deontology (or at least the doctrine of double effect, which I mentioned in the comments to a previous post) doesn't say it's prohibited to risk people's lives, only that it's prohibited to take actions whose good effects are only obtained when those people die. In the case of medicine you are hoping that everyone lives, and for each person who lives, their survival is not logically dependent on the death of the people who die. So it's OK.
I am very skeptical of this kind of argument relying on a few ridiculously convoluted and out-of-distribution counter-examples. See "2.2.3.1. Focus on the big picture" from Beckstead's On the Overwhelming Importance of Shaping the Far Future.
This is damning for any ex-ante deontologist which seems to be the actual target of the paper, but it doesn’t generalize to classic deontology.
The issue here is that any serious deontologists will simply bite this bullet. They have as far as I can tell long stood by the fact that doing the right thing under their view may mean doing something that makes everyone worse off, but that won’t change their view.
Furthermore this whole thing just begs the question, and presupposes that we ought to do things that leave people better off. I agree with this, but a deontologist would not.
Furthermore in part 6 the preference cycle does not seem to accurately describe what’s happening.
It’s not as far as I can tell an actual preference cycle, deontologists would hold that all the outcomes where you sacrifice someone (b, d, f) are to be preferred to ones where you do not (a, c, e, and especially g). But you may only permissibly perform a, c, e, or g, even if b, d, f are strictly preferred they are simply not permissible moves.
This is a very important difference because if it was as described in the paper it would be much more damning. To us committed utilitarians it’s liable to seem equally insane but it is their view.
There is a similar argument that I think is better which is the Russian roulette case which shows that deontology reacts very strangely to uncertainty which I think is a greater systematic internal flaw to the theory than this which is more of external critique.
Who are the deontologists who have "long stood by the fact that doing the right thing under their view may mean doing something that makes everyone worse off, but that won’t change their view." In addition, that doesn't help with all the scenarios discussed.
The preference cycle is wrt which acts are better to perform than which other ones.
Any classic deontologist that holds that the good will is what determines the moral valence of actions and that that the consequences aren’t important. This is obviously how you get the murderer at the door problem, but Kant was infamously willing to take that on something like these grounds.
The preference cycle still won’t work if they don’t care about the consequences (so not concerned about Pareto)
It will then just be reworked as a,c,e> g > b,d,f
Because it doesn’t matter that nobody was made worse off, to them the difference between the killing and letting die is that significant. There is no cycle, all outcomes where you kill are just worse to them than any where you do not (merely let die). The real way to probably force them on this would be to show that killing and letting die is a spurious distinction.
Now I agree this seems insane but, it’s their view, and while it substantially undermines the views attractiveness it doesn’t rly undermine the view itself.
I’m not well read on moral philosophy around deontology, but…it seems like none of the positions described are deontology as I understand it? Surely in all the scenarios presented a deontologist would not kill some to save the lives of others, because it is wrong to kill? The positions presented all seem to be ones where “it’s okay to break the rule if X”: yet isn’t the whole point of deontology that it’s never okay to break the rule? “It’s okay to kill someone if everyone would expect to be better off” seems like utilitarianism with extra steps.
But as I show, it's not just that it violates ex ante Pareto. It violates Pareto, holding that you should take the course of action that's worse for everyone.
Worse in utilitarian terms, but why shoulda deontologist accept that it’s worse for two people to die (or three, or 100,000,000) than it is to kill someone?
If you killing them would have been better for them--making their death painless rather than painful--and would have saved someone else, then you should do it is the intuitive view.
Also, I'm not quite sure which horn of the various dilemmas you are advocating taking.
It seems to me that a deontologist would choose not to push the person (whether in a suitcase or no) onto the tracks, no matter the circumstances, because it is wrong to kill someone. Isn’t that deontology’s primary deal? You know, won’t lie to the Nazis when they come knocking because lying is wrong?
There is a consistent version of deontology that holds that but it has a bunch of other problems as I argue. And most deontologists think you should lie to the Nazi at the door!
