Keywords: generative ai, meta, deliberative polling, privacy

Sociotechnical imaginaries concept focuses on desirable AI futures while also addressing undesirable developments that should be avoided. It reveals visions of the future that are informed by the historical and contemporary experiences of society. Shared values and normative expectations in society and the experiences of everyday people are essential for understanding public imaginaries. This study aims to examine public imaginaries of AI and the associated factors that could shape theses imaginings. It explores correlations between participants’ use of AI, knowledge of AI, perceptions of AI technology and favorable attitude and imaginaries. We analyzed data from the deliberative polling conducted in 2023 by Stanford University’s Deliberative Democracy Lab in Brazil, Germany, Spain, and the United States. It explores changes in public imaginaries of AI future after deliberation. The research uses factor analysis and regression models to identify the factors that drive changes in AI attitudes and imaginaries and to determine whether these factors differ across imaginaries. It shows that public imaginaries are directed towards more desirable visions of “user consent and privacy protection”, steering clear of any undesirable human rights violations, and risks of AI replacing human interaction. This article highlights the dynamic public imaginaries surrounding the AI interactions with human, which could facilitate the shaping of the AI governance process.