Takeaways From the First Section of The Trolley Problems

Kamm is mostly right, though wrong in various places

I’m currently reading the book The Trolley Problems, a compilation of various scholars works on the trolley problem. Francis Kamm has a rather devastating critical treatment of Thompson’s proposal, though perhaps it will be revealed to be less decisive when I read Thompson’s proposal—one which I have not yet read.

Thompson claims that one ought to, in the version of the trolley problem where they’re the driver and can turn the trolley so that one dies rather than five, turn the trolley, but that one ought not, as a bystander, flip a switch to turn a trolley. The explanation for this is given by the allegedly salient distinction between killing and letting die—the bystander would kill one if they act, while only let five die if they don’t. In contrast, the trolley driver would kill five if they don’t act, while they’d kill one if they do act.

Kamm has a few replies, not all of which I will discuss. First, she claims that it’s too permissive. Suppose that the trolley driver can use some device to knock a fat man off a bridge in front of the trolley that they’re driving that will otherwise save five. It seems that they ought not do that—after all, that case, called Topple, seems structurally analogous to the original case where one can push another off a bridge to prevent five deaths, an action prohibited by both standard deontology and common sense. Kamm’s piece is not intended to address the more revisionary implications that I’ve argued for elsewhere—it takes common sense judgments about trolley problems as a given.

Second, this seems to be, in other cases, not permissive enough. Most people have the intuition that the bystander should flip the switch. Similarly, it requires accepting the unintuitive result that I highlighted in my most recent article.

There’s another reason, one not raised by Kamm, to think that spectators should turn the switch, that doesn’t require merely an appeal to initial intuition. Consider the following case

Bystander + Driver Cooperation: The trolley can only be turned if both the driver turns the steering wheel and the bystander flips the switch. The driver has already turned the steering wheel. Should the bystander do so?

It seems like if the bystander should, in Bystander + Driver Cooperation, flip the switch if and only if they should flip the switch in Bystander. But it seems obvious that in Bystander + Driver Cooperation, they should flip the switch. This is supported by the following principle.

Enablement: if some action would enable another to do the right thing, all else equal, you should do it.

In this case, the bystander throwing the switch allows the driver to turn the trolley. But the driver should turn the trolley. Therefore, by Enablement, this is what they should do.

Thompson has an interesting argument—one which Kamm, in my view decisively, refutes. Suppose that the bystander has two options.

1. Flip the switch to the left, killing a bystander.

2. Flip the switch to the left, killing themselves.

3. Not flip the switch at all, thereby allowing the trolley to continue and kill five people.

Thompson argues that it would be permissible to either do three or two, but not one. One is impermissible because it is strictly less permissible than two. She appeals to a principle, according to which if some act can impose a cost on oneself or a third party, it’s impermissible to impose it on the third party (E.g. if you can either sacrifice your own hand to save the world or cut off someone else’s hand, it would be impermissible to cut of their hand). This principle doesn’t strike me as ultimately plausible, but I think I’d be in the minority about that.

So, Thompson argues that in this case, 2 and 3 are both permissible—3 is permissible because you’re not obligated to sacrifice yourself to save five. However, she implicitly appeals to the following principle

Permissibility combination: If A and B are permissible and C is impermissible when they’re all options, then if only B and C are options, B is impermissible.

This principle is, I think false, on plausible views of permissibility. To illustrate this, suppose that you can either donate to a charity that can save lives for 5,000 dollars, one that would save lives for 10,000 dollars, or not donate to charity at all. Suppose the donor is very poor, such that donating is not obligatory. In this case, it seems permissible to donate to the efficient charity or not donate but impermissible to donate to the less efficient charity. However, if you can only donate to the charity that saves lives for 10,000 dollars or not donate at all, it seems permissible to donate to the charity that save lives for 10,000 dollars.

This is because permissibility covers a wide range of cases. A and B can both be permissible and A and C can both be permissible, even if B is better than C. This principle may run into some moral problems, but this is the way that permissibility is widely conceived to work.

However, Kamm gives a counterexample to this principle which is not, I think, a genuine counterexample. In fact, I think that it serves to challenge the distinction between Bridge and Switch. In the example, a person is standing on a bridge, and one can flip the switch to redirect the train to kill them on the bridge, or push them off the bridge to stop the train, wherein they’ll just be paralyzed. In this case, she thinks you should push the man—after all, that just leaves him better off and no one worse off—but this doesn’t mean that you shouldn’t push the person in the ordinary bridge case, where the person will be paralyzed or that you shouldn’t flip the switch.

But I think it does mean that. Permissibility is not a property that obeys completeness. A property is complete if, between any two things, either one is more of that thing, one is less, or they are equal in that regard. Sacrificing oneself and allowing five to die are neither more, less, nor equally permissible—both are permissible if one is choosing between them. Permissibility doesn’t come in precise degrees.

However, whether some action is preferrable to another does seem to come in degrees. If A is preferrable to B and B is preferrable to C, then it seems that A is preferrable to C. If this is correct, this means that Kamm’s alleged counterexample fails—and the arguments I have made here do not fail because of Kamm’s considerations.