The emotional roller coaster of talking to the AI chatbot: an idiodynamic investigation of fluctuations of EFL learners’ enjoyment, anxiety, and boredom

Despite the growing integration of AI into foreign language education and the recognised centrality of emotions in language learning, little is known about the real-time emotional dynamics that unfold when learners speak with AI chatbots. To address this gap, this study pioneers the use of the idiodynamic method to investigate the dynamics of and relationships between enjoyment, anxiety, and boredom among six Chinese university EFL learners in AI-mediated L2 speaking practice. The Anion Variable Tester V2 software was used to capture learners’ real-time emotional ratings. Stimulated recall interviews were subsequently conducted to elicit the triggers and mitigators of emotional changes. The results revealed highly individualised emotional trajectories, with enjoyment generally predominating over negative emotions. The moment-to-moment relationships between the three emotions were dynamic, frequently shifting between positive and negative correlations within the same individual. Qualitative analysis identified four factors shaping emotional fluctuations: perceived usefulness of AI’s feedback, perceived competence and performance, emotional relatedness with the AI agent, and smoothness of communication. Interpreted through the lens of control-value theory, these factors operate via learners’ real-time appraisals of control and value. Based on the findings, pedagogical implications and directions for future research on dynamic emotion-appraisal processes in technology-mediated L2 learning are provided.