New studies on moral judgement delve into the psychological complexity of dealing with dilemmas
[Note: This will be a two-part post. In the first part, I simply want to describe a new paper, including details of how the study was done and its results; in the second part, I raise some questions about the moral scenarios used in this study]
Some moral prohibitions, such as those against harming children or brutalising and killing innocent people, are blandly uncontroversial. We feel the intrinsic force of these proscriptions so deeply that they have a self-evident quality, and rarely need explicit argumentative defence (among regular folk at least – psychopaths are notable for their lack of concern about violating these moral codes). Yet many people, if not most, would agree that in some cases harming or even killing someone may be morally permissible (although probably not obligatory). Take the following scenario (I’ll be numbering these so I can refer to them later):
 Enemy soldiers have taken over your village. They have orders to kill all remaining civilians. You and some of your townspeople have sought refuge in the cellar of a large house. Outside you hear the voices of soldiers who have come to search the house for valuables. Your baby begins to cry loudly. You cover his mouth to block the sound. If you remove your hand from his mouth his crying will summon the attention of the soldiers who will kill you, your child, and the others hiding out in the cellar. To save yourself and the others you must smother your child to death. Would you smother your child in order to save yourself and the other townspeople?
Few things could be more abominable than killing a baby. Yet in this case there’s a compelling logic for doing so. Your baby, if he continues crying, is going to be killed, along with all the townspeople. If you smother your baby he will die, but everyone else will live. So despite the awfulness of smothering the baby, you’re not making his future any worse (that’s a very cold sentence to write!). Obviously this logic isn’t enough to make your choice a no-brainer — just imagine what it takes to hold your hand over a baby’s mouth until he stops breathing — which is why this is a moral dilemma and not simply a maths puzzle.
Dilemmas such as this — and philosophers have developed many subtle variants of this kind of scenario — suggest that harming or even killing other people isn’t always morally outlawed as a means to an end: it depends on whether the harm inflicted on other people makes them worse off than they would have been anyway (and, of course, whether it benefits other people).
In a new paper published in Mind & Language, Bryce Huebner, Marc Hauser and Phillip Pettit have explored in depth how such considerations factor into moral judgements — and also how they interact with other features of moral dilemmas (Ref. 1). (You can read a pre-print of the article here.)
As a first step, Huebner et al. re-analysed data collected in earlier studies by Joshua Greene and colleagues on a wide range of moral dilemmas that involve inflicting harm on another person. (All the dilemmas used in this latest study are available here.) In the cases that Huebner et al. examined the harm inflicted on a person is done so a means to achieving some other end — namely, saving the lives of other people. Take the well-known ‘footbridge dilemma’:
 Matt is standing near to the railroad tracks when he notices an empty boxcar rolling out of control. It is moving so fast that anyone it hits will die. Ahead on the main track are five people. There is one person standing next to Matt who is wearing a heavy backpack. If Matt does nothing, the boxcar will hit the five people on the main track, but not the one person next to him. If Matt pushes the one person next to him on to the tracks, the boxcar will hit this person and then slow down and stop before the five people on the main track. Is it morally permissible for Matt to push the person?
Now this one feels different from the crying baby case, and probably less of a dilemma. If you’re like the majority of people, you’ll think pushing the man is morally unacceptable (just 27% of people in Huebner et al.’s new study said it was OK for Matt to push the man, a result in line with previous studies). In the footbridge dilemma, Matt clearly making the man worse off than he would have been (if Matt leaves him alone, he lives happily ever after; if Matt pushes him, he dies). In the crying baby dilemma, this isn’t so. The baby’s fate is sealed not matter what we do, but this isn’t true of the townspeople, who can be saved. Philosophers call the option of smothering the baby who will die anyway to save the townspeople a ‘Pareto improvement’ (again, this sounds ghastly in this context). This essentially amounts to making the best of an already bad situation. (I only introduce this terminology as it makes the following description of Huebner et al.’s studies a bit easier.)
