A Tangled Loop by Kerah Gordon-Solmon

That is my sketch of the trolley track in Judith Thomson’s loop variant of the trolley problem. The thing that looks like a lollipop is a trolley track. The thing that looks like a wonky musical clef is a trolley. Outside the frame, there is a switch for operating the trolley.

The trolley is heading toward the loop, bearing left. It is on course to circle the loop over and over, in a clockwise direction. But, if someone flips the switch before the trolley enters the loop, the trolley will bear right, to circle in a counter-clockwise direction.

Now we can add some people! Five small people are at 9 o’clock. One large person is at 3 o’clock. Whichever group the trolley hits first will stop it, preventing it from hitting the other group.

What is the rationale for turning the trolley? The one large person will stop the trolley, preventing it from hitting the five.

What is the rationale against turning the trolley? The five small people will stop the trolley, preventing it from hitting the one. Notice that, if the five wouldn’t stop the trolley, you’d have to turn it. The one person would die regardless, in almost the same moment; there would be nothing you could do for him. Turning the trolley, you would at least save the five.

Is it permissible to turn the trolley? If you turn it, you’re exploiting the presence of the one to save the five. If you don’t turn it, you’re (passively) exploiting the presence of the five to enable the one to survive. From the point of view of the Doctrine of Double Effect — or any principle that discriminates against treating people as means in ways that lead to them suffering harm — it’s a wash. The remaining moral considerations are doing vs. allowing, 5 vs. 1, etc.

The upshot: suppose it is permissible to turn the trolley in the loop case. This is compatible with the view that it matters, in doing or allowing harm, whether one is using one’s victim (as a tool, as an exploitable opportunity, etc.). Contra, Thomson, the former is not a true counter-example to the latter.



— Quinn, W. (1989). “Actions, Intentions, and Consequences: The Doctrine of Double Effect,” Philosophy & Public Affairs, 18/4: 334-351.
— Thomson, J.J. (1985) ‘The Trolley Problem’, Yale Law Journal, 94: 1395–415.

17 Replies to “A Tangled Loop by Kerah Gordon-Solmon

  1. Hey Kerah! Thanks for this. Can you say more about what it means to passively exploit the five? Here’s what I can imagine someone saying in response: letting the trolley stay its course isn’t passively exploiting the five, because even passive exploitation requires more than simply letting an event unfold (especially when one did not have a hand in bringing about that event, when they don’t stand to benefit from the event unfolding etc.). If we think that we’re exploiting people simply by not intervening on such an event, then we’ve identified a sense of exploitation that’s far weaker than the one that’s usually operative in our notion of using people merely as means.

  2. Interesting! I’d always imagined the loop case more asymmetrically, i.e. such that without switching, it would never enter the one-person part of the loop at all (you can imagine an exit track just after the five, so their presence isn’t essential to protecting the one). This asymmetric variant still counts as a “true counterexample”, right?

  3. Hi Richard,

    We have something in common! For years, I pictured the loop case exactly as you describe it. But it always struck me as obviously wrong to turn the trolley in that case — I couldn’t see a relevant difference between it and the bridge variation. It baffled me that others had different intuitions.

    It was only after I belatedly read Thomson’s “The Trolley Problem” — and saw that her original loop is basically as I described it in the post — that I thought “THIS CHANGES EVERYTHING.”

  4. Hi Jordan,

    This is such a great question. I think passive exploitation enters the case in virtue of one’s being a morally motivated agent on the scene, who processes what’s going on and has the impulse to help.

    An indifferent observer — a nastier character — wouldn’t be passively exploiting the five, because, presumably, they’d be disinterested in what happened to any of the people on the track. The fact that the bodies of the five will stop the trolley and save the one plays no role in their practical deliberations, it doesn’t feature in their plans, etc.

    But the morally motived agent is different. Their reason for not the trolley (if they don’t) is that the bodies of the five will stop the trolley, protecting the one. It precisely parallels the reason they would have for turning the trolley, namely, that the body of the one would stop it, protecting the five.

    It’s a fascinating case in this respect, I think, that just in virtue of being a morally motivated agent on the scene, one’s intentions are implicated in an unsavoury (uncomfortable?) way.

  5. Ha, fair enough! I confess my intuitions there go more with Thomson. While I agree that it’s hard to see a “relevant difference” from the bridge variation, it seems even harder to see a relevant difference from the basic switch case. (Putative grounds for deontic constraints have always struck me as of dubious moral relevance, though. Why should the killed one care whether his death is instrumental to saving the five or incidental to it? If anything, I’d rather my death be useful…)

  6. Hello Kerah,

    Really interesting post. I agree that even if it’s permissible to turn in this loop case, it isn’t a counterexample to DDE for the reasons you give.

    Now consider the asymmetric loop case, where the killing of the five wouldn’t prevent the trolley looping around to the one, and so they wouldn’t be useful. I don’t have strong intuitions about that case. But suppose that it is permissible to turn from the five to the one. Is this an objection to DDE?

