Exploring Teenagers' Trust in Al Chatbots: An Empirical Study of Chinese Middle-School Students
Siyu Qiu, Anqi Lin, Shiya Wang, Xingyu Lan
TL;DR
This study investigates how four developmental traits—AI literacy, ego identity, social anxiety, and psychological resilience—shape Chinese middle-school students' trust in AI chatbots. Using a mixed-methods design with an online survey (N=152) and 15 semi-structured interviews, the authors find that psychological resilience significantly predicts trust, while AI literacy, ego identity, and social anxiety do not show robust direct effects; age moderates the negative influence of social anxiety on trust among older teens. Qualitative results reveal that teens typically trust AI in learning contexts and treat it as a tool rather than a confidant, with trust shaped by perceived functionality, self-report biases in literacy, and social-media narratives. The work offers practical guidance for educators and policymakers to foster healthy, context-aware teen–AI interactions and highlights avenues for future research on adolescent AI use, moderation effects, and bidirectional trust dynamics.
Abstract
Chatbots have become increasingly prevalent. A growing body of research focused on the issue of human trust in AI. However, most existing user studies are conducted primarily with adult groups, overlooking teenagers who are also engaging more frequently with AI technologies. Based on previous theories about teenage education and psychology, this study investigates the correlation between teenagers' psychological characteristics and their trust in AI chatbots, examining four key variables: AI literacy, ego identity, social anxiety, and psychological resilience. We adopted a mixed-methods approach, combining an online survey with semi-structured interviews. Our findings reveal that psychological resilience is a significant positive predictor of trust in AI, and that age significantly moderates the relationship between social anxiety and trust. The interviews further suggest that teenagers generally report relatively high levels of trust in AI, tend to overestimate their AI literacy, and are influenced by external factors such as social media.