Here’s a variation on the classic trolley problem that I haven’t seen before:
Five innocent people are trapped in the path of an oncoming trolley. You could turn the trolley onto a sidetrack, away from the five. If you do so, the five will be saved, but, foreseeably, the trolley will hit and kill one innocent person who’s trapped on the sidetrack.
As it happens, one of the five people threatened by the trolley (call her Susan) is standing on a lazy susan. You have an alternative means of saving Susan: you could rotate the lazy susan on which she’s standing, thereby moving her out of the trolley’s path.[1]
I am assuming that:
- It is neither difficult nor costly for you to turn the trolley or rotate the lazy susan.
- It is impermissible for you to do neither of those things: saving Susan, at least, is an easy rescue for you to perform. It would cost you nothing, and would not require your harming anyone.
- Other things equal, it is permissible to kill one person as a foreseen side effect of saving five, but it is not permissible to kill one person as a side effect of saving any fewer than five. (That is, other things equal, it would not be permissible to kill one person as a side effect of saving four.)
- It is impermissible for you firstto rotate the lazy susan, thento turn the trolley. This is because, once you’ve rotated the lazy susan, only four people remain under threat. Turning the trolley at that point would kill one innocent person as a side effect of saving four.
My question is:
Given that you have the power to save Susan without harming anyone or incurring costs yourself, is it morally permissible for you to turn the trolley instead of rotating the lazy susan? That is, does your having the powerto save Susan without harming anyoneexcludesaving all five and killing one as a side effect as a morally permissible option?
I’m not certain what I think about this case, but I’d love to know what others think.
Kerah Gordon-Solmon (Queen’s University)
[1]This case is the product of an e-mail conversation between Jonathan Parry and me; it was constructed jointly by both of us.
This is a very nice case! Given your stipulations, I think that we must conclude that turning the trolley is wrong in this case.
Let us start by analysing your case as one in which you have four options:
1. Turn the Lazy Susan and the trolley
2. Turn the Lazy Susan and not the trolley
3. Turn the trolley and not the Lazy Susan
4. Turn neither the Lazy Susan nor the trolley
Option 4 is certainly inferior to option 2 – since option 2 saves Susan, and otherwise option 4 has the exactly same relations as option 2 to all of the other five individuals involved (viz. you don’t kill the one on the side track, and you fail to save the remaining four).
You have stipulated that Option 1 is impermissible. If so, then by my lights this would have to be because it is (perhaps just barely) inferior to Option 2. Option 1 is in one way worse than Option 2 – it involves actively killing the one on the side-track (even if only as a foreseen side-effect of one’s action) – and in another way better than Option 2 – it involves saving the four. But let us stipulate that the respect in which Option 1 is worse outweighs the respect in which it is better, so that on balance Option 1 is worse.
Surely, however, Option 3 is in all relevant respects morally on a par with Option 1. Both save all five (including Susan) on the main track, and kill the one on the side track (as a foreseen but unintended side-effect). It surely makes no difference to your relationship to Susan whether you save her by moving her out of the path of the trolley by turning the Lazy Susan, or by moving the trolley onto a path where it no longer endangers her, by turning the trolley. And otherwise, Options 1 and 3 involve exactly the same relations to all the individuals involved. So if Option 1 is worse than Option 2, surely option 3 must also be.
Hence, all available options are inferior to Option 2, which is therefore the only permissible option.
That’s a nice version of the case. I think of Ralph’s four options, options 1, 2, and 3 are all permissible. You may turn both the Lazy Susan and the trolley because in turning the trolley you are saving 4 (the 4 aside from Susan) by redirecting a threat without in doing so thereby infringing upon anyone’s stringent rights and I think that that is permissible (in fact I think even a two-one version of the trolley problem turning the trolley is permissible–just so long as the turning minimizes harm). Option 2 is permissible because it is not required of you to turn the trolley to save the four you could save aside from Susan (in general turning the trolley is permissible but not required (contra Frowe (2017)). Option 3 is permissible because you save Susan without harming anyone and so that is permissible. Option 4 is impermissible because you have available to you option 3.
Sorry, I didn’t mean for that last comment to be anonymous.
Interesting.
