摘要
This study investigates how generative artificial intelligence (AI), particularly ChatGPT, is positioned and utilized by Taiwanese graduate students in cross-linguistic academic contexts, and explores its pedagogical implications in higher education. The participants were 25 master's and doctoral students who attended a one-month overseas academic program in the United States in 2023. Data were collected through semi-structured interviews and participant observation, and analyzed using grounded theory with open, axial, and selective coding.Findings reveal that ChatGPT was not merely perceived as a technical learning tool, but increasingly as a co-learning partner mediating translation, conceptual clarification, and language production. Its use significantly reduced students' language anxiety and cognitive load, thereby enhancing classroom participation and learning confidence. However, concerns also emerged regarding cognitive offloading, weakened deep processing, and blurred ethical boundaries in learning practices. At the same time, teachers' roles were observed to shift from knowledge transmitters to learning designers and ethical facilitators guiding students toward responsible and reflective AI use.Based on these findings, this study proposes three key implications: (1) generative AI should be positioned as a cognitive mediator rather than a substitute for human learning, (2) both teacher AI literacy and students' critical verification abilities should be strengthened, and (3) institutional governance frameworks should be established to ensure ethical and sustainable AI integration. Overall, the study contributes to rethinking human-AI co-learning and cross-linguistic academic practices in the emerging AI era. © 2026 IEEE.