In philosophy, there's an ethical question called the trolley problem. If you had to push one large person in front of a moving trolley to save a group of people on the tracks, would you? This abstract idea has taken hold in programming self-driving cars: what happens if it's impossible to avoid everyone?
Researchers from the Toulouse School of Economics decided to see what the public would decide, and posed a series of questions to online survey-takers, including a situation where a car would either kill 10 people and save the driver, or swerve and kill the driver to save the group.
They found that more than 75 percent supported self-sacrifice of the passenger to save 10 people, and around 50 percent supported self-sacrifice when saving just one person. However, respondents didn't actually think real cars would end up being programmed this way, and would probably save the passenger at all costs.
The questions were answered by paid participants on Amazon Mechanical Turk (compensated to the tune of 25 cents for their time), and in total 913 participants took three surveys with different questions. Since it was an online survey, the participants could have been from anywhere in the world.
One version randomized the amount of people that would be killed if the driver did not swerve (between 1 and 10) and asked if the car should sacrifice the passenger or bystanders. The second version tested how people would program cars themselves—always sacrifice the passenger, always protect the passenger, or random, and asked to rate the morality of each. The third group were read a story where ten people were saved because the car swerved, killing the passenger. They were asked to imagine themselves as the passenger, and then a bystander, and assess the morality on a slider.
When analyzing the demographics of their respondents, the team also found that those excited for self-driving cars were younger and less religious.
Researchers wrote that survey respondents were generally receptive to autonomous cars making decisions that prioritized many over few (called utilitarianism in philosophy). However, they foresee regulations being tricky, asking whether the public would support a law that requires cars to sacrifice their passengers under certain circumstances.
This is an issue that will undoubtedly plague the autonomous car industry even after they enter widespread production. Sure, driverless cars can reduce traffic fatalities by up to 90 percent. And like the field of ethics itself, what happens in the other 10 percent is still up for debate.