TOKYO: In the not-too-distant future, driverless cars may have to choose between saving their passengers or pedestrians when faced with unavoidable accidents. But how should they decide?
It's one of the thorniest issues faced by policymakers and manufacturers as we edge closer to a future where autonomous vehicles fill our roads, and a new study offers some potential principles based on a survey of millions of people.
The researchers behind the study designed an online quiz with a variety of tough scenarios: should a car favour the lives of the young, or the old? Should it save the fit over the infirm? Is saving more lives always preferable to saving fewer?
They wanted to see if people around the world would settle on some fundamental guidelines.
"We identified three main principles on which people more or less agreed," said Jean-Francois Bonnefon, study co-author and a professor at the Toulouse School of Economics.
They were: "protect human life (over animals), save the largest number of people, and place priority on saving children," he told AFP.
"But even with these strong preferences there was variation from one country to another."
The study, published Thursday in the journal Nature, found sometimes surprising differences among regions, including a "strong preference for sparing women and... fit characters" in France and French-influenced areas, as well as countries in Latin America.
"But the fact that broad regions of the world displayed relative agreement suggests that our journey to consensual machine ethics is not doomed from the start," the study said.
The quiz, dubbed the "Moral Machine", was put online in June 2016 in 10 languages, and the study is based on responses over an 18-month period.
So far, more than two million people have taken the test, which remains online.
Edmond Awad, a postdoctoral associate at MIT and first author of the study, said the data was intended to help guide policymakers.
"Experts don't have to cater to the public's interests, especially when they find these preferences problematic," he told AFP.
"But they should at least be aware of the potential reaction of the public when something goes wrong and their regulations are in place."
The authors of the study acknowledged their decision to use an online quiz to reach the maximum number of participants meant their data set has some shortcomings.
"Our sample is self-selected, and not guaranteed to exactly match the socio-demographics of each country," the authors wrote.
"The fact that our samples are not guaranteed to be representative means that policymakers should not embrace our data as the final word on societal preferences."
And it remains an open question whether government and regulators would be comfortable effectively enshrining a hierarchy of human life into law.
To date, few governments have attempted to codify guidelines for driverless cars, but one example is a 2017 report by a German commission.
It states explicitly that "any distinction based on personal features (age, gender, physical or mental constitution) is strictly prohibited."