Blog Editor


  • Copyright 2005-20012 by Adam Kolber
    All rights reserved.

« Australian Neurolaw Database Launch Event, Dec. 8, 2015 | Main | PEBS Neuroethics Roundup (JHU) »


I would humbly suggest that your critique is not just hating on trolleyology "just like a philosopher would", but rather raises important critiques that anyone steeped in the scientific method should be attentive to. The whole idea of doing experiments is to understand how the world works - well, OK, maybe that is just one reason for doing experiments, but it is and should be the primary objective. If the trolley vignette is unrealistic, or lacks ecological validity, it may or may not get at the real moral intuitions that real people have in the real world. It is a general problem in the design of contrastive vignettes, and one that my research group worries over a great deal [ ].

There is a (partial) solution that can be offered: late in the experiment, one can ask respondents how realistic they felt the vignette was, using those answers both to get a sense of what respondents "think they think" about the realistic nature of the matter at hand as well as insuring that they don't consider one arm of the trolley problem more realistic than another.

But the fundamental critique here is pretty much spot on.


I think your argument raises a significant challenge, though I'm not sure I'm getting why you think the weird features taint the contrast too.

One reason I suspect the contrast might be tainted (maybe this is the sort of thing you have in mind?) in weird cases such as those is that - for example - the weird features of scenario A might affect people's judgments in a manner that's different from the manner in which the weird features of scenario B affect them, and yet scenario A and scenario B are compared in order to learn about contrasts.

For example, the common "switch" problem is often compared with the "fat man" problem. Both problems require that the person making the assessment assume that if she doesn't push the button, five people will certainly die.
This might be somewhat problematic because - among other issues - when people try to imagine the scenarios, perhaps some (many, most, etc.) can't help but make an unconscious intuitive probabilistic assessment, assigning a probability that is high but significantly less than 1 to the hypothesis that the five people will die, even if they imagine the five people to be unconscious, tied up, etc. But at least, that's similar in both scenarios.

However, the scenarios differ in that the "fat man" scenario also requires that one assumes that the fat man will certainly stop the trolley (which might strike most people as extremely improbable), and that he will certainly die if pushed onto the tracks (perhaps not so improbable), whereas the "switch" scenario requires that if one flips the switch, the five people on the track will not die (perhaps intuitively much more probable than the effectiveness of pushing a man), and the person on the other track will die (perhaps not more probable than the probability that the man will die in the "fat man" scenario).

If some (many, most) of the people who answer the questions about permissibility, etc., are intuitively guessing that the man will probably die but fail to stop the trolley if pushed in the "fat man" scenario, but on the other hand, switching will (for example) more likely reduce than increase the number of casualties in the "switch" scenario, then perhaps that is the cause of the differences in people's judgments, rather than the fact that a person is used as a means in the "fat man" scenario.

That said, in this particular case, I do think whether a person is used as a means is often morally significant, and also that people often consider it so. Based on that, I'm inclined to think that that factor may well be driving at least part of the differences in people's judgments in the "switch" vs. "fat man" cases, if not all of them. But I'm not sure how much of it I can properly get from the results of the experiments.

Am I getting your idea right, or do you have in mind other ways in which contrasts might be tainted?

Dear Peter,

thanks a lot for your comment! Joshua Greene, for instance, has adopted the "partial solution" you hint at above by asking subjects about their "implicit realism". People who didn't think the fat man would succeed at stopping the trolley, for instance, were removed from the statistical analysis. This definitely seems like a step in the right direction, but I think it has serious limitations (which is indeed why it can only ever be a "partial" solution), because people may well be unaware of their implicit beliefs or their ability to discount their influence.

One question: do you think that the argument from contrastive responses ("contrasts matter even if indiviudal vignettes are unrealistic") does anything to defuse the lack of realism problem?


Dear Angra,

thanks a lot for your comment! That is a really good point -- I think both cases are possible. One is that contrastive judgments don't help overcome the lack of realism problem because the contrasts may be due to some intransparent difference. For instance, if there is imaginative resistance, people may well fill in details into the stories, which then end up triggering their differential judgments. This is, I think, the case you describe with differential probability assessment "tainting" the contrasts.

But I think contrasts will be tainted even if noch such additional hidden factors come into play. Consider ecological evalidity. Maybe people's judgment about trolley problems are not ecologically valid. So what, says the trolleyologist, as long as we those interesting differences in people's judgments? But if those differences themselves do not reflect any difference that we have reason to think would be ecologically valid, then the insistence on contrastive judgment does nothing to alleviate the ecological validity charge.

The comments to this entry are closed.