The Reasonable Robot Standard: How the Federal Government Needs to Regulate Ethical Decision Programming in Highly Autonomous Vehicles

Authors

Laura Emmons

Abstract

(Excerpt)

A steam train is chugging along down a country track. You are a passenger, watching red and orange clustered hills pass by; fall is already here. You start to daydream about the last time you took this trip, how the summer had just begun and how quickly it went. Suddenly, reality hits. You awake from your daydream when you glance up and see five people strapped to the train tracks ahead of you. You know there is not enough time for the train to brake. Luckily, you happen to be sitting right next to the emergency switch, which would divert the train to another path at the fork just before the train reached the five captives. They would be saved. Just as you grasp the switch with both hands, ready to pull as hard as you can, the train approaches the fork revealing another unfortunate person, strapped to the alternate tracks.

What should you do? If you pull the switch, your action will result in the death of one person. Five people will be saved, but you will have made the choice to alter the train’s path when you knew it would kill one person. The eager answer might be to save as many lives as possible and allow harm to come to as few lives as possible. But does your action of pulling the switch knowing someone will die count as killing someone? If you do not pull the switch, a chain of causation that has already been set in motion will result in the deaths of five people. You will not have acted, so you would not be responsible for their deaths. Or would you be? Does the fact that you could have acted to save their lives make you more responsible? Is inaction less culpable than action, in this case?