You can't say that a view isn't a problem for someone who bites bullets. If you have to bite bullets, that is by definition a problem! I mean, the end result in the shuffle case + the button case just seem so nuts. And you can't even say it's question-begging because lots of deontologists don't accept it!
To call it bullet biting is then perhaps the wrong way to phrase. It is just the view, if I ‘bite the bullet’ on the tortured child utopia case for example, this should not bother me, there isn’t a problem here for my view.
I agree it seems nuts but it shouldn’t move a committed deontologist(that’s not an ex-ante deontologist, for them it is rather decisive).
As for many deontologists not accepting it, there is an unfortunate habit for a lot of people to take exceptions to their own views in order to tweak them at the edges on the basis of their intuitions that only makes their position less tenable. This is an example of that, it seems like a committed deontologist should just accept that it violates Pareto and make peace with that. By trying to make Pareto fit they broke everything as the post goes to show.
Most deontologists hold that it's sometimes permissible to kill (e.g. in self-defense, fighting just wars, or in the trolley switch case), but other times not (e.g. in trolley footbridge), and so they develop intricate theories about when and *why*. If it was just a list of exceptionless commandments it wouldn't be a very interesting (or plausible) theory. The explanation is also important: moral rules aren't supposed to be arbitrary. But if they're grounded in rights, which are in turn grounded in the protection of individual interests, then it seems like this shouldn't give rise to rules that are predictably worse for *everyone*...
Far too much of ethics as far as I can tell seems to consist of exactly this long winded attempt to save Kant from himself. My frustration with this is part of why I decided to not pursue a philosophy degree to graduate school.
All the arguments for (most of) these exceptions seem to fail when you really get into the weeds. Kant’s arguments were ultimately rather airtight as far as I have been able to tell even if they lead to weird conclusions. But the attempt to modify those arguments to coincide with our intuitions never seems to go well.
None of these are arguments against deontology, they are arguments against not pushing one person off a bridge to stop a train from killing two people.
You are trying to turn morality into a decision-making process in a bubble, or one that is only concerned with short term outcomes. The implication with using the veil of ignorance at a micro, trolley-problem level, is that it ignores the larger game, or assumes there is no larger game, which just isn't true.
In the larger game, we want people to make decisions based on human rights, not to throw everyone's rights out the window as long as they think they know what's best. The game you created is "if we are unlucky, exchange our bad luck with someone else if it results in net good"
The implication, or underlying assumption of utilitarianism is that we have zero control over our situation and circumstances we find ourselves in, because only if we are never responsible for our situation can we be interchangable like people in suitcases.
Are there situations that we find ourselves in based on pure bad luck? Yes. Are all situations we find ourselves in based on luck? No. And we lack the knowledge in any given situation to say someone whether someone caused their situation or whether it was bad luck, both of the person in a bad spot, and the people we are contemplating exchanging with them like suitcases.
Again, the only way any of this utilitarianism makes sense is if we are never responsible for our actions and their outcomes, and if morality exists, then we are responsible for our actions and their outcomes.
And no, I didn't read the whole thing. I couldn't make it past the third paragraph because of this.
Reply to the suitcase argument: Better odds are not the same as consent.
The suitcase argument claims that pushing is permissible because each person is better off in expectation. If no one knows who is in which suitcase, then pushing lowers each person’s chance of death from 2/3 to 1/3. So, the argument says, everyone would rationally agree to pushing.
But this confuses expected benefit with actual consent.
A person may reasonably say:
“Even if this rule gives me better odds, I do not consent to being intentionally killed as a means to save others.”
That is not irrational. Some people may care not only about their chance of survival, but also about whether they are being used as a tool. Deontology can say that this distinction matters.
Consider a clearer version. Suppose there are three people:
Two agree to the rule: “It is permissible to sacrifice one to save two.”
One refuses and says: “I do not consent to being sacrificed, even if the lottery improves my odds.”
