If someone makes a decision you don’t like you can complain to the manager. But what if the decision-maker’s a computer not a person? Computer programs are making more and more decisions from the most mundane request right up to deciding the length of a prison sentence. How can we design ethical artificial intelligence?