Ethical Learning, Natural and Artificial
Peter Railton
Abstract
Abstract How might artificial systems become sensitive to ethically relevant considerations? This chapter argues that the question we should ask is not “How can we build ethics into robots?” but rather “How can we build robots with a capacity for ethical learning?” It is evident from the continuing disagreement over overarching ethical theories that we do not know a set of consensus ethical principles with sufficient definiteness to “program” ethics into a machine. Recent research in artificial intelligence has shown the power of general-purpose learning to acquire autonomous human-level competencies. Intriguingly, recent research in developmental psychology suggests that children may also acquire independent social and ethical competence via similar mechanisms. Instead of programing ethics into robots, this chapter proposes that artificial systems could gain ethical competence through general-purpose learning, comparable to infants learning human ethics in part by observing adult behavior.