It would be unfair to force the non-consenting person into the sacrifice lottery. The two who accepted the rule may bind themselves to it, but the person who refused should not be treated as if they consented merely because the rule benefits them in expectation.
So if someone must be placed on the bridge, the non-consenting person should not be pushed. If anything, the risks created by the sacrifice-rule should fall only on those who accepted that rule.
This shows the weakness in the article’s claim that “everyone would rationally vote for pushing.” Perhaps many would. But not everyone must. And unless actual consent is given, the fact that an action improves someone’s odds does not automatically make it permissible to intentionally kill them.
The core reply is this: a better gamble is not the same as moral permission.
Even if pushing is statistically better for each person beforehand, deontology can still object that no one may be intentionally used as a means without their consent.
While I rely heavily and strongly believe in utilitarianism, the length of the article belies its weakness. The closer you try to get to perfect logic the complexity grows exponentially. It assumes you can predict the future and this is difficult at best, especially when there is little time to make an ethical choice. What if unknown to you, there is a switch that diverts the train to a safe track, or the train is equipped with a new brakes that can stop in time, or someone you cannot see is able to derail the train before it hits anyone. The strength of deontology is that it can be used to make socially defensible choices very quickly by people who give little thought to ethics. The strongest approach combines utilitarianism with deontology as well as care and virtue ethics. The best choices are supported by all ethical frameworks. If an impossible choice must be made, like Kirk in Star Trek did, find a way to change the assumptions.
I didn’t get through all of this because it seems to make a mistake all these trolley problems make: it posits perfect knowledge. I.e., you know for certain that you can kill one to save two.
Would that it were so simple. Picture this situation in real life: you come upon a trolley emergency and it seems to you—not a trolley engineer, not a trolley driver—that you can save two people by committing murder. But you might be wrong! You might wind up pushing someone onto the track for no reason.
Given this uncertainty, a “don’t kill people even when it seems like a greater good would result” is a useful heuristic. In the real world, e.g. communism, and in fiction, e.g. Thanos, the problem with killing people in the name of the greater good isn’t that it violates some abstract law; it’s that people are fallible and you are probably wrong that killing the kulaks is either necessary or sufficient to bring about a better world.
Probably good reason to think there are not intrinsic principles of morality to be honest. Evolutionary debunking arguments complicate any case for them even if they aren’t decisive for similar reasons to why fine tuning is not.
> So far we have seen that the deontologist inevitably must support doing things that are worse for people in expectation. When acting under uncertainty, the deontologist must support doing things that make one person worse off in expectation and benefit no one. That is a bad result.
I am very confused. This is just the core claim of deontology restated. I am aware that deontology makes people "worse off in expectation" as evaluated from a utilitarian standpoint. That's the whole bit! I just don't particularly care about that or think it requires me to abandon deontological principles.
The core claim of deontology is that there are constraints, not that you should make everyone worse off in expectation--or take sequences of acts that are good for some and bad for none.
The constraints apply *even when* they make everyone "worse off" in expectation. Not sure what you mean by "sequences of acts that are good for some but bad for none."
That is decidedly NOT the deontological core claim! There is a difference between choosing A over B, when B has higher expected value, and choosing A over B, when B does nothing, and A is worse for some and better for no one. Obviously deontology sometimes supports the former, no one is disputing that. But the latter is certainly not the core claim of deontology. Thinking they are the same amounts to flattening our evaluations of actions into their expected values, ignoring the structure of harms and benefits to different people--which is a little rich for a deontologist to hold, I must say.
I always thought of the core claim of deontology as being that there are constraints against harming *one* person as a means to helping *others*. We might generally expect this to be better at least for the one!
I'm surprised that you've always thought about deontology that way, but after spending an hour reading your blog posts and subscribing to your Substack, I can't find any hard evidence to the contrary.
I did find a post where you defended a (hypothetical) Nazi concentration camp guard foiling a Jew's attempt to escape because he knew staying in the camp would ultimately make the Jew wiser. It seems like a deontological prohibition against holding people in concentration camps against their will would apply here...
