In the previous post, I described Huebner and colleagues’ study of the factors influencing moral judgments. I presented a number of the moral scenarios they used as probes, and also reported their results on how people judge these scenarios, what these patterns of judgment tell us about the separate elements that go into a moral judgment, and how they interact with each other. At the end of the post, however, I alluded to some issues I had with the scenarios that were used. And I’d like to expand on these issues here.
In my original post I focused on the set of ‘boxcar dilemmas’ that the trio used, and didn’t describe the analogous scenarios they came up with involving wrecking balls, street gangs, murderous tribes, or burning houses. (You can see them all here if you like.) So in this post I’m going to keep the focus on the boxcar scenarios, but I think that the concerns I raise about the boxcar vignettes apply equally to the other scenarios (though I think they’re particularly clear in the boxcar cases).
OK, so what are the issues I have in mind? Let’s take the boxcar scenarios that do not allow for Pareto improvements, and contrast these with those that do. The canonical non-Pareto case is the original footbridge dilemma : should Matt push the man off the bridge to stop the train? The Pareto counterpart to this is whether Dave should push the trapped man to make him scream and thus alert the imperilled five . (Both  and  involve direct contact, and harm as a means to an end: saving the lives of the five walkers.) In addition, there’s the mediated Pareto improvement case , which is like  except Steve can hurl a rock at the trapped man instead of shoving him.
Ideally, to make comparisons between the non-Pareto and Pareto cases meaningful, everything should remain the same except for the possibility of a Pareto improvement. That means that in Pareto harms should not only involve the same physical actions as the non-Pareto cases, but also inflict the same harm. But this doesn’t seem to apply in the cases used in this study.
The fundamental problem is that the harms inflicted in these cases are dramatically different. In the footbridge dilemma , the harm inflicted on the man is not simply whatever pain or discomfort is caused at the moment your hand makes contact with him — it’s the harm that results from the following causal sequence of events, which is the man being dead on the tracks. The harm inflicted on him is his death.
Now, in the case of the trapped man , the scenario says that if Dave “pushes the one person, he will remain stuck on the track but will scream”. That is, pushing him doesn’t put him on the tracks, or even keep him on the tracks: he’s there already, and he is going to die whatever Dave does. But what does Dave do? He gives him a shove, causing the trapped man to yelp, perhaps as you might if someone barged into you in the street. The harm inflicted is very minor, and presumably morally comparable to walking up to the trapped man and giving him a charley horse (or ‘dead leg’, as we call it in the UK). And that level of harm is drastically less than pushing someone to their death.
So it seems plausible that the different pattern of judgments in these scenarios primarily reflects this fact. One involves merely pushing someone; the other, pushing someone to his death. It is little wonder that the majority of people say the former is permissible, while a majority say the latter is impermissible. If so, doesn’t this cofound any inferences drawn from comparisons of the patterns of moral judgments generated by these Pareto and non-Pareto cases?
It’s noteworthy that in the crying baby dilemma , the harm the baby will suffer if discovered is death, and the harm that is inflicted on the baby to achieve the Pareto improvement is also death. But this is true of the Pareto improvement versions of the footbridge dielmam ( and ). To make a Pareto case out of the footbridge dilemma that can be directly compared with the original footbridge dilemma  the harm inflicted should be the same i.e., death? And they’re not.
It’s tricky to come up with possible scenarios that meet these criteria, but here’s a far-fetched stab:
“There’s a large man standing on a bridge. Dan happens to know that in 2 minutes time an assassin perched on a nearby hill will shoot the man in the head, killing him instantly. At the same time, an out-of-control boxcar is bearing down the track and will hit 5 people unless Dan pushes the man in front of the boxcar in the next 15 seconds, as it passes under the bridge. If Dan does nothing, the 5 people will be killed by the boxcar, followed by the man on the bridge shortly after when the bullet enters his brain. If Dan pushes the man, it will kill him but stop the boxcar, and save the 5. Is it permissible for Dan to push the man?”
Might that scenario generate different responses than either the original footbridge dilemma, or the trapped-man variant devised by Huebner et al.?
There are others things I find puzzling about the vignettes used by Huebner et al., and the judgments they elicited. If the only harm inflicted by Dave in the trapped man case  is a relatively harmless shove, as I’m saying, then why do 30% of anyone object to doing that, given that it saves life? I don’t pretend to have the answer, but I personally find the case of Dave pushing the man quite odd to think about. It’s not explicitly stated that the only reason to push the man is to generate a scream, nor that this is the only effect pushing him will have. As such, I think there’s an ambiguity about what’s going on when Dave pushes the man, and perhaps it seems like he’s somehow more causally involved in the trapped man’s death, even though it is explicitly stated that regardless of what Dave does, the trapped man is done for. But I don’t know.
It’s curious — to me, at least — why even more people thought it was OK to throw a rock at the trapped man  than to merely give him a shove  (89% versus 70%). Here’s why I think this is a little odd. In both cases, the goal is to inflict sufficient pain on the person to elicit a scream (at least that’s my reading of the scenarios). Yet isn’t it worse to throw rocks at people than to shove them? What would you prefer to be on the receiving end of? While directly caused harms might generally be seen as worse than those achieved through indirect means, I would’ve imagined that most people would judge hurling rocks at someone as worse than pushing them. (Obviously, if the person was perched on the edge of a cliff, this would be a different matter). The fact that more people found pushing the man to be objectionable than hitting him with a rock again suggests that there was something confusing about the shoving case. (Just for completeness, it’s easy enough to think of a mediated-harm counterpart to my direct-contact assassin case: instead of being in a position to push the man in front of the boxcar and to his death, Dan can throw a rock at him, knocking off the footbridge and in front of the train.)
I think the problems I’ve alluded to in these moral scenarios create some difficulties in using the judgments they elicited as evidence for what drives those judgments. I also think that they beset the other scenarios Huebner et al. devised (and which I didn’t describe) to explore the factors that influence moral judgment. That’s not to say I find the substantive conclusions implausible; far from it. I just think these specific claims could benefit from some confirmatory studies that use less troublesome dilemmas.