Unpublished Thoughts: Thomson’s Trolley-problem Problem

Welcome to the first in a new series of posts: Unpublished Thoughts. Today, I’ll take up a significant problem in Judith Jarvis Thomson’s (2008) “Turning the Trolley.” She offers an argument intended to demonstrate the that Trolley Problem is in fact a non-problem. Her argument, however, is logically invalid. For anyone perhaps unfamiliar with the Trolley Problem, it stems from the following story. A trolley (say on the hills of San Francisco) is careening out of control down a hill. At the bottom of the hill on the tracks are five track workers who will certainly be killed. (Imagine they can’t see and move out of the way in time.) A bystander sees all this and notices a switch next them that, if pulled, will divert the run-away trolley onto a different track, saving those five people. Alas, however, that there is one different track worker on this alternate track that will be killed if the bystander pulls the switch (to the right). People typically report that if would permissible for the bystander to flip the switch (c.f., Waldmann and Dieterich 2007, Cushman et al. 2006). The problem is that this answer runs contrary to the deontological view that killing is worse than letting die. (There are a number of other problems, such as inconsistent answers on similar situations like the Fat Man vignette, but I’ll leave those aside for now.) Thomson wants to argue that it actually isn’t true that the bystander may choose to kill the one worker to save the five. She intends to demonstrate that “the (so-called) trolley problem is a nonproblem.” Her argument for that conclusion, however, does not work. To make her case, Thomson modifies the bystander story, adding a third option. In this variant, the switch has three positions (instead of the orignal two), and so there are three options:

  1. The bystander to do nothing, letting the five people die.
  2. The bystander to pull the switch to the right; the trolley diverts to a second track with the one worker, killing him.
  3. The bystander to pull it to the left, diverting the train onto a track that the bystander himself or herself is standing on, thereby killing the bystander.

To resolve this three-option case, Thomson offers two principles. Let’s call the first principle P1 (her third in the original article), which stipulates that when an option to sacrifice oneself exists to save the five, one must not instead kill someone ele to save the five. Basically, if (3) exists as an option, then the bystander must not choose (2). So, with (2) eliminated. There remains only (1) and (3) as live (morally permissible) options for the bystander. Let’s say we accept Principle 1 and Principle 2 for the sake of argument. (I always have at least one student object to Principle 1 and remain unconvinced.) Thomson doesn’t offer an argument, but it seems reasonable, so we’ll take it. By itself P1 doesn’t tell us what is permissible or obligatory for the bystander. To finish that job, Thomson offers a second principle (it’s her fourth in the article, but we’ll call it P2). P2 states that if the only permissible option of saving five is to kill oneself, one may let the five die. Thomson then immediately returns to the original, two-option version of the bystander vignette. She imagines that a bystander might initially not know if the third option exists or not, but nevertheless recognize that (2) is impermissible if (3) is a genuine option (by P1). Then the bystander decides he is unwilling to kill himself to save the five. After so deciding, the bystander looks at the switch carefully (why only after?), and realizes that (3) isn’t an option because the switch can’t go to the left; (3) isn’t really an option. All the same, according to Thomson, “Since he wouldn’t himself pay the cost of his good deed if he could pay it, there is no way in which he can decently regard himself as entitled to make someone else pay it.” Let’s assume that Thomson is actually making a normative claim here, instead of the apparent predictive claim. Assuming that’s the case, then there is a logical fallacy here. Thomson’s argument is that in (1) is permissible because (2) was ruled out. But (2) was only found to be impermissible so long as (3) is actually an option, not if it might be an option. P1 was stated as a conditional. In the original, two-options version, there was no third option. So in the Trolley Problem (the real one with two options), the condition under which we can eliminate (2) don’t pertain. (2) remains a viable option. Thomson wants to pretend that because (2) might be impermissible, therefore it is. There’s another option. It might be that Thomson doesn’t mean to present a logical argument here. Rather she might be presenting an intuition that regardless of whether or not it is a real possibility, we must consider whether any sacrifice we’d force others to make, we’d be willing to make ourselves instead. If not, then we can’t force someone else to do so (even if we can’t). While such a golden-rule type of intuition isn’t necessarily wrong, it might be nice to have more than the Thomson’s intuition in support of it. Without something more, the Trolley Problem looks just as problematic as ever.