    I’m not sure. Here’s how we might reason about this case. I turn the trolley in order to prevent the five from being killed by the trolley going the way it’s going. A side-effect of doing this is that the one will be killed. But the fact that the one will stop the trolley is also useful in preventing the five from being killed in another way – by the trolley looping around.

    Friends of the DDE might argue: it is permissible to kill one as a side-effect of saving the 5 from being killed in one way; and the fact that it is worth saving the 5 from being killed in this way only because the side-effect killing will also be useful in saving the 5 from being killed in another way does not make this wrong.

    Here’s another way to say more or less the same thing. If I turn the trolley: 1) the 5 will be saved from being killed in one way without using the one. The death of the 1 is a side-effect of my saving the 5 from this threat. However, saving them from this threat gives rise to another threat, and the five will be saved from being killed by this new threat only because the 1 is useful.

    I am at least somewhat inclined to think this permissible, even though I’m a friend of the DDE.

    The looping, I think, makes the structure a bit harder to see. But think about this case:

    Tiger: I can turn a trolley away from 5 people towards one who will be killed. Doing this, though, will also release the door of a cage containing a tiger. The tiger would eat the 5, but the death of the one person will attract the zookeeper, who will immediately lock the cage.

    I find it pretty clear that it is permissible to save the 5 in this case, even though their being saved from the tiger relies on the death of the one.

  7. Hi Victor,

    I love this comment, even though I think it peaked in the first sentence.

    I share your intuitions about your tiger case, and am persuaded by your analysis of it. That’s awkward for me, because I don’t share you intuitions about your Asymmetric Loop variation. Like I said to Richard, it always seemed self-evidently wrong to me to turn the trolley in that variant, for DDE reasons.

    So let me try pushing back against the seeming symmetry between those two cases. In Tiger, the button that saves the five from the trolley also creates a new threat to them (the tiger). In your Asymmetric Loop, the trolley coming at the five from the front and from the back are not two discrete threats. It’s the same threat, redirected; absent the person on the loop, the redirection would be right back onto the same victims.

    I’m resistant to the thought that it’s licit to segregate the threats the trolley would pose from the two different directions when we analyze the case. I find it a lot more natural to categorize the threat as the runaway trolley, full stop.

    Another variation for you — what if we moved the loop up the track, so it’s:

    Trolley___________________ 5

    Here, again, it seems to me wrong to turn the trolley; your Asymmetric Loop seems much akin to this one than it does to Tiger.

  8. Oh, dear. This is what happens when one tries to type diagrams into a comment thread. Mangled! I was going for a stretch of main track, with a side track looping off and then back around somewhere in the middle, and the five trapped at the end of the line after the loop diverges and reconnects.

  9. Thanks Kerah,

    Like you, I think the intuitive view is that it is wrong to turn the trolley in your last case and that it is permissible in Tiger. These results would be explained by the difference between creating a new threat that the person is harmfully used to avert and the person being harmfully used to avert the same threat. And that view would also explain the fact that the asymmetrical loop case divides opinion – it isn’t clear whether there is a new threat or the same threat in that case, and our different judgements about that might then result in different intuitions about permissibility. More work would have to be done, though, to show that this view not only has intuitive implications, but that it can be vindicated.

    One thing that is clear, though, is that alternative explanations for the intuition (insofar as people have it) that it is permissible to turn the trolley in both loop cases suggest that these cases don’t cast doubt on the DDE. We can find an explanation for these intuitions that is consistent with the importance of the distinction between harmful using and harming as a side-effect.

  10. Love the post + thread! I agree with a lot of things in the comments here. So, rather than rehash, I thought I might add some Thomson trivia.

    Thomson, as you may know, has a neat test for when you are “using” someone as a mere means. I often hear it glossed as “you use someone if you couldn’t have succeeded without their presence.” But Thomson actually uses an “only if.”

    “If the agent chooses to engage in the course of action, then he uses the one to save the five only if, had the one gone out of existence just before the agent started, the agent would have been unable to save the five.” (P. 1402,

    I think Thomson says “only if” instead of “if” precisely because of the lollypop loop case. She doesn’t want to say that you (passively) use the five to save the one if you do nothing. Although it’s true that, without the one, turning the trolley wouldn’t save the five…

    “It is also true that if the five go wholly out of existence just before the agent starts to turn the trolley, then the one will die whatever the agent does. Should we say, then, that the agent uses one to save five if he acts, *and* uses five to save one if he does not act? No: What follows *and* is false. If the agent does not act, he uses nobody. (I doubt that it can even be said that if he does not act, he lets
    them *be used*. For what is the active for which this is passive? Who or what would be using them if he does not act?).” (P. 1403 in the 1985 paper, fn. 8, italics replaced with asterisks.)

    Food for thought!

  11. Hi Daniel,

    Thanks for that.

    Thomson’s test can’t be right, though – counterfactual dependence of success on a person’s presence isn’t required for using in pre-emption cases.

