Exploring Trust and Literacy in Engagement With Generative AI and Science Information Behavior

Abstract

As generative AI (GenAI) becomes increasingly embedded in everyday information environments, understanding how citizens engage with this technology is critical for science communication. This study examines public engagement with GenAI in Denmark, focusing on trust, AI literacy, experience with GenAI tools, and exposure to science-related information. Denmark provides a relevant case due to its high levels of institutional and scientific trust. Using data from a nationally representative survey conducted in 2024 (<em>n</em> = 514) as part of the cross-national ScI-AI project, we analyze how respondents encounter GenAI, assess its trustworthiness, understand its technical and epistemic features, and engage with science-related information across platforms. Descriptive results show moderate trust in GenAI, uneven AI and GenAI literacy, and concentrated experience centered primarily on ChatGPT, alongside pronounced concerns about misinformation and societal risks. To examine how these dimensions relate, we apply a probabilistic graphical model to 29 variables spanning trust, literacy, experience, science-related information exposure, and demographics. The analysis reveals that trust occupies a central position, mediating between technical understanding of GenAI’s functioning and epistemic beliefs about the reliability and truthfulness of its outputs. Science-related information exposure is largely disconnected from trust and GenAI literacy and links to general AI literacy primarily through gender. Overall, the findings highlight the importance of treating trust and literacy as multidimensional and context-sensitive constructs for understanding how GenAI reshapes science-related information encounters.

Ähnliche Arbeiten