The rise of AI chatbots has enriched the scenarios for practicing communication in a foreign language. Some pioneering studies investigated the relationships between learner's willingness to communicate (WTC) and communication anxiety (CA) in such AI-supported interactions compared with those in traditional human-human interactions, yet it remains underexplored how individual learner's WTC and CA fluctuate and correlate in real time and what factors contribute to such dynamics across the two interaction contexts. To address this issue, this study adopted the idiodynamic approach to capture moment-to-moment variations in L2 WTC and CA among five university students engaged in structured dialogues with both an AI chatbot (Call Annie) and a human interlocutor. Quantitative correlation analysis revealed three distinct WTC-CA patterns (negative, positive, and near-zero) across individuals and contexts, while qualitative interviews informed a framework which highlighted contextual and individual factors in shaping these dynamics, with most of them functioning differently across contexts. The findings underscore the significance of individual differences and contextual variations in influencing the real-time dynamics of WTC and CA. This study offers pedagogical implications by recommending a human-human and human-chatbot collaborative learning mode that integrates the efficiency of AI chatbots with the depth of human interaction.
Liu et al. (2025) studied this question.