Self-driving cars are already being tested on the road, but tricky questions about their ethics remain. Should autonomous vehicles protect their passengers at all costs? Or should they be programmed to sacrifice their passengers, if it means protecting others? A new study says “yes” and “yes.” Scientists asked 451 online survey-takers whether—in the event of an inevitable accident—it was more appropriate to sacrifice passengers or bystanders, a quandary known as the trolley problem in ethics. When the ratio was one-to-one, that is, one passenger to one pedestrian, about 75% of respondents said the passenger should be saved, they report online today in Science. But as the number of pedestrians went up, survey-takers started to change their minds: When there were five pedestrians and one passenger, 50% said the passenger should be saved. And when there were 100 pedestrians, that number dropped to about 20%. Those responses are in line with the results of a second survey, which asked 259 people whether automakers should program vehicles to preserve the “most life,” a stance known as the utilitarian approach. On a 100-point scale, respondents gave an average approval rating of 70. But when asked whether they would buy cars programmed to sacrifice drivers, they were much less interested, giving the idea an approval rating of 30. This raises a challenge for policymakers and companies like Google and BMW, who support the widespread adoption of self-driving cars as a way of cutting pollution and saving lives. Although some argue that roads would be much safer with autonomous cars in charge, the very programming that provides this safety may prevent them from ever being on the roads.