Ensuring the Consent of the Governed: Surveying the U.S. Public’s Attitudes Towards AI Governance Regimes

Abstract

The rapid advancement of artificial intelligence (AI) capabilities has generated numerous proposals for oversight and governance frameworks to ensure that the technology is developed safely and responsibly. There has been little focus, however, on the question of who the public would trust to provide this oversight. This paper examines the public’s perceived need for oversight, as well as the types of organizational structures they would trust to manage increasingly powerful AI. Using a nationally representative survey of 2,000 U.S. adults, we found that the American public very strongly favors monitoring and oversight of the most powerful AI systems. We then examined the types of institutions that Americans would trust most to manage such systems. Respondents were asked to choose from among six different organizational forms are commonly represented in proposals for AI governance: the US government, a foreign government, a private company, a public-private partnership, a multinational partnership, or no oversight. The results were striking. Even though the respondents had earlier rated U.S. dominance in AI as very important and had rated the potential of Chinese dominance very unfavorably, the preferred forms of governance were multinational and public-private partnerships. This suggests that even as the American public views AI development partially through the lens of national competition, many Americans also regard multiparty and international bodies as trusted providers of effective AI oversight. These results provide important empirical evidence for debates over the feasibility and perceived legitimacy of international AI governance alternatives.