Building Trust in AI: The Role of Technical Capacity, Social Risk, and Corporate Institutional Accountability

초록

This study advances understanding of public trust in artificial intelligence (AI) by distinguishing between overall trust in AI as a system and trust in specific AI components, and by disentangling the roles of perceived capacity, risk, and personhood. Drawing on nationally representative survey data from 1099 U.S. adults collected in 2023 (AIMS dataset), the study estimates multiple regression models to examine how these evaluations shape trust across technical, organizational, and institutional dimensions. The results show that perceived cognitive capacity is the strongest positive predictor of both overall and component-level trust, while emotional and autonomous capacity primarily enhances trust in specific system components. Perceived social risk consistently undermines trust across all levels, whereas perceived personal risk mainly erodes trust in technical components. Importantly, support for granting AI legal or institutional status significantly increases trust, while moral consideration of AI exhibits limited direct effects, highlighting a critical distinction between institutional accountability and ethical concern. Together, these findings demonstrate that public trust in AI is not a unitary attitude but reflects multidimensional judgments about capability, risk, and governance. The study underscores the importance of institutional accountability and risk mitigation-alongside transparent communication about AI capabilities-for fostering sustainable public trust in AI.

키워드