A Problem of Metrics
Recently I've been using the expression "trolley problem" as shorthand for a whole family of "you can't save everyone" problems in which saving one group of people brings about the deaths of another group of people. However, it now turns out that the traditional formulation of the trolley problem may not be measuring what its creators intended.
It turns out that people give very different answers to this sort of dilemma depending on whether it is presented abstractly or in more concrete terms. In one study, one group of participants was read a scenario in which there were two cages, one with five mice and the other with one. There was a twenty-second timer, and when it runs down, a painful but non-lethal shock would be delivered to the five mice -- but there was a switch by which one could redirect the shock to the cage with only one mouse. The other group was actually confronted with two cages of mice and a laptop with a countdown timer, and had to decide whether to push the button. At the end they could see that neither cage received any shocks, but because of the countdown timer, the participants had to make a choice beforehand.
Interestingly enough, it was the participants who were presented with actual living, breathing rodents who were more likely to push the button, sparing the five mice at the expense of the one. Even more interesting, when participants were presented with the more traditional formulations of the trolley problem beforehand, as part of their orientation, the answers they gave to those practice rounds did not predict how they would respond when actually confronted with the mice in cages.
This squares with another experiment in which participants were given VR versions of the traditional trolley problem scenarios via an Oculus Rift headset. Yet again, when confronted with a more realistic presentation of the scenario, people were willing to be more pragmatic and sacrifice the one for the good of the many.
The researchers hypothesize that, as long as the dilemma is presented in a hypothetical way, people become more concerned about how their responses will be perceived by both the testers and by anyone else in the room. That is, their responses show as much concern for their reputation as moral and ethical members of society as their own moral grounding. But when the dilemma is presented in concrete terms that make it feel like real people's lives are on the line, and especially when under time pressure (as is often the case in actual situations where people have to make life-and-death decisions), people tend to become more pragmatic in their approaches to harm mitigation -- and it's quite possible that the people who refrain from acting in fact are freezing rather than making a choice of non-action under the principle of "better many natural deaths than one murder."
Unfortunately, it is all too common that tests end up actually testing something ancillary or even unrelated to what is intended. I still remember the "oral comprehension" test in a foreign language class that ended up actually testing our ability to chunk incoming information and avoid having it overflow our short-term memory. I have strong reason to believe that the creators of the test didn't even realize that they were able to store an incoming stream of speech in multi-word chunks, or that short-term memory was a limited resource, so they never thought to train us on that process by having us listen and translate lengthy sentences in class. As a result, when we all requested again and again that the proctor slow down and say the sentence again, they just told us we needed to listen closer and work harder.
It turns out that people give very different answers to this sort of dilemma depending on whether it is presented abstractly or in more concrete terms. In one study, one group of participants was read a scenario in which there were two cages, one with five mice and the other with one. There was a twenty-second timer, and when it runs down, a painful but non-lethal shock would be delivered to the five mice -- but there was a switch by which one could redirect the shock to the cage with only one mouse. The other group was actually confronted with two cages of mice and a laptop with a countdown timer, and had to decide whether to push the button. At the end they could see that neither cage received any shocks, but because of the countdown timer, the participants had to make a choice beforehand.
Interestingly enough, it was the participants who were presented with actual living, breathing rodents who were more likely to push the button, sparing the five mice at the expense of the one. Even more interesting, when participants were presented with the more traditional formulations of the trolley problem beforehand, as part of their orientation, the answers they gave to those practice rounds did not predict how they would respond when actually confronted with the mice in cages.
This squares with another experiment in which participants were given VR versions of the traditional trolley problem scenarios via an Oculus Rift headset. Yet again, when confronted with a more realistic presentation of the scenario, people were willing to be more pragmatic and sacrifice the one for the good of the many.
The researchers hypothesize that, as long as the dilemma is presented in a hypothetical way, people become more concerned about how their responses will be perceived by both the testers and by anyone else in the room. That is, their responses show as much concern for their reputation as moral and ethical members of society as their own moral grounding. But when the dilemma is presented in concrete terms that make it feel like real people's lives are on the line, and especially when under time pressure (as is often the case in actual situations where people have to make life-and-death decisions), people tend to become more pragmatic in their approaches to harm mitigation -- and it's quite possible that the people who refrain from acting in fact are freezing rather than making a choice of non-action under the principle of "better many natural deaths than one murder."
Unfortunately, it is all too common that tests end up actually testing something ancillary or even unrelated to what is intended. I still remember the "oral comprehension" test in a foreign language class that ended up actually testing our ability to chunk incoming information and avoid having it overflow our short-term memory. I have strong reason to believe that the creators of the test didn't even realize that they were able to store an incoming stream of speech in multi-word chunks, or that short-term memory was a limited resource, so they never thought to train us on that process by having us listen and translate lengthy sentences in class. As a result, when we all requested again and again that the proctor slow down and say the sentence again, they just told us we needed to listen closer and work harder.