Enhancing apology sincerity in AI bots: The role of anthropomorphic cues, perceived experience, and AI literacy

초록

AI bots' apologies are frequently perceived as programmed and mechanical, which makes them less satisfying than human agents. This study utilized online experiments to investigate the impact of anthropomorphic visual and linguistic cues in the contexts of perceived experience and sincerity across airline (Experiment 1), retail (Experiment 2), and food delivery (Experiment 3). The results indicate that human-like images (visual cues) and empathetic language (linguistic cues) increase bot usage intention, with perceived experience and apology sincerity as mediators. Notably, AI literacy moderated linguistic effects: empathetic wording proved more effective among users with lower literacy, who perceived it at genuine, whereas users with higher literacy regarded it as scripted. The findings demonstrate that there is a sequential mechanism in symbolic recovery, where visual cues enhance perceived emotional capacity, whereas linguistic cues enhance sincerity. Theoretically, this reveals how AI service recovery works through an emotional chain mechanism. Practically, the study suggests that chatbot designers should prioritize evidence-based sincerity—clear facts, responsibility, and remedies—while also using empathic cues to enhance perceived emotional capacity. These results underscore the significance of context-sensitive design and identify AI literacy as a critical factor influencing the effectiveness of anthropomorphic cues in service recovery. © 2026 Elsevier Inc.

키워드