What’s the argument for your third assumption? I don’t think Foot thought it mattered whether or not the trolley was heading towards five people or four people (or two people, for that matter), so long as it was more than one person.
Do you think it matters? For my part, while I’m only 90% convinced you should turn the trolley in the 5v1 case, I am 99% convinced that if you should turn it in the 5v1 case then you should turn it in the 2v1 case.
That would avoid the conundrum your bystander finds themselves in.
Playing off Thomas’s point: while the third assumption is the most suspect, the question KGS poses remains interesting if there are just two persons on the main track, one of them being Susan on the lazy susan. Although it then would seem to lead to a different conclusion. Standard consequentialism would be indifferent whether we turn the susan or the trolley as each minimizes the number of deaths (=1). Deontologists would (typically?) say that we should not sacrifice one person to save another person under threat, so we must then turn the susan. Now suppose that the only initial action available is starting a mechanism which turns the susan and trolley simultaneously; both would say we may do this, but if this opens up a second option for turning back the trolley but not the susan, then again Cists say this is indifferent, Dists (typically?) that this is forbidden. None of these results seem any more puzzling than those from past trolley variations, but they do offer a nice way of illustrating the differences between these families of theories.
Thank you for these comments – I’m so pleased this case piques others’ interest as well, and grateful for the analyses.
Ralph, I’m intrigued by your approach, of assessing the case in terms of betterness rankings.
I’ve been thinking of it in the following terms:
In your option 1 (turning first the lazy susan, then the trolley), you would be killing one as a side effect for the sake of saving four. I’ve stipulated that this is normally impermissible. (Although I take your point, Peter and Thomas, that it’s a controversial stipulation, which might prove false.)
In your option 3 (turning only the trolley), you would be killing one as a side effect, for the sake of saving five. By stipulation, this is normally permissible.
So there’s a relevant difference between 1 and 3 in terms of the justification for turning the trolley. This is the main reason I don’t think we can infer the impermissibility of option 3 from the impermissibility of option 1.
On the other hand: because option 2 (turning only the lazy susan) is on the table, it is not the case that it is necessary to turn the trolley to save each of the five. It is necessary to turn it to save only four of the five. That this is so makes me think that option 3 probably isn’t permissible. But even if this is right, more of an explanation is wanted.
Thomas –
I don’t have a robust argument for my third assumption. The thinking is, roughly, that nonconsequentialists take complying with constraints against harming very seriously, such that there needs to be a significant difference between the severity of the harm that one would prevent, and severity of harm that one would inflict (or the numbers of victims one would save vs. numbers of victims one would harm to a comparable degree, etc.) in order to justify contravening constraints against harm.
It’s the consensus view (Judith Thomson excepted) that in 5 vs. 1 cases, you’re justified in contravening the constraint against killing one as a side effect for the sake of saving 5 lives. I think – although I could well be mistaken on this point – that the consensus breaks down when it comes to killing one as a side effect for the sake of saving fewer than 5 lives. I don’t have a fixed view on this point. But I’m disposed to say (in deference to the constraint in question), that the difference in numbers needs to be greater than one life to justify killing one person as a side effect for the sake of saving n people.
Scott –
I’m afraid all I’ve got for you is agreement!
Very interesting!
A natural thought is that it matters whether the choices are made in sequence. If the Susan’s already turned, then it’s clearly wrong to turn the trolley. (Assuming that four savings won’t justify lethal diversion.) This leads to a kind of dynamic inconsistency: it’s fine to turn both if you do both at once, or if you do the trolley first, but not if you do the Susan first. Perhaps that’s not so bad.
…well, ok, it’s a little bad.
Suppose Ralph is a bystander, and he’s itching to turn the Susan because he doesn’t think you have the guts to turn the trolley. You really want to save all five, but you’re unwilling to do wrong, and Ralph can act first. The only way to stop Ralph is to do something quite undesirable — pay gobs of cash, harm the four others on the track besides Susan (with consent), etc. Now presumably you should be willing to do this nasty thing: it’s a necessary means to permissibly saving four whole splendid lives. The weird thing is that the nastiness isn’t really necessary for *saving* anyone; just for the saving to be *permissible*. You could always just turn the trolley after Ralph has rescued Susan — though, sadly, it would be wrong.