It just seems objectionably fetishistic to care about rules that don't help *anyone*! You might as well have a deontic constraint against non-procreative sex. (Many historically have, of course, but I'd expect contemporary deontologists to be a little more principled, humanistic, or "patient-centered", in the sense of trying to develop a theory that actually makes some kind of sense rather than just being transparently arbitrary rule-worship.)
Edited to add: if anyone reading along is curious about the context of the concentration camp case, see the section "Respecting Moe's Normative Perspective" in this post:
https://www.goodthoughts.blog/i/141529086/respecting-moes-normative-perspective
(It involves the time-travelling future version of the prisoner begging the beneficent guard to *not* let their present self escape.)
The upshot (also relevant to the present discussion):
> "There’s something very strange about preferring an outcome, for S's sake, that diverges from what S would (rationally) prefer or choose for themselves. It seems disrespectful of S’s normative perspective: like an objectionable form of paternalism, only worse because you’re not even serving their interests. When beneficence and respect for autonomy both point in the same direction, it’s hard to see how any competing reason could stand against their combined force. Other reasons just can’t possibly be so significant."
It just seems objectionably fetishistic to "maximize individual autonomy" by keeping someone in a concentration camp they are actively trying to escape. You might as well think that soil nematode qualia are more important than all of mankind, past and present. (Many currently do, of course, but I'd expect intelligent utilitarians to be a little more principled, humanistic, or "patient-centered", in the sense of trying to develop a theory that actually makes some kind of sense rather than just being transparently arbitrary "hypothetical rational impartial observer" soil nematode worship.)
If you seriously think that word salad rationally (and not just syntactically) parallels my argument then I won't waste any further time on you. Muting.
lol.
Well said!
I think the core issue is that the linked paper is specifically challenging ex-ante deontology and not deontology as a whole.
That's just false if you read it.
i recommend this paper all the time when people ask me why i love analytic philosophy. the precision to discover a non-obvious implication of deontology is just beautiful.
It's so good!
imagine reading this and thinking continental is better! haha
Nice writeup! Finally, some love for Caspar Hare, who I should mention likes numbers.
My own take (which is far from unique) is that this literature, as well as the post-Frick literature on contractualism, is valuable but not dispositive. The examples are high noise, low signal, because they don’t specify whether the relevant parties actually consent. (“People in Suitcases” never even uses the word “consent.”)
If everyone does consent to your pushing the opaque suitcase…great! Push away. If they don’t consent…what’s the problem with saying you shouldn’t push? Consent matters! That’s the whole point of deontology.
The only real issue arises in action-sequences where it’s foreseeable that some parties will withdraw consent. But this is just the familiar problem of inalienable rights, which everybody already knew conflicted with Pareto. (See eg Kamm on agreements in MM2.) And you can also be a deontologist without believing in inalienable rights. (See eg Thomson in the Realm of Rights.)
This is far from the final word on the subject. (Often we don’t know whether people would consent and must choose anyway. Perhaps the cases should be recast explicitly along those lines.) But I thought I’d mention why I’m not yet feeling devastated :)
I don't think that this helps at all. In each of the cases, we can imagine that you can't ask for their consent because they're in a suitcase (maybe they're also unconscious). In such a case, I can't see what the deontologist can say that's plausible at all. Don't see why any of this depends on any claim about inalienable rights.
Suppose rights are alienable. Then people in the opaque suitcases might alienate their rights against being killed—and indeed it would be rational for them to do so, since ex ante this maximizes their chance of survival. Then you should kill one to save five *even if* it’s revealed to you who’s in the one suitcase.
It’s underspecified what deontologists should say about cases where people can’t consent. But suppose you take the simple view that in such you should defer to what people would rationally consent to. Then, if rights are alienable, people would be rational to alienate their rights. So you should kill one to save five in the opaque suitcase case, and you can even kill them if the suitcase becomes transparent after you made your choice. (You also don’t need resolute choice to get this; it’s enough that the right is alienated, i.e. permanently waived.)
