Trusting machines with morality - delegating moral decisions to AI

Research suggests that individuals are generally skeptical about the use of artificial intelligence (AI) in moral contexts, favoring human decision-makers over AI. Yet, in two experiments involving a total of 5639 participants, we find that individuals facing a real-life moral decision delegate significantly more often when they can delegate to AI rather than to a human counterpart. This result highlights AI’s relative appeal as a moral delegate, indicating that individuals’ preferences for AI’s involvement change when they themselves assume the role of a decision-maker. Responsibility shifting, previously studied as a motive for delegation to humans, extends to AI delegates. Moreover, it appears to be facilitated by individuals adapting their beliefs about AI’s capability in a self-serving manner. Ambiguity surrounding that capability allows them to interpret it in ways that justify delegation. These findings add nuance to assumptions about algorithm aversion in moral domains and raise critical questions about accountability and the ethical implications of relying on AI for morally sensitive decisions.