Empowering EFL Learners Through Generative AI: A Qualitative Exploration of Inclusive and Autonomous Learning in Multicultural Contexts

Abstrak

Generative artificial-intelligence (GenAI) systems are rapidly permeating language education, yet their effects in multilingual, resource-varied classrooms remain under-examined. This 14-week qualitative multiple-case study investigated how GenAI mediates learner autonomy, inclusive participation, multicultural identity work, and teacher practice in four Indonesian EFL settings (three senior-high schools and one university). Data comprised 24 hours of classroom observation, 76 semi-structured interviews, four focus-group discussions, and 6 752 learner-AI log entries. Reflexive thematic analysis produced four inter-locking themes. AI-mediated autonomy emerged as students progressed from copying teacher prompts to crafting and iterating their own, evidenced by a 142 % rise in self-initiated prompting and increased lexical diversity (type–token ratio =.58). Inclusive voice was fostered through translanguaging cycles, text-to-speech, and complexity-controlled outputs, elevating the unique-speaker index from.54 to.82. In multicultural identity negotiation, learners embedded local idioms (e.g., sipakatau) and deployed iterative counter-prompts to correct Western-centric bias, averaging 2.1 revisions per cultural task. Pedagogical re-positioning saw teacher talk-time fall from 61 % to 34 %, as instructors shifted from grammar transmitters to AI-literacy mentors who facilitated prompt-engineering and bias-hunt workshops. These benefits were conditional on equitable connectivity and critical-AI scaffolding; absent such supports, GenAI risked reinforcing dependency and cultural erasure. The study advances a blended sociocultural–critical-digital-literacy framework, offering practical design principles (prompt workshops, bias-mitigation routines) and policy guidance (minimum-access packages) for the equitable integration of GenAI into language education.