Enjoyment, anxiety, and breakdown fluency when Chinese EFL learners speak with AI chatbots: An idiodynamic approach

The rise of AI (Artificial Intelligence) chatbots has witnessed its significant application in second language acquisition. Recently, studies have focused on how the use of AI chatbots influences emotions, especially enjoyment and anxiety. However, previous studies have not focused on the dynamic emotional fluctuations and granular factors that occur during L2 speakers interacting with AI chatbots. Therefore, this study adopts an idiodynamic approach to investigate the relationship between enjoyment and anxiety, as well as the reasons behind their dynamic change, during the speaking process of five Chinese English as a Foreign Language (EFL) students with two AI chatbots: ELSA Speak (Anguera & Van, 2016) and Chat Generative Pre-trained Transformer (ChatGPT). Additionally, the relationship between breakdown fluency and the two variables is examined to determine if emotional states affect real-time language fluency. The result showed that even though we detected most of the negative relationships between enjoyment and anxiety, the positive relationship also appears, which illustrates that their complex relationship is affected by intricate factors. We found that a few participants demonstrated significant correlations between breakdown fluency and enjoyment/anxiety, which, triangulated with the interview, indicates a reciprocal influence; however, its significance depends on individual differences. Furthermore, the psychological processes of each individual were mapped, and the recurring factors were identified, which offer significant pedagogical insights into the use of AI as a tool for Foreign Language (FL) oral practice.