In short: if turning the Susan means no turning the trolley, then keeping the Susan in place is weirdly important, even though it’s causally and intrinsically worthless.
So I guess I think the dynamic inconsistency is too weird, which tempts me to say that it’s wrong to turn the trolley even when I face my choices all at once. (Simplified a bit: {Save None, Turn Susan, Turn Trolley}.) I can’t justify turning the trolley rather than saving Susan.
This means I don’t have to pay Ralph (etc.) to stop him from turning the Susan, since it really ought to be turned. I shouldn’t want Susan to be on the tracks — an extra threatened innocent — just so that I can shore up justification for saving the four; there is no justifiable way to save them, once the Lazy Susan is in the picture.
Daniel: Very clever. I agree that this dynamic inconsistency is weird. But of course, rather than conclude that turning the trolley is wrong even to save five, we might instead reject the provisional assumption that it is wrong to do so to save only four (or three, or two, of course).
Thanks, Kerah!
You say: The difference between Option 1 (turning the trolley and the Lazy Susan) and Option 3 (turning just the trolley) is that Option 1 involves “killing one as a side effect for the sake of saving four”, while Option 3 involves “killing one as a side effect for the sake of saving five”.
I am afraid that I need to know more about what you mean by “for the sake of”… After all, since if I take Option 3, I don’t intend to kill the one (I merely foresee that I will kill them), it’s not true that I am killing the one as a *means* to the end of saving the five.
Admittedly it is true that, in turning the trolley in the classic Trolley Problem, I regard killing the one as a price worth paying, given that this course of action will save the five. But what matters is, surely, not what I *take* to be a price worth paying, but what the balance of reasons really *is*. On my view, what is permissible or impermissible is determined by the balance of reasons, and reasons all correspond to betterness rankings of various sorts between the available options — and these rankings are ultimately grounded in naturalistic features of those options. So I don’t think that my regarding killing the one as a price worth paying can be a part of the fundamental reason that makes it permissible to turn the trolley in the original Trolley Problem case.
At all events, whatever you mean by “for the sake of” here, I also don’t think it’s right to say that in taking Option 1, one is killing the one as a side-effect “for the sake of saving the four.” My end in taking Option 1 is *saving the five*. The fact that we can break Option 1 down into two components (turning trolley and turning the Lazy Susan) is surely irrelevant — just as it’s irrelevant that we can break the action of slowing down on the freeway into two components (taking one’s foot off the gas pedal and pressing down on the brake pedal). It’s no more significant that the difference between getting onto the bus with two steps and with one. The whole course of action of Option 1 should be assessed as a unit, I think!
Daniel, that is such a great comment.
Scott, I think the line Daniel’s offering holds up, even if I’m mistaken about the minimum number of lives to you need to be saving to justify killing one person as a side effect. I.e., suppose that minimum number is three; we can tweak the case to say that it’s Susan and two other people who are in the path a runaway trolley….
Daniel, again – you end saying, “I shouldn’t want Susan to be on the tracks — an extra threatened innocent — just so that I can shore up justification for saving the four.” This is intuitively plausibly; why it’s so is the big question. It can’t be for Susan’s sake. If there was no-one on the sidetrack, and you needed the minimum weight of 5 bodies on the main track to enable pulling the lever to turn the trolley, it would be wrong to rotate Susan. Not only that, but the reason in favour of rotating Susan would be extremely weak: it’s certain (or virtually certain) that the threat to Susan would not eventuate in her harm even if you leave her where she is (because you intend to turn the trolley, saving all five).
Thanks, Ralph –
Fair enough – I ought to have gone with the simpler formulation “kill one as a side effect of saving five,” rather than “kill one as a side effect for the sake of” formulation five.
With respect to your last point – once you’ve rotated Susan, she’s no longer under threat from the trolley. So, if you rotate her first, she’s already safe when you turn the trolley; you can’t justify turning the trolley partly with reference to its saving Susan.
Here’s one more, even stranger variation on the case. (I’m sticking with the simplifying assumption that five is the minimum number of lives you’d need to be saving to justify killing one person as a side effect – but feel free to adjust the numbers according to your intuitions!)
* * *
Six people are trapped in the path of a trolley. You could turn the trolley away from the six, onto a sidetrack. But doing so would kill one person, Sid, who’s trapped on the sidetrack.
