Calibrating AI Assistance in EFL Reading: The Graduated AI-Mediated Reading Support Framework

Abstract

Generative artificial intelligence (GenAI) is increasingly being used to support second and foreign language reading through vocabulary explanation, simplification, summarization, background-knowledge provision, question generation, and feedback. These capabilities can simultaneously improve access to challenging texts and support reading development or substitute for the cognitive processes underlying independent reading competence. This article develops the Graduated AI-Mediated Reading Support (GAIRS) framework to address this tension in EFL reading. Drawing on research on reading cognition, instructional scaffolding, productive difficulty, cognitive offloading, learner autonomy, and GenAI-mediated learning, GAIRS distinguishes assistance that reduces barriers while preserving learner cognitive responsibility from assistance that performs interpretive, inferential, integrative, or metacognitive work for the learner. Three dimensions anchor the framework: contingency (when help is given), directiveness (how much cognitive work the support performs), and fading (how responsibility progressively shifts back to the learner). These dimensions inform four graduated configurations: independent attempt, responsive prompt, targeted explanation, and collaborative interpretation. GAIRS shifts the focus from the amount of assistance to how it sustains access while progressively expanding cognitive responsibility.