Imagine you are a train-yard operator who sees an out-of-control boxcar running down a track that five workers are repairing. The workers won’t have time to get out of the way unless you flip a switch to change the car to another track. But another worker is on the second track. You have just seconds to make a decision: let the five workers die — or kill the one. What do you do?
This dilemma is a famous philosophical conundrum that was originally called the “trolley problem.” Now a team from Michigan State University’s psychology department has used virtual-reality technology to test how we respond psychologically and physiologically when faced with this problem.
The two opposing philosophical approaches to the trolley problem are the utilitarian one (kill one guy in order save the others) and the do-no-harm approach (let God or nature take its course, but don’t make an active choice to kill another person).
In many years of surveys, the vast majority of people — usually about 90% — have chosen to kill the one and save the five. But until now, there’s never been a study examining how people would react in a lifelike setting with real-looking potential victims.
In the Michigan State study, led by psychologist David Navarette, the 147 participants made their choice while wearing a head-mounted virtual-reality device that projected avatars of those who could die. (Watch a simulation here.) One chilling factor of the test: the potential victims were screaming as the boxcar approached.
The 147 subjects also had electrodes attached to their skin in order to measure their autonomic responses, the involuntary nervous-system responses that can spike when we are faced with stress. Navarette and his team found that, once again, 90% of us would kill the one to save the five. Among the 147 participants, 133 pulled the switch.
Interestingly, those who were more emotionally aroused during the simulation — based on measurements of electrical conductivity along the skin — were less likely to kill the one. Those who were colder and more calculating did what I would do: run that guy down as fast as possible.
In another test, the Michigan State team changed the experiment so that the train would kill the one person unless it was diverted to kill the five. In other words, this time the participants had to choose passive, restrained action: just let the train continue on its course and mow down the one guy. Once again, 90% chose to save the five over the one. This group was also, on average, less emotionally excited than the 10% who had to act to save the one life.
In short, those who can control their emotions are more likely to murder one in order to let five live. Still, the authors note limitations to their study. First, a virtual-reality world is just virtual: there are no legal consequences to killing an avatar. Also, surveys have shown that when people are asked whether they would use their hands to push a lone person from a track in order to save five, only approximately half would do it. We don’t want to get our hands dirty.
Finally — and most important — when the one person you would have to kill to save five is your child, parent or sibling, only approximately one-third of us will opt to protect the five people. (I’m getting those stats from a 2010 paper in the Journal of Social, Evolutionary, and Cultural Psychology.)
What does all this teach us about human nature? Evolution has hardened us into brutal and selfish creatures. We make split-second calculations that result in murder — unless a family member is at stake. But let me ask: What would you do? Please comment below or on our Facebook page. We will compile your responses and report the results a week from now, on Dec. 12.
John Cloud is a senior writer at TIME and the host of the Healthland Podcast. Find him on Twitter at @johnashleycloud. You can also continue the discussion on TIME Healthland’s Facebook page and on Twitter at @TIMEHealthland.