One of the six, Susan, is on a lazy susan. You could rotate her out of the path of the trolley, and onto the sidetrack with Sid.
* * *
I’ll stick with Peter’s assumption that (absent the lazy susan) it would be permissible to do nothing – that is, it would be permissible to allow the six to die. If that’s the case, then it would likewise be permissible to rotate Susan. At least then you’d been saving someone (or allowing one less person die).
But, of course, if you rotate Susan, you’d no longer be permitted to turn the trolley: once Susan is rotated, you’d be killing two as a side effect of saving five, which, I’m assuming, is impermissible.
Rotating Susan in this variation strikes me as the obviously morally inferior option. You should* leave Susan where she is and turn the trolley. The question, again, is why this is so.
* “Should” in the less-than-obligatory sense.
Thanks all, this is a very illuminating and helpful discussion!
Daniel – I agree with your analysis of why we ought to turn the Lazy Susan, rather than leave Susan in place and save the five (“I shouldn’t want Susan to be on the tracks — an extra threatened innocent — just so that I can shore up justification for saving the four; there is no justifiable way to save them, once the Lazy Susan is in the picture.”) Interestingly, it was that kind of thought that got Kerah and myself discussing these cases.
In response to Daniel, Kerah wonders what the underlying explanation might be. I agree with Kerah that it is not really Susan’s interests that are doing the work. Instead, I think the explanation should focus on the rights of the 1 innocent person on the side-track. The thought is something like this: Our rights don’t simply impose a constraint on harming us; they also prevent others from gerrymandering the moral situation so as to sustain or create justifications for overriding that constraint. What seems objectionable about leaving Susan in place (and so makes it obligatory to turn the Lazy Susan) is that the rationale for doing so would be to circumvent the constraint on killing Sid. What do people think?
Kerah, I think Daniel’s argument holds only if there is some minimum number N of lives saved which makes killing 1 permissible where N > 1. But consequentialism rejects this: any N > or = 1 makes it permissible (and for N > 1, generally obligatory). This eliminates the weirdness Daniel pointed to. For if the bystander turns the susan and still leaves >1 person on the main line, it remains permissible (or even obligatory) to turn the trolley anyway; there is no moral angst, and certainly no need to want the susan turned just to make this permissible. If there are only 2 on the track, and the susan is turned leaving 1, then turning the trolley remains permissible, just not obligatory. In the trivial case, we start with 1 and turning the susan eliminates the threat and makes turning the trolly forbidden–and we certainly don’t want that not to happen! Hence I think Daniel’s argument presents an interesting case for consequentialism (not necessarily a decisive one), whether he intended it as such or not.
Aha, I spent so long typing that I missed Kerah’s variation case. I take it (?) the variation case offers a challenge to the analysis that Daniel offered, and to which I suggested an underlying explanation. The challenge is something like this: In the initial case, Daniel and myself suggested that it is impermissible to leave persons in harm’s way in order to maintain (or ‘shore up’) a justification for killing to save the greater number (and hence why you ought to turn the Lazy Susan). But, if we share Kerah’s intuition in the variation case: that we ought to leave susan be, and thereby permissible save the greater number by killing one, then that looks like a counter-example to the explanation offered in the initial case.
Kerah. Regarding the third assumption again. Do you think there also needs to be more than a single life saved before the driver of the trolley can herself divert onto the one? To my mind, the driver should always divert down the less populated track.
But that sits ill, it seems to me, with the bystander only being permitted to divert the trolley if there is more than a single life being saved (setting aside how much more).
Imagine the trolley is heading towards 2 and the driver has fainted. The bystander is able to divert the trolley, but is not permitted. He’s willing, with all his might, and shouting and screaming trying to wake up the driver so the driver will divert the trolley. “What are you doing?” asks a passerby. “Trying to wake up the driver so he’ll divert down that track!” “But you can do it yourself, why does the driver need to do it?!” And so on. That seems like a strange state of affairs to me. (It won’t generalise, of course, to instances with villains etc, but in this particular case, it seems strange.)
Thomas, I really like your case.
I agree that the driver should always direct the trolley down the less populated track. If one’s only options are killing the greater or lesser number (and other things are equal), one should kill the lesser number.