That doesn't help with most of the problems. It doesn't help e.g. with the Hare worry or the incentive for ignorance worry or the footbridge case where your eyes are closed and you push the people while their backs are turned.
It seems that it runs all the same objections as the proposal for resolute choice--it's in no one's ex-ante interest for *their* right specifically to be killed to save to be waived. So it will have to be something like "if people are rational to prefer that some class of act is allowed, then they waive their right against that class of act." But then you just get the same issues: everyone now is made better off in expectation by waiving their right to not have killings to save.
It's like each new day I discover a new problem for deontology
The entire thought experiment just smuggles in the repugnant conclusion as if it's a ground truth and then proceeds from there (AKA, every consequentialist argument ever made). Literally don't even need to read past the third paragraph to dismiss the entire piece. Nice try though!
Assume this is a joke?
Only in tone, not substance! This claim:
"Morality, if it means anything, means doing what’s best for everyone."
simply begs the question. That is a consequentialist assumption in the first place -- especially when you go on to claim that such a thing can and should be measured.
You're misreading the claim. It's not that morality is doing what's best in the aggregate, it's that if some action is better for every single person, then you should do that! Perhaps you should read the post before snarking about it.
Even if it were a misread, which it's not, because again, "some action" being better for "every single person," and the idea that this is an actually meaningful, measurable thing that we should be trying determine ahead of time, is still just a restatement of consequentialism -- but regardless, let's pretend it was a misread: how would "reading the post" address a misread?
The notion that actions can benefit people is not the same thing as consequentialism. If you read the post rather than sounding off after a few paragraphs, you might be able to see why your reading was a misreading! Not a guarantee, it might be that you are just incurably confused.
The notion that one both can and should work out the benefits of each action, whether on a per-person basis or in aggregate, ahead of time, and then act accordingly, is quite literally the founding principle of consequentialism. It's practically a tautology, given the name.
That this notion is patently absurd, meaningless, and impossible (and, in many cases, either disastrous or outright evil), is the general rebuttal from all other schools of meta-ethical thought. Sorry, I know you probably usually find success by asserting your position as the null hypothesis and then pretending to proceed in good faith from there, but I'll call that out when I see it.
But denontogists don’t believe that, is the key thing
They don’t believe that morality has to be Pareto optimal
It seems to me that "whoever is in the suitcase you are about to push off the bridge" is a person (even if we don't know many facts about them, such as what they look like), and so it's not crazy to say that killing that person violates a deontological rule.
Well see my objections to that proposal!
What objections are you referring to? I skimmed your post for the second time and still couldn't find anything relevant.
Sections 5, 6, 8.
In Section 5 it seems that the crucial step is the assumption that "If an action lowers a person’s risk of death and affects no one else, then you ought to take it." But that means we need to deal with the definition of what "a person" is. My proposal at the start of this thread was that "the person in the suitcase you are about to push" is indeed a legitimate definition of a person, and that that person is indeed negatively affected by you pushing the suitcase off the bridge (or whatever weird manipulations you do in Section 5).
In Section 6 you write "d>c because it’s better for A and worse for no one." But d is a prohibited action so it can't be better than a non-prohibited action. Similarly, in Section 8 (1) can't be better than (2) since (1) is prohibited and (2) isn't. In both cases you point out that the consequences are better for the prohibited action than for the non-prohibited action. But for the deontologist, this is not a decisive consideration because deontologists are not pure consequentialists.
//My proposal at the start of this thread was that "the person in the suitcase you are about to push" is indeed a legitimate definition of a person, and that that person is indeed negatively affected by you pushing the suitcase off the bridge (or whatever weird manipulations you do in Section 5).//
Each time you push a button you don't know that there is a harmed person. There is one person you know--A, say--and their risk of death goes down. Of course, you can designate people as "guy in the top suitcase," and he is harmed by the sequence of acts, but you can do that in the case of any action that ever has downsides. Nevertheless, it makes sense to talk about ex-ante improvements.
