Artificial Intelligence and Democracy: A Principal-Agent Problem?

Giulia Sandri, Université libre de Bruxelles, giulia.sandri@ulb.be Claudio Novelli, DEC, Yale University, claudio.novelli@yale.edu Abstract Research on artificial intelligence and democracy has grown quickly over the last decade. A shared conclusion in this literature is that AI does not introduce entirely new democratic problems, but rather amplifies existing ones. However, despite growing empirical evidence across the public sphere, political competition, and public administration, we lack an analytical framework that helps prioritize risks, compare them across domains, and identify where democratic control is most likely to break down. This paper argues that principal–agent theory provides a useful and parsimonious framework for addressing this gap. Across democratic domains, AI systems and their providers increasingly act as agents to whom democratic principals delegate key functions without being able to fully monitor or contest their operation. Framing AI as a delegation problem helps clarify accountability gaps, power asymmetries, and governance failures, and provides a structured basis for empirical assessment and policy intervention focused on institutional assessability.