<![CDATA[As generative language models like ChatGPT and Google Gemini gain prominence in education, their efficacy in specific contexts, such as Indonesian English as a Foreign Language (EFL) instruction, still needs to be explored. This study investigates the pedagogical affordances and constraints of these models as perceived by Indonesian EFL sophomores, aiming to understand their contribution to active learning in language acquisition. Using a qualitative approach, we conducted open-ended questionnaires with 40 sophomore students from an Indonesian university's English department. Thematic content analysis was employed to analyse the data. Findings reveal that ChatGPT offers authentic conversational simulations and versatile content-based instruction, while Google Gemini's strength lies in its multilingual capabilities. However, limitations such as linguistic complexity and rigid conversational structures were also identified. The study suggests these models can enhance active learning experiences, particularly in conversational practice and interdisciplinary content exploration, though their efficacy depends on factors like learner proficiency and internet access. We conclude that integrating these models into EFL instruction requires careful consideration of their affordances and limitations. This study contributes culturally-specific insights to AI in education research, with implications for curriculum designers, educators, and policymakers in developing countries, emphasising the need for adaptive and inclusive approaches in AI-enhanced EFL education.]]>
Copyrights © 2025
