« Back

“I’m not normally afraid of change but this scares me”: Trust, transparency and the uncertainties of generative AI

Author(s): James White, Stefan Larsson

Friday 16  |   13:20-13:40

Room: TP52

Session: Robots in social life: interactions, hierarchies, and intimacies

AI is a complex technology that already affects many spheres of social and economic life. Significant attention has been directed towards the risks that this entails, both present and future, and policy and regulation debates have responded accordingly. However, recent advances in AI, most notably in terms of coherent text and image generation, have quickly been made available to the public through new and well-established apps and services. This poses a challenge to existing risk analyses and to regulatory efforts based upon them. In this paper, we explore changes in the risk landscape of AI brought about by the rapid emergence and deployment of its powerful generative capabilities. This is done principally through the presentation and analysis of the results of a survey of more than 1,000 Swedish people conducted in December 2023 and January 2024. Drawing on the conceptual language of agnotology and risk sociology, we stress the importance of respondent uncertainty, less with respect to how AI functions, than to how it threatens established epistemic practices and seemingly renders the future more unknowable. This has important consequences for policy and regulation aimed at improving AI transparency and trustworthiness. Our findings suggest that the risks of generative AI, as perceived by the Swedish public, may best be addressed by ensuring full disclosure of direct interactions with AI and its use as a source of media and information, and by properly mediating the involvement of industry in regulatory development.