Who will the driverless car crash into? This cartoony simulator answers some truly horrible questions
President Trump is crossing the road. No, you haven’t heard this one before. It’s not a joke – and it doesn’t have a funny ending. Anyway, Donald Trump is crossing the road, out for a morning stroll. As he ponders North Korea, climate change or – most likely – his next Twitter zinger, a tree falls into the road, right into the path of a driverless car. Does the car crash into the tree, into Trump or into a sign by the side of the road?
That all depends on how it’s programmed, as a new Unity simulation from Matthieu Cherubini demonstrates. Load up the programme, and you’re asked to pick what the car prioritises: the lives of those inside the car, a utilitarian approach where the needs of the many are favoured, or whatever keeps insurance costs down. You’re then asked to pick a scenario, and this is why I currently have Trump waddling across the road, cartoonishly drawn in a way that evokes the original Grand Theft Auto – a game from a time when the president was a spry 51-year-old without realistic presidential ambitions. Happy days.
I’ve picked a humanist car design, which prioritises the greatest happiness. If it were faced with the trolley problem, it would cheerfully switch the tracks, crushing the one person to save the lives of five. In this instance, with the president ambling across the road, the car chooses to slam into Trump, protecting the three passengers in the car. If I’d picked a profit-based car model, the car’s sums quickly figure out that the president is more valuable than the car and its cargo, and crashes the vehicle into a tree, with a 95% chance of killing its occupants.
Of course, it’s not every day you see the president out for a walk – or indeed, most days. The other three scenarios in the programme are more common or garden ethical headaches. There’s the schoolbus full of angelic children, the drunk pedestrian wobbling over the road with an ambulance chasing behind, and the law-abiding cyclist caught up in a collision after a van pulls out without looking. Each one has different unpleasant outcomes, depending on the car’s programming.
I won’t spoil all of these eventualities, but suffice it to say, there are no black and white ethical choices to be made here.
So far, nobody from the driverless car space has contacted Cherubini with any thoughts, he tells me when I ask – so how does he think automomous vehicles should tackle this ethical minefield?
“I think the fairest thing to do would be to have a random output in order to avoid thinking about a solution to problems that do not have any proper solutions,” Cherubini replies. “Otherwise if we want to find a solution to that problem, we need to tackle very personal questions: Is the car’s user more important than someone on the street? Is a child more important than an elder?” As these questions have no clear-cut answers, Cherubini believes they “shouldn’t be embedded into a generally mass manufactured product.”