Kuan-Yi Ho
Graduate Student National Taichung University of Science and Technology Department of Multimedia Design Taichung City, Taiwan
E-mail：dwvsmedia02@gmail.com
Yu-Han Wang
Associate Professor National Taichung University of Science and Technology Department of Multimedia Design Taichung City, Taiwan
E-mail：han@gm.nutc.edu.tw
Abstract
Generative AI applications in EMI courses suggest that the interactive personalities of AI may influence learners’ emotions. This study explored how AI virtual teaching assistants with different personalities affect learners’ emotions across English proficiency levels. Previous studies have mainly focused on the learning outcomes or usage intentions of AI tools, with less attention to the relationships among AI assistant personality, English proficiency, and learning emotions. Therefore, this study developed extroverted and introverted AI virtual assistants and implemented them in an undergraduate EMI course in a design department. A total of 45 valid quantitative responses were collected, and semi-structured interviews were conducted with 13 students. The results showed a significant interaction effect between assistant personality and English proficiency on positive emotions. Higher-proficiency students reported significantly higher positive emotions after interacting with the introverted assistant than with the extroverted assistant, whereas lower-proficiency students showed no significant differences in positive or negative emotions between the two assistant conditions. Qualitative results indicated that higher-proficiency students valued response efficiency and information clarity, while lower-proficiency students showed inconsistent preference patterns. This study contributes to emotional research on AI personalization and learner ability differences, suggesting that AI for EMI should be designed according to language proficiency, task type, and response quality.
Keywords：Chatbots, Artificial Intelligence, EMI, Learning Emotions, Virtual Teaching Assistant