Speaking Through Chatbots or Text: How Format Shapes Information Agreement, Reactance, Environmental Awareness, and Trust

Abstract

Recent advances in large language models have made conversational agents (CAs) equipped with environmental knowledge a promising tool for fostering environmental awareness. Yet it remains open whether CAs outperform static text in increasing agreement with the information, environmental awareness, and trust. In this preregistered, online, multi-week, mixed-methods intervention (N = 449), we implemented a CA and compared it with static text. Information format (CA, text) and valence (positive, negative, and neutral) were between-subjects factors, and time was a within-subjects factor. Results showed higher agreement and lower reactance to information delivered by the CA compared with text, regardless of the valence of the information and time. No effects were found on environmental awareness or trust. Environmental awareness increased over time, especially for participants with low environmental awareness, but this remained independent of format and valence. Trust in CA also increased over time in negative and neutral valences, but not in positive