You can say that you're simply never allowed to kill, but again, as I explain in section 6 that involves doing what is worse for everyone--including the victim of your act. And as I explain in section 8, this will struggle to vindicate the normal permissibility of ex-ante improving medical treatments.
-"You can say that you're simply never allowed to kill, but again, as I explain in section 6 that involves doing what is worse for everyone--including the victim of your act."
Right. I think a helpful analogy would be the question of whether it is ethical to force people onto suicide missions, assuming that if the suicide mission fails then your entire country will die. It is not obvious to me that this is ethical, even though it makes everyone less likely to die including the person you are coercing. I guess this is just an irreconcilable difference of intuitions between deontology and utilitarianism.
-"And as I explain in section 8, this will struggle to vindicate the normal permissibility of ex-ante improving medical treatments."
No, because deontology (or at least the doctrine of double effect, which I mentioned in the comments to a previous post) doesn't say it's prohibited to risk people's lives, only that it's prohibited to take actions whose good effects are only obtained when those people die. In the case of medicine you are hoping that everyone lives, and for each person who lives, their survival is not logically dependent on the death of the people who die. So it's OK.
Okay, this is epic.
This makes me want to be a philosopher just to think of weird-and-fun conceptualizations of common conundra
I am very skeptical of this kind of argument relying on a few ridiculously convoluted and out-of-distribution counter-examples. See "2.2.3.1. Focus on the big picture" from Beckstead's On the Overwhelming Importance of Shaping the Far Future.
This is damning for any ex-ante deontologist which seems to be the actual target of the paper, but it doesn’t generalize to classic deontology.
The issue here is that any serious deontologists will simply bite this bullet. They have as far as I can tell long stood by the fact that doing the right thing under their view may mean doing something that makes everyone worse off, but that won’t change their view.
Furthermore this whole thing just begs the question, and presupposes that we ought to do things that leave people better off. I agree with this, but a deontologist would not.
Furthermore in part 6 the preference cycle does not seem to accurately describe what’s happening.
It’s not as far as I can tell an actual preference cycle, deontologists would hold that all the outcomes where you sacrifice someone (b, d, f) are to be preferred to ones where you do not (a, c, e, and especially g). But you may only permissibly perform a, c, e, or g, even if b, d, f are strictly preferred they are simply not permissible moves.
This is a very important difference because if it was as described in the paper it would be much more damning. To us committed utilitarians it’s liable to seem equally insane but it is their view.
There is a similar argument that I think is better which is the Russian roulette case which shows that deontology reacts very strangely to uncertainty which I think is a greater systematic internal flaw to the theory than this which is more of external critique.
Who are the deontologists who have "long stood by the fact that doing the right thing under their view may mean doing something that makes everyone worse off, but that won’t change their view." In addition, that doesn't help with all the scenarios discussed.
The preference cycle is wrt which acts are better to perform than which other ones.
Any classic deontologist that holds that the good will is what determines the moral valence of actions and that that the consequences aren’t important. This is obviously how you get the murderer at the door problem, but Kant was infamously willing to take that on something like these grounds.
The preference cycle still won’t work if they don’t care about the consequences (so not concerned about Pareto)
It will then just be reworked as a,c,e> g > b,d,f
Because it doesn’t matter that nobody was made worse off, to them the difference between the killing and letting die is that significant. There is no cycle, all outcomes where you kill are just worse to them than any where you do not (merely let die). The real way to probably force them on this would be to show that killing and letting die is a spurious distinction.
Now I agree this seems insane but, it’s their view, and while it substantially undermines the views attractiveness it doesn’t rly undermine the view itself.
I’m not well read on moral philosophy around deontology, but…it seems like none of the positions described are deontology as I understand it? Surely in all the scenarios presented a deontologist would not kill some to save the lives of others, because it is wrong to kill? The positions presented all seem to be ones where “it’s okay to break the rule if X”: yet isn’t the whole point of deontology that it’s never okay to break the rule? “It’s okay to kill someone if everyone would expect to be better off” seems like utilitarianism with extra steps.
