We’re going to play an ethical game. It’s a classic one that you have probably heard of before but is now beginning to have real world applications that I will explain a little later. Let us begin.
You are on a bridge overlooking a train track. Tied to the track are 5 people and there is no way they can escape. Hurtling towards them is an unstoppable pump car (one of those weird things with two guys pumping at the handles to make it move along the track) but fortunately for the hapless victims the switch to move the car onto a different line is within your reach. Unfortunately, tied to the alternative line is 1 person. What do you do? If you do nothing 5 people die, if you act only one person dies, but you have to chose for them to die.
Let’s make it more personal. We still have 5 people about to be turned into mush but this time you are on the pump car. It cannot be stopped except by you pushing your fat companion off the front of the car and under the wheels. Do you push them?
This is a decades old thought experiment with myriad other permutations and up until now it has just been a fun way to think about the value of human life and what we would do to save it. But now, with the advent of driverless cars, the people that write the algorithms that determine the car’s behaviour are actually having to come up with real world answers.
Imagine a driverless car is going along its merry way and is either empty or its passenger is not in a position to take control. A person runs out into the path of the vehicle and it has to decide what to do. It probably brakes hard but should it also swerve? If there is a car in the way should it hit the car in preference to the pedestrian? What if the other vehicle is a school bus? What if the pedestrian is a child? All these ‘decisions’ the vehicle makes are as a result of the code written by the programmer and are, therefore, effectively their decision. Hands up who wants that job?
Ultimately, like everything, this question will be answered by the data. Right now we don’t really have any, driverless cars are still a new technology, but a few years from now we’ll have amassed the info we need to ensure the best statistical outcome. Bad things will still happen but I’m confident they’ll happen far less than they do now. I, for one, welcome our new robot overlords.