Human–AI Collaboration in Task-Based Language Learning: A Lesson from EFL Students’ Experiences of Using Gemini for Academic Writing
DOI:
https://doi.org/10.36312/jllet.v1i2.75Keywords:
Academic writing, EFL learners, Google gemini, Human–AI collaboration, Task-based language learningAbstract
The increasing integration of generative artificial intelligence into EFL writing has created new opportunities for linguistic scaffolding while raising concerns about learner autonomy, reasoning, and authorship. This qualitative study explored how EFL students experienced Google Gemini during planning, drafting, and revision in task-based academic writing and examined how such collaboration shaped confidence, autonomy, reasoning, and perceptions of authorship. Fifteen third-semester students from the English Language Education Study Program at Universitas Pendidikan Mandalika were purposively selected based on their experience with online academic writing and Gemini-assisted tasks. Data were collected through semi-structured interviews and analyzed thematically through familiarization, initial coding, category development, theme refinement, and interpretation. The findings showed that Gemini performed stage-dependent roles: it functioned as an idea generator and organizational scaffold during planning, a linguistic and rhetorical resource during drafting, and an immediate feedback and revision partner during revision. Students reported increased confidence and easier access to writing support; however, these benefits were accompanied by tensions involving technological dependence, cognitive offloading, loss of personal voice, and uncertainty regarding authorship. Autonomy and reasoning were strengthened when students compared, questioned, modified, or rejected AI suggestions but weakened when outputs were accepted without critical evaluation. The study concludes that Gemini is most educationally valuable when used as a scaffold rather than a substitute for learner cognition. Task-based writing pedagogy should therefore structure AI use around critical evaluation, reflective decision making, verification, and explicit learner responsibility for the final text.