But as I show, it's not just that it violates ex ante Pareto. It violates Pareto, holding that you should take the course of action that's worse for everyone.
Worse in utilitarian terms, but why shoulda deontologist accept that it’s worse for two people to die (or three, or 100,000,000) than it is to kill someone?
If you killing them would have been better for them--making their death painless rather than painful--and would have saved someone else, then you should do it is the intuitive view.
Also, I'm not quite sure which horn of the various dilemmas you are advocating taking.
It seems to me that a deontologist would choose not to push the person (whether in a suitcase or no) onto the tracks, no matter the circumstances, because it is wrong to kill someone. Isn’t that deontology’s primary deal? You know, won’t lie to the Nazis when they come knocking because lying is wrong?
Like, my imagined deontologist is saying “Dying isn’t morally wrong, but killing is.”
There is a consistent version of deontology that holds that but it has a bunch of other problems as I argue. And most deontologists think you should lie to the Nazi at the door!
But that is just their view, unless we are talking about ex ante deontology but that’s just consequentialism in a jacket.
It is significant if non ex ante deontology violates ex post Pareto
Yeah I agree, but it doesn’t seem like any of this is a problem for a committee deontologist willing to bite bullets.
You can't say that a view isn't a problem for someone who bites bullets. If you have to bite bullets, that is by definition a problem! I mean, the end result in the shuffle case + the button case just seem so nuts. And you can't even say it's question-begging because lots of deontologists don't accept it!
To call it bullet biting is then perhaps the wrong way to phrase. It is just the view, if I ‘bite the bullet’ on the tortured child utopia case for example, this should not bother me, there isn’t a problem here for my view.
I agree it seems nuts but it shouldn’t move a committed deontologist(that’s not an ex-ante deontologist, for them it is rather decisive).
As for many deontologists not accepting it, there is an unfortunate habit for a lot of people to take exceptions to their own views in order to tweak them at the edges on the basis of their intuitions that only makes their position less tenable. This is an example of that, it seems like a committed deontologist should just accept that it violates Pareto and make peace with that. By trying to make Pareto fit they broke everything as the post goes to show.
Most deontologists hold that it's sometimes permissible to kill (e.g. in self-defense, fighting just wars, or in the trolley switch case), but other times not (e.g. in trolley footbridge), and so they develop intricate theories about when and *why*. If it was just a list of exceptionless commandments it wouldn't be a very interesting (or plausible) theory. The explanation is also important: moral rules aren't supposed to be arbitrary. But if they're grounded in rights, which are in turn grounded in the protection of individual interests, then it seems like this shouldn't give rise to rules that are predictably worse for *everyone*...
Far too much of ethics as far as I can tell seems to consist of exactly this long winded attempt to save Kant from himself. My frustration with this is part of why I decided to not pursue a philosophy degree to graduate school.
All the arguments for (most of) these exceptions seem to fail when you really get into the weeds. Kant’s arguments were ultimately rather airtight as far as I have been able to tell even if they lead to weird conclusions. But the attempt to modify those arguments to coincide with our intuitions never seems to go well.
None of these are arguments against deontology, they are arguments against not pushing one person off a bridge to stop a train from killing two people.
You are trying to turn morality into a decision-making process in a bubble, or one that is only concerned with short term outcomes. The implication with using the veil of ignorance at a micro, trolley-problem level, is that it ignores the larger game, or assumes there is no larger game, which just isn't true.
In the larger game, we want people to make decisions based on human rights, not to throw everyone's rights out the window as long as they think they know what's best. The game you created is "if we are unlucky, exchange our bad luck with someone else if it results in net good"
The implication, or underlying assumption of utilitarianism is that we have zero control over our situation and circumstances we find ourselves in, because only if we are never responsible for our situation can we be interchangable like people in suitcases.
