Website CV Paper
Fields
Behavioral & Experimental
Job market paper
Human Oversight of AI Redistributive Decisions
As consequential decisions are increasingly delegated to AI, both law and public opinion require human oversight of AI-powered systems. Yet, little is known about how people oversee AI decisions with direct consequences for others. This paper investigates whether people oversee redistributive decisions made by an AI differently from equivalent decisions made by another human. I develop a theoretical framework for the revision of redistributive decisions, and run two pre-registered online experiments with representative U.S. samples. I find that overseers judge the decision, not the nature of the decision-maker. They do not intervene more when the decision-maker is an AI rather than a human. When they lack information, however, they misjudge how the AI decides, making more unnecessary interventions on AI decisions and missing more warranted interventions on human ones.