Who Decides in AI-Mediated English Teaching? Distributed Agency and the Reconstruction of Teachers' Professional Roles

Abstract

When English teachers began using generative AI, researchers asked sensible first questions: do teachers accept it, what do they think of it, how do they use it? These questions have produced useful descriptions, but they leave something important untouched. Each time a chatbot drafts a reading passage, corrects an essay, or decides which version of English sounds natural, a pedagogical decision has moved from one place to another. This conceptual article asks who now makes those decisions and what that means for teachers' professional agency. Drawing on the ecological model of teacher agency and on a critical reading of recent research on language teachers and generative AI, the author proposes a framework of distributed and contested agency. It distinguishes five domains of pedagogical decision and four mechanisms through which agency is reshaped: silent delegation to a model's defaults, an evaluation paradox, institutional foreclosure or manufacture of agency, and the entanglement of teacher and learner agency. Four propositions follow and are examined against cases from Vietnam, Hong Kong, South Korea, and China. The cases suggest that reluctance can be a professional judgement, that training in tool operation does not build the judgement teachers need, and that mandates and bans can both sidestep teachers. The article questions technology acceptance as an explanation of classroom practice and the "teacher in the loop" as a safeguard, and it draws implications for teacher education, policy, tool design, and research in Asia and in other systems where English is taught under scarcity and high-stakes assessment.