Designing Chatbots for Misinformation Correction: Examining the Roles of Chatbot Expertise and Anthropomorphism

Abstract

In response to the growing threat of scientific misinformation, the development of advanced AI chatbots presents an opportunity to provide customized and timely misinformation corrections on a large scale. Informed by the computers as social actors (CASA) paradigm, we conducted a 2 × 2 between-subjects online experiment, examining how chatbot anthropomorphism and expertise influence the effectiveness of chatbot delivered corrections. We found that anthropomorphic cues of the chatbot strengthened participant’s trust in the chatbot and promoted correction sharing behavioral intentions. Furthermore, we also found that chatbot expertise strengthened the influence of its anthropomorphism on perceived correction credibility and intentions to use the chatbot. This study highlights the potential for AI chatbot to tackle misinformation, providing practical design insights for misinformation correction and advancing our understanding of the CASA paradigm.