Reclaiming or Relinquishing? Learner Agency in GenAI-Mediated EFL Writing at a Chinese Higher Vocational College
Lei Cao, Amelia AbdullahPurpose
This study examines how generative artificial intelligence (GenAI) reshapes learner agency in EFL writing at a Chinese higher vocational college. Grounded in Bandura’s social cognitive theory, it analyzes the writing process through four agentic properties (intentionality, forethought, self-reactiveness, and self-reflectiveness) and the triadic reciprocal (P–B–E) model to consider how personal, behavioral, and environmental conditions shape autonomy, self-efficacy, and ownership.
Design/Methodology/Approach
Twelve first-year EFL learners at a vocational college in eastern China were purposively selected for contrasting GenAI engagement patterns. Data from an eight-week cycle, comprising interviews, reflective journals, classroom observations, and GenAI interaction logs, were analyzed through case-based reflexive thematic analysis, structured around three provisionally identified engagement patterns that were tested and refined across cases, to trace how agency was reclaimed, negotiated, or relinquished.
Findings
Three patterns recurred. Strategic Orchestrators used GenAI selectively, keeping authorship; Reactive Optimizers fell between, controlling output mainly after it appeared; and Passive Reliants ceded judgment with little self-regulation. GenAI magnified learners’ prior self-beliefs rather than equalizing them, and fluent output could deepen dependence while masking disengagement.
Originality/Value
The study reframes the “digital divide” in GenAI as psychological rather than technical, showing that better textual quality can conceal weak agency. It argues for teacher-supported use, in which affective validation and process-oriented assessment enable GenAI to serve as a genuine scaffold rather than an exit from active learning.