Understanding Citizens' Acceptance of AI in Policymaking: How Policy Domains, Levels of AI Involvement, and Stereotypical Perceptions Shape Public Support
研究英国公民对AI参与政策制定的接受度，发现技术性议题中AI接受度最高，意识形态议题最低，且对AI和人类决策者的刻板印象（能力与温暖）显著影响公众支持。
ABSTRACT This study examines UK citizens' perceptions of Artificial Intelligence (AI) in policymaking. We first explored the interactive influence of three empirically derived policy issue types (technical, mixed, and ideological issues) and five different levels of AI involvement (ranging from no involvement to full involvement) on public acceptance (RO1). Our results showed that acceptance of AI is highest in technical issues, particularly in a co‐decisive role. For mixed issues, AI was moderately accepted, with advisory roles being most preferred. In contrast, ideological issues were met with the lowest levels of acceptance, with most respondents favoring little or no AI involvement. Secondly, we examined how stereotypical perceptions of AI and human decision‐makers influence AI acceptance using the competence–warmth framework (RO2). While humans are perceived as both competent and warm, AI is seen as competent but lacking warmth. Moreover, across all three policy issue types, higher perceived human competence reduced AI acceptance, whereas higher perceived AI competence increased it. Additionally, perceived lack of AI warmth undermined acceptance in mixed and ideological issues. Together, these findings highlight the complex factors influencing public acceptance of AI in governance and underscore how stereotypical perceptions function as underlying drivers of these preferences.