You Don’t Have to Hit the Grandma

Re: autonomous vehicles and the whether-to-hit-the-grandma-or-small-child debate

This is a choice that companies and the public sector are trying to ensure that autonomous vehicles never need to make. Here’s how.

Level 5 autonomous cars (truly driverless cars) will rely on in-vehicle technologies that help them see and understand the world around them, communicate with other vehicles on the road, and take action based on the information at-hand. By doing these three things at once, each autonomous car will have a real-time map of real-world conditions behind them, in front of them, and to the side of them, not to mention having predictive capabilities about what it thinks each of the other actors on the road/sidewalk will do.

You’re probably thinking, well, what if there’s a hack into the car and some evil genius re-routes the car to hit the kid, and then the person in the car overrides the override and hits the grandma instead. Then you still have a philosophical conundrum; it’s just transposed onto the driver. Right?


Although automobile manufacturers would like to have you believe that autonomous cars will be able to do everything on their own, without any external infrastructure and without any government intervention (apart from maybe the initial regulatory framework that establishes universal standards for safety and communication), the scenario I just described is why that’ll never happen.

Companies and Cities (municipal governments) are already working on infrastructure solutions that would act as fail-safes when vehicles even start acting abnormally. These solutions would detect strange behavior and create geo-fences around people and other vehicles or automatically beam out potential overrides for the override or simply kill the car dead, with different tactics depending on whether the vehicle has an electric or combustion engine.

Of course, all of these solutions kind of assume every vehicle on the road is autonomous. It’s very likely that that won’t be the case in the short-term, or maybe ever. So what happens if there’s one non-automated car on the road that’s driving directly towards the autonomous vehicle and the autonomous vehicle could hit the non-automated car or a grandma and a kid or maybe crash into a wall?

This is a conundrum that auto companies are planning for already by embedding lower levels of autonomy into new vehicles — things like automatic braking. So new vehicles, whether they’re fully autonomous (Level 5) or Level 3 or 4, will have some capacity to regulate. Cars turn over, on average, every 15 years, so we’re not locked into our current car fleet forever. As for the next 15 years? Well, likely what we’ll see is that cities will physically separate autonomous traffic from non-autonomous traffic on high-capacity corridors, with their own lanes on highways and on really busy streets. And on less busy streets, we’ll see. Some companies have already decided that their cars will prioritize the driver. [Cough, cough, Mercedes.] Others are contemplating imbuing cars with moral frameworks, but then the question is which philosophical tradition will win out.

Long story short, “Hit the grandma or the kid?” is still an interesting question, still probably valid in the short-term, but won’t be valid if companies and governments collaborate for a smooth roll-out of autonomous vehicles. [One can dream.] Also, let’s be real, the insurance industry in the U.S. has probably already answered this question anyways by valuing our lives based on metrics they think aren’t arbitrary.