CORE S197   Ethics, Algorithms, & AI

Machine-learning algorithms and autonomous systems introduce a wide range of morally significant questions: about rights, fairness, consent, accountability, trust, transparency, exploitation, and sustainability, among others. For example, are there moral costs to the design and training processes of such algorithms? ls it acceptable for an algorithm to classify people differently on the basis of things they have no control over, even if it's very accurate? How do we tell a machine we want its outcomes to be fair? What should we do when there is entrenched disagreement about moral values? More broadly, is it a problem if a machine-learning system develops a standard that is too complex to be recognizable as a human moral concept, or even understood by humans at all? If a machine reaches a certain level of sophistication, can it acquire moral status, e.g. responsibility for its decisions?

Students consider questions like these systematically and philosophically, with knowledge of the predictive reasoning underlying such systems. Special attention is paid to the difficulty of narrowing the gap between mathematical precision and human intuition. Readings come from a range of fields: computer science, contemporary philosophy, statistics, cognitive science, and law.