© Colleen Hayes / NBCThe trolley problem as featured in the hit TV show The Good Place
Would you kill someone if it would save the lives of five others? This classic thought experiment is known as the trolley problem, and is taking on growing importance as we train self-driving cars to take to the road. It's also had a recent rise in recognition thanks to its role in philosophical sitcom
The Good Place. But the first real-life enactment of the problem in a lab - using mice - suggests we may have been approaching it wrong.
The
trolley problem involves imagining that a runaway rail car is going to hit and kill five people - unless you pull a lever, diverting the car onto a different track, where it would only wipe out one. Rerouting the car would logically cause the least harm, but some people struggle with the hypothetical guilt of hurting someone through their direct actions and
say they wouldn't be able to pull the lever.
Dries Bostyn of Ghent University in Belgium and his colleagues wanted to know if people would show this reluctance in a real-life version of the test. To do this, they used mice as the victims instead of people, and recruited about 200 volunteers. Each person entered a room and was told that a very painful but non-lethal electric shock was about to be applied to a cage of five mice in front of them. But if the person pressed a button, the shock would be diverted to a second cage, containing just one mouse.
Less emotionalHappily, no mice were harmed during the study. A 20-second timer counted down while participants had to make up their minds - but at the end, there was no shock.
Overall, 84 per cent of people opted to press the button.
This was significantly more than when a separate group of people were asked to imagine the same scenario. In that experiment, only 66 per cent of participants said they thought they would press the button.
This contradicts the current assumption that, faced with the dilemma in real life, more people would feel too guilty to pull the lever than in hypothetical thought experiments, says Bostyn. Studies have so far assumed that the more real it gets, the more emotional we get -
but the reverse is true, he says.Self-driving carsOne problem with Bostyn's study is that, in the live mice version, some volunteers didn't fully
believe the mice would truly be hurt - on average, the participants said they were 55 per cent certain the experiment was real.
However, Bostyn says that most people got upset at the prospect of choosing which animals would suffer. "Some of them were visibly shaken by it."
If the team is right, and people really are more likely than they say they are to pull the lever, this could have implications for how we program
autonomous cars, which, for instance, may have to "decide" whether to crash into an oncoming vehicle or swerve off the road into pedestrians.
Research groups, including a team at the Massachusetts Institute of Technology, are investigating what people think self-driving cars should do over a range of different situations. People can do the MIT tests online on a site called the
Moral Machine. "This paper casts a bit of a shadow over that work," says Josh Cowls of the Alan Turing Institute in London.
"There's a gap between how we answer these questions hypothetically and what we do when faced with a real-life situation." Experimental ethicsLast year, Germany released ethical
guidelines for self-driving cars that said in potential collisions they should not discriminate between people based on factors such as age, gender or physical condition. But the mouse experiment suggests that, when facing dilemmas like this, we may have been getting an imperfect picture of human ethics.
"If we want to make human-like machines, rather than sitting down and have a big old think about it, we need to broaden our understanding of what ethics is in practice," says Cowls.
However, we shouldn't generalise too much from one study, says Cowls, especially as it involved potentially hurting animals rather than people. Bostyn's team plans next to recreate the experiment but by really shocking people instead. "It's easier to get ethical approval to shock humans than mice."
Journal reference:"Of Mice, Men, and Trolleys: Hypothetical Judgment Versus Real-Life Behavior in Trolley-Style Moral Dilemmas", Dries H. Bostyn, Sybren Sevenhant, Arne Roets
, Association for Psychological Science,
DOI: 10.1177/0956797617752640
Reader Comments
to our Newsletter