Designing and governing generative AI for language education: a testable sociotechnical framework (RAiLE)
TL;DR
By reframing GenAI adoption as a sociotechnical design-and-governance problem rather than tool uptake, RAiLE offers a pathway for building plural, accountable, and agency-preserving futures for language education.
Abstract
Generative artificial intelligence (GenAI) is rapidly becoming infrastructural in language education, supporting drafting, revision, translation, and interactional rehearsal. Yet most institutional deployments remain organized around personalization and efficiency logics that treat learning as an optimization problem. This article argues that personalization-first GenAI, when implemented without explicit pedagogical and governance constraints, may amplify three risks in language education: epistemic substitution (outsourcing judgement to the system), normative compression (privileging dominant varieties and genre conventions as “neutral”), and governance opacity (data capture and surveillance drift). Drawing on rhizomatic learning theory and recent work on epistemic authority redistribution in AI-mediated learning, the article proposes Rhizomatic AI for Language Education (RAiLE), a testable sociotechnical framework integrating pedagogy, system design, and governance. RAiLE specifies evaluable design commitments and indicators beyond narrow proficiency metrics and outlines a research agenda for empirical refinement through design-based, discourse-analytic, and policy studies. By reframing GenAI adoption as a sociotechnical design-and-governance problem rather than tool uptake, RAiLE offers a pathway for building plural, accountable, and agency-preserving futures for language education.