Interacting with Healthcare Chatbot: Effects of Status Cues and Message Contingency on AI Credibility Assessment

作者：Yujie Dong, Yuheng Wu · 发表于：International Journal of Human-Computer Interaction · 年份：2024 · DOI：10.1080/10447318.2024.2387396 · 被引用次数：12 · 研究领域：Death Anxiety and Social Exclusion、Mental Health via Writing、Misinformation and Its Impacts

How do status cues of AI agents alter human users’ interpretations and evaluations of these agents during human-AI interactions? Drawing upon social status theories and the CASA paradigm, this study explores the influences of status cues in AI-powered healthcare chatbots on AI credibility assessment via perceived AI agency and AI anxiety. The moderation effects of message contingency are also examined. A 2 (status cues: high-status vs. low-status) × 2 (message contingency: high vs. low) between-subjects experiment was conducted online (N = 209). Results revealed that healthcare chatbots with high-status (vs. low-status) cues were perceived as having more AI agency, which was further associated with higher perceived AI credibility. Perceived AI agency and AI anxiety sequentially mediated the effect of social status cues on perceived AI credibility. Results also showed that high (vs. low) message contingency significantly ameliorated users’ AI anxiety for high-status healthcare chatbots. Theoretical and practical implications are discussed.