Presentation Information
[4Yin-A-20]AI as a Source of Scientific Information: A Cross-National Study of Persuasion and Source Credibility
〇Ruijun Sophia Liu1 (1. University of Tokoy)
Keywords:
Source Credibility,AI-Mediated Persuasion,Human-Centered AI,Cross National Comparison
As Gen-AI systems increasingly mediate communication, it remains unclear how they function as sources of persuasion and credibility. This study examines how labeling scientific information as generated by an AI system (ChatGPT) versus a human expert shapes persuasive outcomes and perceived source credibility, and whether these effects differ across national contexts. We conducted an online experiment with participants in Japan and the United States (n=671), holding message content constant while manipulating source labels and topic domain. Post-treatment attitudes were analyzed using ANCOVA models controlling for baseline attitudes, alongside measures of perceived source credibility. Across both countries, AI-labeled messages were at least as effective in shifting attitudes as expert-labeled messages relative to control, though the magnitude of persuasion was weaker in Japan. In contrast, AI was evaluated as less credible than a human expert in the United States, but not in Japan, suggesting a dissociation between credibility perceptions and attitude change.