You are an autonomous car. Inside you carry three passengers. A catastrophic brake failure prevents you from stopping in time at the traffic light, while three other people cross the street. If you run over those three people, you will cause their death, but your passengers will be unharmed. On the other hand, if you decide to maneuver, you will save the pedestrians, and your passengers will die in the crash. What would you do? This and other exercises are explored on the Moral Machine portal, which invites us to share a human perspective on the moral conflicts that many automated systems will have to face.
"Kobayashi Maru". A scenario in which you cannot win. A "test of character", according to a certain captain who solved that scenario... by cheating. The problem with the Kobayashi Maru is that it was designed to put under fire what are undoubtedly human qualities, or in the specific case of Star Trek, the qualities of its officers. But what happens if we place logic in that situation? Ones and zeros. Artificial intelligences. In a matter of milliseconds they will evaluate all available options, and probably choose the most efficient, or the most effective. However, what about the morality and ethics behind that option? This is not a theory. Artificial intelligences will make decisions of that caliber very soon, when they control cars, vans, and trucks. The human factor is the main cause of traffic accidents, but mechanical failures will not disappear. What is "the right thing"? What is "the lesser evil"?
The Moral Machine portal, developed by MIT, helps us explore the answer to both questions by assuming the role of external observer before a scenario that an autonomous car must resolve. The exercises are nothing but "binary traps", since there are no alternatives, no shortcuts, no middle ground. One or the other choice inevitably leads to the death of someone, whether person or animal, but other elements are also incorporated. What is worth more, the life of a person crossing on a red light, or that of the one who waited their turn? In total, Moral Machine presents thirteen exercises, and at the end we can see a summary of our decisions. Number of lives saved, obedience to the law, gender preferences or social status, and even the level of intervention are evaluated.
Each exercise is randomly generated, so there is nothing like a "definitive result", and if you decide to start from scratch, you will surely see other recorded values. As the portal indicates in its fine print, Moral Machine does not seek to judge anyone, but to share with the public the importance and complexity behind ethical and moral decision-making by systems run by artificial intelligences.