    Suppose that you are on a flipper on a bridge. I press a button flipping you in front of the trolley, stopping it to save five. I’ve used you. Now suppose that had you gone out of existence, a moose would have wandered onto the flipper. I’ve still used you. But my saving the five didn’t counterfactually depend on your not being present. (There was a bit of literature on this a while back – I think the moose case might have been Alec Walen’s, though not sure).

    Intentional causal involvement in the success of my plan, or something like that, is probably a better way of understanding using, but causation doesn’t require counterfactual dependence.

    It’s also worth noting that many people think that any principle against using should cover passive using to explain Quinn’s Guinea Pig case, where I allow a person to die in order that others can learn about a disease they are suffering from. Many people think that this is wrong, and have the intuition that it is wrong for the same reason it would be wrong to harmfully experiment on a person to save others.

    I tend to think that passive using is not as bad as active using though (if you’re really interested, I can send you a ref), and that might play a role in these loop cases.

  12. Hi Victor and Daniel,

    Victor, I’m well pleased — I 100% agree with everything you said in your last two comments!

    Dan, you always bring the best references. In response to Thomson’s footnote — alongside what Victor said — I would repeat what I said to Jordan upthread. Thomson’s footnote anticipates Jordan’s comment (or is it the other way around?).

  13. Victor, thanks for those great cases. I agree with you on both counts (assuming you agree with the majority about Quinn’s case—would love to see your reference on that).

    As for the worries about preemption, I’m sure we can just borrow a simple, counterexample-free solution from the causation literature…….

  14. Hi Daniel

    In To Do, To Die To reason Why (currently available in all good bookshops – movie unfortunately delayed due to COVID) at p.97 I offer this pair of cases:

    Distract: A murderous psychopath is about to attack Gertie. She can save her life only by altering him to the five people hiding in the cupboard. He will then kill the five rather than her. If killing Gertie distracts the psychopath, he will not find the five in the cupboard for that reason. If killing the five distracts him, he will not kill Gertie for that reason.

    Hide: A murderous psychopath is about to attack Gertie. She can hide under the bed upstairs. If she does so, he will find the five people who are hiding in the cupboard and kill them. If killing Gertie distracts the psychopath, he will not find them in the cupboard for that reason. If killing the five distracts the him, he will not kill Gertie for that reason.

    I suggest that the intuitive view is that Gertie is not permitted to point to the five in the cupboard in Distract, but that she is permitted to hide in hide. That is so even though the five would be useful to her in both cases in preventing the psychopath from killing her. That suggests that active using is harder to justify than passive using.

    Similar reasoning might lead us to think that it is wrong to turn the trolley in Kerah’s original loop case, though for reasons I offered above, the case has additional features that I think make turning permissible.

  15. Hi Victor,

    Oh, this is fascinating. I admit, I haven’t read your book, yet (I was holding out for the movie; those casting rumours…!) but I’ll bump it to the top of my queue.

    When you said upthread that active using is harder to justify than passive using I thought: ‘Yes, harming is harder to justify than allowing harm; likewise, instrumental harming is harder to justify than instrumental allowing harm. At least, that’s true when other things are equal. In the lollipop loop, there’s also the difference in numbers, which plausibly offsets the former difference. The lollipop loop parallels the classic trolley in that respect.’

    But I’m seeing now that you might have intended something different, namely, that active vs. passive using is — well — distinct — from other nonconsequentialist distinctions. Is that right? Certainly, in your pair, we have active vs. passive using, but no killing vs. letting die.

    I absolutely share your intuitions about your case pair. But I’m still inclined to see ‘using’ as augmenting, or amplifying, the pro tanto wrongs of things like harming, allowing harm, and etc. Gertie’s behaviour in Distract goes beyond merely allowing harm; it amounts to collaborating in the murder of the five. Nonconsequentialists would frown harder at collaborating in harm than at (merely) allowing harm. We would keep frowning even if it weren’t the psychopath’s killing the five that saved Gertie, but her convincing him that she was on his side by answering his questions (unfortunately, about the location of the five) honestly.

    I think it also makes a difference in Hide that Gertie is both the agent and the one who’s life is pitted against the five. People generally have greater license to favour themselves when doing so involves allowing harm to others. But that license doesn’t extend to third parties. If Gertie were unconscious, and you had the choice of whether to hide her from the psychopath, or… (etc. etc.), I think it would be wrong of you to hide her, knowing the implications for the five.

    It feels very gauche of me to respond to the case before doing the reading. An anticipatory 🤦🏻‍♀️in case the book explains why all of this is wrong!

  16. Hi Kerah

    Thanks for that. Yes, I more or less agree with what you say here. I use these cases to illustrate Doing and Allowing. And I agree that Gertie harms (or its moral equivalent) in Distract but not in Hide (the fact that we agree so much is heartening for me, but should surely be dispiriting to you!).

    One reason to talk about active and passive using, though, is that the principle against using applies not only to cases involving harm, but also to harmless rights infringements.

Comments are closed.