Are there situations that we find ourselves in based on pure bad luck? Yes. Are all situations we find ourselves in based on luck? No. And we lack the knowledge in any given situation to say someone whether someone caused their situation or whether it was bad luck, both of the person in a bad spot, and the people we are contemplating exchanging with them like suitcases.
Again, the only way any of this utilitarianism makes sense is if we are never responsible for our actions and their outcomes, and if morality exists, then we are responsible for our actions and their outcomes.
And no, I didn't read the whole thing. I couldn't make it past the third paragraph because of this.
Reply to the suitcase argument: Better odds are not the same as consent.
The suitcase argument claims that pushing is permissible because each person is better off in expectation. If no one knows who is in which suitcase, then pushing lowers each person’s chance of death from 2/3 to 1/3. So, the argument says, everyone would rationally agree to pushing.
But this confuses expected benefit with actual consent.
A person may reasonably say:
“Even if this rule gives me better odds, I do not consent to being intentionally killed as a means to save others.”
That is not irrational. Some people may care not only about their chance of survival, but also about whether they are being used as a tool. Deontology can say that this distinction matters.
Consider a clearer version. Suppose there are three people:
Two agree to the rule: “It is permissible to sacrifice one to save two.”
One refuses and says: “I do not consent to being sacrificed, even if the lottery improves my odds.”
It would be unfair to force the non-consenting person into the sacrifice lottery. The two who accepted the rule may bind themselves to it, but the person who refused should not be treated as if they consented merely because the rule benefits them in expectation.
So if someone must be placed on the bridge, the non-consenting person should not be pushed. If anything, the risks created by the sacrifice-rule should fall only on those who accepted that rule.
This shows the weakness in the article’s claim that “everyone would rationally vote for pushing.” Perhaps many would. But not everyone must. And unless actual consent is given, the fact that an action improves someone’s odds does not automatically make it permissible to intentionally kill them.
The core reply is this: a better gamble is not the same as moral permission.
Even if pushing is statistically better for each person beforehand, deontology can still object that no one may be intentionally used as a means without their consent.
While I rely heavily and strongly believe in utilitarianism, the length of the article belies its weakness. The closer you try to get to perfect logic the complexity grows exponentially. It assumes you can predict the future and this is difficult at best, especially when there is little time to make an ethical choice. What if unknown to you, there is a switch that diverts the train to a safe track, or the train is equipped with a new brakes that can stop in time, or someone you cannot see is able to derail the train before it hits anyone. The strength of deontology is that it can be used to make socially defensible choices very quickly by people who give little thought to ethics. The strongest approach combines utilitarianism with deontology as well as care and virtue ethics. The best choices are supported by all ethical frameworks. If an impossible choice must be made, like Kirk in Star Trek did, find a way to change the assumptions.
In ethics we imagine simplified scenarios to isolate morally important features.
I didn’t get through all of this because it seems to make a mistake all these trolley problems make: it posits perfect knowledge. I.e., you know for certain that you can kill one to save two.
Would that it were so simple. Picture this situation in real life: you come upon a trolley emergency and it seems to you—not a trolley engineer, not a trolley driver—that you can save two people by committing murder. But you might be wrong! You might wind up pushing someone onto the track for no reason.
Given this uncertainty, a “don’t kill people even when it seems like a greater good would result” is a useful heuristic. In the real world, e.g. communism, and in fiction, e.g. Thanos, the problem with killing people in the name of the greater good isn’t that it violates some abstract law; it’s that people are fallible and you are probably wrong that killing the kulaks is either necessary or sufficient to bring about a better world.
Yes, no one denies that it's a useful heuristic. The question is whether it--as utilitarianism posits--merely a heuristic or as non-utilitarianism suggests an intrinsic principle of morality. To test that, we want cases where we isolate the rights-violation. For more see https://utilitarianism.net/types-of-utilitarianism/#multi-level-utilitarianism-versus-single-level-utilitarianism
Probably good reason to think there are not intrinsic principles of morality to be honest. Evolutionary debunking arguments complicate any case for them even if they aren’t decisive for similar reasons to why fine tuning is not.