Another interesting and weird paper (to me, anyway
:
http://csjarchive.cogsci.rpi.edu/Proceedings/2010/papers/0626/paper0626.pdf
In experiment 1, there are two situations (people on the trains can't affect outcomes):
I. Train C – no one on board- is going to hit train A – 5 people on board -, unless it's diverted into another track, where it will hit train B – 1 person on B.
II. Train C – no one on board - is going to hit train A – 5 people on board -, unless B – 1 person on board - is diverted into its path.
Weirdly, the experimenters contend that II is a case in which a person is used as a means, and so that the Doctrine of Double Effect correctly predicts a difference in patterns of judgments, so the judgment that diverting the train in a. is acceptable will be more common than the judgment that diverting the train is acceptable in II – which is observed.
But that's false. DDE does not make that prediction. The person on train B is never used as a means to an end. What is used as a means to an end in II. is the train.
To be fair, when discussing experiment 3, the researchers seem to realize that the person who dies is not actually used as a means to an end, though they seem to doubt that people (other) would construe it in that matter (i.e., they seem to believe “people” will believe that that's a means to an end).
Experiment 3:
3.I. There is 1 person on B, 5 on A, and B is going to hit A killing the 5, but not the one on B, who is in the back of the train. C is redirected to hit B on the back, killing one but saving 5.
3.II. There is 1 person on B, 5 on A, and B is going to hit A killing the 5, but not the one on B, who is in the back of the train. C hits B so that the passenger is pushed towards the brake system, and that kills the passenger but stops B.
3.III. C – no passengers – is going to hit A. B – 1 passenger – is diverted to stop it. Acceptability ratings are very low.
3.IV. There is one person on D, which is not moving. The intervention sends empty train C on a collision course with D. That kills the passenger, but allows C to go on and hit empty train B, which is thus preventing from hitting train A, and killing its five passengers.
Comparing all four conditions, acceptability ratings are the highest and II, slightly lower in I, lower in IV and still lower in III.
Both results and interpretation are weird to me.
Apart from their “fairly high” and “fairly low” assessments (even III got 3.76 on a 1-6 scale asking whether a person should carry out the intervention; that question is also vague, complicating interpretation to some extent, but that sort of problem seems common in trolley experiments), researchers say that both conditions I and II refute DDE, because in both cases, the person on B is used as a means to an end, but in both cases, there is similarly high acceptability.
While it's true that the results are evidence against DDE (as an account of the observed patterns, regardless of whether it's true), it's not true that in both cases, a person is used as a means to an end. Only in condition II, a person is used like that.
In 3.III, the researchers count that as a person's being used as a means to an end. But that's not true, and I don't know that the mistake will be common.