But I think I’m willing to accept what you find strange. In your case, the runaway trolley is heading toward two people; there’s one person on the sidetrack. The driver is unconscious; you’re in the trolley with him. If the driver were awake, he would be obliged to turn the trolley, killing one rather than killing two.
Your agential perspective is different from the driver’s. Your first two options are turning the trolley yourself, saving two and killing one as a side effect, or doing nothing, allowing two to die, and killing no-one. Your third option is doing everything in your power to wake up the driver. If you succeed, you would have enabled two to be rescued, and – as a side effect – enabled one to be killed. Suppose you’re forbidden from killing one as a side effect of saving two. This need not entail that you’re forbidden from enabling two to be rescued, enabling, as a side effect, one to be harmed.
Jonathan, I’m curious: would you say, of Thomas’s case, that you’re not permitted to try to wake the driver because this would be ‘moral gerrymandering’ to bring about a constraint being overridden?
Perhaps I missed some crucial assumption, but I’m baffled by the last two messages by Thomas and Kerah. Why would the bystander who could wake the driver be able but not be permitted to turn the trolley to kill 1 instead of 2, but the driver is able (if awake) and so permitted? Are you assuming that the “driver” would kill the 2 if the trolley is not turned? But this doesn’t sound right at all; this is not a “driver” like that of a car who can direct the trolley any way one pleases. The driver, aka engineer, is simply the usual person who directs the trolley onto one track or another; but barring some particular conception of professional duties which I haven’t seen argued for here at least, epistemically and morally he seems to be in the same position as the bystander: he finds himself able to turn the trolley which, if nothing is done, will kill 2. What am I missing here, if anything?
Jonathan, to your follow-up comment – yes, exactly! At least, the variation case is evidence that there’s a more fundamental explanation underlying the one you and Daniel offered.
Here’s a go at a different analysis:
Original case:
You may not do nothing; you also may not turn first the lazy susan, then the trolley. The remaining alternatives are rotating Susan (only), or turning the trolley (only).
If you rotate Susan, you would be saving 1 (Susan), and allowing 4 to be killed.
If you turn the trolley, you would be saving 1 (Susan), saving the 4 trapped with her, and killing 1 (the person on the sidetrack) as a side effect.
1 (Susan) will be saved, whichever option you pick. For that reason, you can take Susan’s interest in being saved out of the equation. This leaves you to compare allowing 4 to be killed, vs. saving 4 and killing 1 as a side effect.
On the stipulation that it’s not permissible to kill 1 as a side effect of saving 4, you’re only remaining permissible option is to rotate Susan.
Variation case (which I’ll replicated so no-one has scroll through the comments):
* * *
Six people are trapped in the path of a trolley. You could turn the trolley away from the six, onto a sidetrack. But doing so would kill one person, Sid, who’s trapped on the sidetrack.
One of the six, Susan, is on a lazy susan. You could rotate her out of the path of the trolley, and onto the sidetrack with Sid.
* * *
Again, I’ll assume, here, that (i) doing nothing isn’t an option: you should save at least one person; (ii) you wouldn’t be morally required to turn the trolley, killing 1 as a side effect of saving 6; and (iii) if you rotate Susan, you would be forbidden from turning the trolley: stipulatively, it is impermissible to kill 2 as a side effect of saving 5.
In this variation, too, Susan will be saved no matter which you pick. So again, you can take her interest in being saved out of the equation. You are left comparing saving 5 and killing 1 as a side effect, vs. allowing 5 to die. On assumption (ii) you are permitted to do either, but, intuitively, saving 5 and killing 1 as a side effect is the morally superior option.
Scott: Yes, the original set up of the case has the driver killing the many (if she doesn’t divert). You might suppose she mistakenly steered the trolley down the wrong track towards the many in the first place and now her only way of not hitting (and killing!) them is by diverting down the spur towards the one.
Kerah: I wonder how far you’ll take that line. What if the bystander is a passenger in the trolley and (by stipulation, and unlike the driver) would not kill the two if the trolley hits them.*
He’s allowed to vigorously shake (kick! pour water on!) the driver trying to wake her up in order for her to (permissibly) turn the trolley, but would not himself be permitted to turn it.
*Quinn (1989?) would deny this, but it’s never seemed right to me.
