Yong-Bing Xiang, Ruolei Gu, Zhuo Chen, Tian Gan, Yan Zheng
The relationship between humans and AI has evolved into a phase of deep integration, with the latest generation of text-based chatbots increasingly serving as trusted companions for emotional support. Relational cues (features in chatbots designed to simulate human emotional and social behaviors) reflect the core characteristics of human social existence and play a crucial role in enhancing human-AI emotional interactions. This research employs a mixed-methods approach to explore how relational cues shape human-AI emotional interaction experiences. Study 1 selected common indicators of emotional interaction experience and used an experimental method (n = 196), which showed that relational cues significantly increased users’ emotional arousal, system usability, emotional attachment and human-AI trust. Moreover, mind perception exerted a negative moderating effect on the relationship between relational cues and pleasure, arousal, dominance, emotional attachment and usability. Furthermore, the efficacy of specific types of relational cues (particularly empathy) was stronger in extreme negative emotional events compared to moderate ones. Study 2 used an online survey method (n = 25) and further revealed that relational cues can improve mood, enhance perceived realism, foster intimacy, and elevate conversational quality. These improvements collectively contributed to enhancements in multifaceted emotional interaction experiences. It is worth noting that empathy cues exhibited the most positive effect in both studies. These findings highlight the beneficial role of relational cues in human-AI emotional interactions and their boundary conditions, providing a theoretical foundation for emotional agent design from the perspective of chatbot, users, and event.