Huebner and colleagues wanted to explore when and why harming people in the pursuit of other moral goals is acceptable, and what factors affect these judgments. So they looked at responses to a range of dilemmas: some allowed for Pareto improvements (helping others by harming someone who is going to be harmed no matter what), while in others helping other people required harming or killing people who would otherwise remain unharmed. (Harming someone to create a Pareto improvement is called a Pareto harm.) Their results suggest that, in general, people are significantly more likely to judge actions that involve Pareto harms as more permissible than those causing non-Pareto harms. Translated a bit, harming someone who would already be harmed in order to save the lives of other people is generally seen as more morally acceptable than harming someone who would otherwise be free from harm to achieve the same end).
Earlier studies have shown that people are not only sensitive to whether harm is inflicted on other people in order to achieve a greater good, but the means by which that harm is inflicted. Harms caused by direct physical contact and ‘personal force’ (the use of muscular energy to harm someone, such as pushing the man on the bridge with hands or a rigid pole) are typically judged as less permissible than those that are achieved through more remote means — say, flipping a switch to cause the man on the footbridge to fall through a trap door onto the tracks (Ref. 2; you can read this paper on ‘pushing moral buttons’ here.)
To probe whether different ways of inflicting harm matter in Pareto improvement cases, Huebner et al. needed some scenarios similar to the footbridge dilemma that not only involved Pareto harms, but which could also be varied in how the harm was caused. Remember that in the footbridge dilemma , the harm is non-Pareto (the man on the bridge will be harm-free if left alone), and is inflicted through direct contact. Here’s the ‘Pareto improvement’ variant of the footbridge dilemma that Huebner and colleagues came up with, which also involves a harm achieved through direct contact:
 Dave is standing near to the railroad tracks when he notices an empty boxcar rolling out of control. It is moving so fast that anyone it hits will die. Immediately ahead on the main track is a person whose foot is stuck in the tracks such that it is impossible to get him out of the way before the boxcar hits him; further ahead of this one person are five more people walking on the tracks. If Dave does nothing, the boxcar will first hit the person whose foot is stuck in the tracks and then the five other people on the main track. If Dave pushes the one person, he will remain stuck on the track but will scream, thereby alerting the five people to the boxcar; the boxcar will hit the one person, but not the five who will move off the main track. Is it morally permissible for Dave to push the one person?
In this case, 70% of respondents said it was morally permissible for Dave to push the person. Remember, the person on the tracks is going to be killed whatever happens. Dave simply inflicts harm (a push) on the person to elicit a scream, thereby alerting the other people in peril. (We can, for the sake of the hypothetical scenario, assume that there is some reason why Dave can’t simply shout out a warning himself.) That is, in contrast to the original footbridge dilemma , this scenario offers scope for Pareto improvement. And this difference is reflected in the permissibility ratings they typically receive: whereas just 27% of people say it’s OK to push the man in the footbridge dilemma , 70% do in Dave’s case . This suggests that Pareto considerations play a big role in our moral judgments.
So far, so straightforward. But does the moral significance of Pareto considerations differ when the harm is brought about through mediated contact? Here’s the scenario Huebner et al. used for to explore this question:
 Steve is standing near to the railroad tracks when he notices an empty boxcar rolling out of control. It is moving so fast that anyone it hits will die. Immediately ahead on the main track is one person, and then further ahead, five more people. If Steve does nothing, the boxcar will first hit the one person and then the five other people on the main track. If Steve throws a rock and hits the one person, this person will scream, thereby alerting the five people to the boxcar; the boxcar will hit the one person, but not the five who will move off the main track. Is it morally permissible for Steve to throw the rock and hit the one person?
Now a full 89% of people said it was OK for Steve to throw the rock at the person (compared with 70% of people who judged it morally acceptable for Dave to push the person in the direct contact vignette ). This suggests that while the possibility of Pareto improvements generally increases the moral permissibility of inflicting harms on people as a means to an end (saving other lives), such considerations interact with the means by which the harms are achieved — namely, whether they are brought about through direct or mediated contact.