Thomas: thank you, your second sentence makes this much more clear. As a good consequentialist, I don’t think that’s a morally relevant distinction in the end, but at least now I see what underlies the discussion in question.
Thanks for these great replies!
J&K: totally agree that Susan’s interests aren’t the explanation, though I don’t think a ban on gerrymandering can be fundamental. My hunch is that the key is the necessity constraint: killing by diversion is only kosher if it’s *necessary* to save five lives. If the Susan’s there, then turning the trolley isn’t really necessary for saving each of the five, only for saving the four.
This is all a bit delicate, since turning the trolley is necessary for saving *all five*. There’s a scope ambiguity. Perhaps we just need to interpret the quantifier as having wide scope. The crucial question: are there five people such that turning the trolley is necessary for saving them? Not if you can swivel one of them to safety.
But come to think of it, that’s not the right interpretation, either. Suppose there are six on the main track, whom you could save by diverting the trolley onto the one, but you also have a lasso, which you can use to save one of the six–no more than one, but any one you’d like. Now there isn’t anyone such that turning the trolley is necessary to save them, but it should still be fine to turn the trolley.
So here’s a new idea: make the necessity constraint comparative. We compare the option where we kill to each of the options where we don’t, and for each pair, we ask: does the killing save at least five lives more than this alternative?
This gets the right answers in both cases. In the original: killing by trolley doesn’t save five more than turning the Susan, so the killing is wrong. In the last case: killing by trolley does save five more than merely lassoing someone, so it’s permissible.
Comparativism to the rescue?
I almost forgot Kerah’s delightful variation case, with six on the main track, one on the side, and the susan set up so that it will move Susan to the sidetrack.
We have four options: {Turn Trolley, Turn Susan, Turn Neither, Turn Both}. Turn Trolley is clearly justifiable over the other way of killing–Turn Both (in which Susan is gratuitously killed). And we can save at least five more with Turn Trolley than with the remaining options: Turn Susan (6 vs. 1) and Turn Neither (6 vs. 0). That means that Turn Trolley gets the green light from the comparative necessity constraint, which seems like the right result.
Daniel, I think you’ve solved it! Invoking the necessity condition is very elegant – and it seems obvious, now that you’ve said it. (Comparativism is just cool.)
Ironically – I started thinking about these cases as riffs on ones Jonathan has discussed in print (in his 2017 PPA article, “Defensive Harm, Consent, and Intervention”). In Jonathan’s cases, you’d be adding a 5th person to a group of 4 on the main track, in order to provide a different agent at the lever with a lesser-evil justification for turning the trolley. I’d anticipated that the case-variations in this post and thread would shed light on his original cases. But now it looks like the solutions to those will have a different form than the solutions here.
Let me first say: congrats to Kerah on her prize! This post certainly got me thinking. Glad to hear other people enjoyed it, too.
Now then. At the risk of diverting a trolley onto a dead horse…
I had a quick thought on the Parry case, which seems to center on proportionality (must save five) rather than necessity.
The case: you can put someone onto the main track, where four are already under threat, which then gives the bystander at the switch enough grounds to permissibly turn the trolley onto the one on the side spur.
What if the bystander isn’t going to turn? Then you can’t put the guy on the tracks; you’re just killing him pointlessly.
What if the bystander turns no matter what? Then the case isn’t interesting and you can do whatever you want.
The interesting case is where the bystander turns iff you put the guy on the tracks. But (conjecture!) this is relevantly like a case where there’s no bystander, and *you* cause the trolley to turn by putting the guy on a special platform that triggers the switch. Even closer is the case where the platform is itself on the tracks alongside the four. But you can’t put the guy on the platform: you’d be diverting a threat onto one in order to save only four, and and that’s not enough!
Parry also has the intuition that no one has a moral reason to put themselves on the tracks to give the bystander the justification to turn the trolley. I feel like there’s something to this, but it’s also mysterious. Surely you at least have *some* moral reason to put the guy on the platform, turning the trolley to (wrongly) save the four. But isn’t that like the case where, instead of directly causing the switch, your act is causally necessary and sufficient for the bystander to do it? Why should the fact that a the causal path includes a human and lever, rather than a platform and trigger, *eliminate* your reason to save the many?