Cars and the Trolley Problem

I suppose that most people are familiar with the trolley problem, an imagined scenario challenging someone with the choice between doing nothing and watching 5 people die and acting to save those five people, knowing that doing that will kill one person. (There’s been some rather interesting research in recent years, suggesting that the particular circumstances of the decision posed can lead to people making rather different choices – see (or, rather, hear) this radio lab discussion.) New developments in automobile technology – especially self-driving cars – have renewed interest in and concern about the problem.

Consider an example that’s perhaps overly simplified, but, I think, not misleadingly so. Imagine you’re driving a car and you suddenly realize that there’s a person immediately in front of you on the road. You have a split second to decide whether to continue on your course and run over the person or veer sharply to the right, missing the person, but slamming the car into a solid concrete abutment. Suppose further that you are absolutely certain that running over the person will kill the person, and that slamming into the abutment will save the person but kill you. Now imagine different versions of the same scenario – imagine that you’re alone in the car and that it’s five people rather than only one in the path of your car. Or that you have your family of four with you in the car, and there’s only one person in the road ahead of you.

As I said, the scenario is overly simplified. It’s unlikely that you could know going into such a situation that death is certain in either case. And it’s even more unlikely that you’d have time to make a considered decision to do one thing rather than the other.

But new technological developments are bringing the problem back into the discussion, because a person writing the software governing a self-driving car has to decide how the car will respond to such situations. And this decision how to write the code is made under circumstances far more supportive of deliberative thinking than the circumstances when a driver of a conventional car is faced with a life or death decision.

It seems to me to be a rather difficult decision, but one that someone will have to make if we’re to have self-driving cars. Of course, given human fallibility, one could reasonably say that it’s less likely that a self-driving car would find itself in that scenario. The car and software presumably won’t fall asleep at the wheel, won’t drive while drunk, won’t try to read or respond to a text message. Still, the person writing the code has to anticipate such situations and write the program that will govern how the car will act.

But standing behind the question how we should decide is the question who will make the decision. I read somewhere recently that those discussing this question have considered the possibility of allowing the person who purchases a car to decide how they will want it to function. Do I want a car that I purchase to favor me and other occupants of my car over others, whether they are pedestrians, cyclists, or occupants of other cars? One with the inclination to favor others’ needs might opt for a car that will privilege the lives of others, even at the expense of one’s life. Another person – perhaps most people – would opt for the algorithm that favors the lives of the car’s occupants.

An interesting question, but it occurred to me today that it’s a problem that we’re already facing. Simply put, car manufacturers, for all sorts of reasons, are designing cars so that they protect the lives of the people occupying the car, in many cases at the expense of others. Cars are larger and heavier, which means that they cause more damage than they would if they were smaller and lighter. Moreover, most vehicles sit higher on the road, so that if they hit cyclists or pedestrians they’re more likely not only to hit them in the more vulnerable midsections than in the lower extremities, but also to push them to the ground and run over them, causing more fatalities and/or more serious injuries. It’s no accident that fatality rates for cyclists and pedestrians have gone up even as new technologies make things safer for the occupants of vehicles.

It’s an arms race out there, and not just for those who purchase cars. Auto manufacturers make a much higher profit manufacturing and selling the larger trucks and SUVs than they make selling compact cars. And they find themselves competing with each other to offer vehicles with more space for occupants and cargo, which means larger and heavier vehicles.

And consumers, faced with these choices, find it much easier to purchase a larger car or truck, and they appreciate the safety features highlighted in those studies showing how safe a vehicle’s occupants will be in a collision. Do those crash tests even rate the survival rate of those outside the vehicle, whether pedestrians, cyclists, or occupants of other vehicles? What would car purchasers do if presented with information suggesting that the driver of a car is significantly more likely to live in a collision, but that a pedestrian or cyclist is significantly more likely to die in a different sort of accident?

Trolley problem, indeed.