Affirming or Refutational? The Roles of Message Sidedness, Communication Style, and Preexisting Attitudes in Low-Carbon Diet Persuasion by AI Chatbots

As generative artificial intelligence is increasingly deployed as a conversational agent in digital health, it raises theoretical questions about how different persuasive strategies operate and whether their effectiveness depends on recipients’ preexisting attitudes. This study investigates how message sidedness, communication style, and recipients’ preexisting attitudes interact to shape post-exposure attitude toward low-carbon diets in a scripted chatbot-like persuasive context. A web-based experiment with a 2 (message sidedness: one-sided vs. two-sided) × 2 (communication style: affirming vs. refutational) between-subjects design was conducted. A total of 300 participants interacted with a scripted AI nutritionist chatbot-like interface regarding low-carbon diets. Simple-effects analyses provided partial support for the moderating role of preexisting attitudes. Among participants with less favorable preexisting attitudes in the one-sided message condition, affirming responses were more effective than refutational responses. In contrast, among participants with more favorable preexisting attitudes, refutational responses exerted more favorable attitudes to participants than affirming responses when reading one-sided message. Theoretical and practical implications are discussed as well.