Description
This study investigated how three AI reliance subtypes — Idea Generation Reliance (IGR), Linguistic Delegation (LDE), and Evaluative Passivity (EVP) — alongside Affective Response (ARS), differentially predict writing gains among 113 undergraduate EFL learners in Vietnam. Using a pre-test/post-test design with the purpose-built AI Dependency in Academic Writing Scale (ADAWS), multiple regression analyses identified IGR as the sole statistically significant independent predictor of reduced writing gains, with pronounced effects on grammatical accuracy (β = −0.358, p = .005) and idea development (β = −0.448, p = .004). LDE and EVP became non-significant once IGR was controlled. High-IGR learners demonstrated significantly lower gains than low-IGR peers. These findings suggest that delegating ideation — rather than surface-level editing — to AI constitutes the primary developmental risk, with direct implications for AI literacy pedagogy and EFL writing assessment.
Thông tin các tác giả
Nguyễn Thanh Thủy: ThS., đang công tác tại Trường Đại học Bách Khoa TP.HCM, 268 Lý Thường Kiệt, Phường Diên Hồng, TP.HCM, email: nguyenthanhthuy@hcmut.edu.vn
Từ khóa
AI over-reliance; EFL writing; idea generation reliance; linguistic delegation; generative AI; ADAWS
