Delegating Moral Decisions to AI Systems

Humans regularly make moral choices. As artificial intelligence (AI) technology advances, task delegation to AI is increasing, leading to machines making moral decisions. Research on task collaboration with AI is emerging. However, the willingness to actively delegate moral decisions to AI systems has not been widely addressed, especially in connection with post-hoc moral evaluations. To this end, this thesis examines the human willingness to delegate moral decisions to AI systems or human agents and the impact of this decision on the perceived moral responsibility for the consequences in moral dilemmas. In two studies (total n = 669), contextual factors (domain and trade-off) and agent-related factors (trust, risk aversion) are investigated by their influence on imagined moral decision-making with the possibility to delegate. It shows that delegation increases when decision stakes are high, and that trust and risk aversion influence delegation behavior. Furthermore, delegation effectively shifts the perceived moral responsibility for the consequences of the decision to the AI system and human agent, leading to a form of shared responsibility. Psychological factors that drive delegation to AI are discussed and translated into recommendations for the ethical governance and implementation of AI systems, particularly in situations involving moral decisions.