Huebner et al. didn’t stop there. They developed even more scenarios in which an out-of-control boxcar is on course to kill five people. In one (scenario 5), this could be prevented by pushing a dead man in front of the boxcar (this represents a Pareto improvement, as a dead man’s fate cannot be made much worse! Correspondingly, 85% of people said it was OK to use the dead body as a brake). In another similar scenario  the man was terminally ill rather than flat dead. (This was not considered a Pareto improvement, as the dying man would become dead and thus worse off; 65% of people said it was OK to push the dying man in front of the boxcar, compared with 27% off the bridge.) In a final twist on the boxcar dilemmas, the question was whether it would be permissible to trick someone into wandering on the tracks and in front of the train — in this case, by saying there’s a $100 dollar bill there waiting to be picked up (scenario 7). People judged psychologically coercing the man into harm’s way as a means to saving lives much like the original footbridge dilemma, in which the choice is to physically push the man: just 20% of people said this trickery was permissible.
It’s time to take stock. Let’s start with the three scenarios that offered the chance of Pareto improvement: scenarios  (trapped/push/yelp),  (trapped/rock/yelp) and  (dead body). These generated permissibility ratings of 70%, 89% and 85%, respectively, and so the average permissibility rating across the Pareto-improvement dilemmas is 81%. By contrast, the average rating for the three scenarios in which Pareto improvements were not possible — original footbridge [2; 27%], dying man [6; 65%], and trickery [7; 20%] — was 37%. This suggests that Pareto harms are much more permissible than other types of harm. At the same time, not all Pareto harms are considered equal: when the harm required to bring about the better outcome is achieved through direct physical contact (i.e., pushing), it’s judged as less permissible than when the harm is mediated by something else (i.e., throwing a rock): to wit, while 89% of people said it was permissible for Steve to throw the rock , only 70% said the same for Dave pushing the person . To reiterate: harms that generate Pareto improvements (Pareto harms) are more permissible than those that do not; and Pareto harms that require direct physical contact are less permissible than those inflicted through indirect means.
Through another round of studies with even more dilemmas, Huebner et al. explored another question: whether the source of threat in the moral scenarios — the thing from which 5 people need saving — is another person, a natural hazard such as a fire, or something mechanical, like an out-of-control wrecking ball. (All of these different threats were weaved into stories that were intended to structurally mirror the boxcar examples; you can see them here.) The judgments in these cases suggest that people are sensitive to the source of threats: in particular, people judge Pareto harms as more permissible when the threat is non-mechanical in nature (e.g., another person).
All told, these are intriguing findings. They paint a rich and complex picture of the way the mind breaks down and represents the component aspects of moral dilemmas: Why are these people in peril, and what’s the threat? If lives can be saved by harming someone else, is that person being made worse off than they otherwise would be? What does harming this person involve? Something up-close and personal, or something more removed?
The answers we give to moral conundrums reflect how we integrate these strands into a final judgment of permissibility. Perhaps this feat of moral calculation is in part carried out by an innate ‘moral grammar’ comprising principles that specify how and when morally relevant distinctions — such as acts versus omissions; means versus side effects; the use of direct, personal force or more remote, mediated means to inflict harm — come into play. (The moral grammar idea is a big topic for another day!) Others might prefer to explain these results as the outcome of an internal dialogue of sorts: one the one side, intuitive feelings elicited by thinking about inflicting certain kinds of harms on other human beings in certain kinds of ways; and on the other, reasoned deliberation about the consequences and outcomes of action or inaction, and the moral weight we attach to each.
Regardless of how we explain these results, it’s worth considering how significant these results are. While I appreciate the ingenuity of the dilemmas used in this study, I’m a bit dubious about them — and therefore unsure of what these findings really tell us. It’s not that I find the conclusions generally implausible. But I do think the scenarios Huebner et al. used have problems that make inferences from judgments about them difficult. That’s the topic of the next post.
1. Huebner, B., Hauser, M. D. & Pettit, P. How the source, inevitability and means of bringing about harm interact in folk-moral judgments. Mind & Language 26, 210–233 (2011).
2. Greene, J. D. et al. Pushing moral buttons: the interaction between personal force and intention in moral judgment. Cognition 11, 364–371 (2009).