Integrating AI-Based Feedback into EFL Writing Instruction: A Case Study from a Vietnamese University

Abstract

There exists inadequacy of traditional feedback mechanisms in EFL writing instruction within Vietnamese higher education contexts. This research aims to investigate the effectiveness and feasibility of implementing AI-based feedback systems in enhancing Vietnamese university students' EFL writing skills. Mixed-methods sequential explanatory design was employed over a 16-week period. Random sampling was used to assign 384 Hanoi Open University second and third-year undergraduate students to experimental (AI feedback, n = 192) and control (traditional teacher feedback, n = 192) groups. Complexity, Accuracy, and Fluency (CAF) measures were used to evaluate argumentative writing tests administered before and after intervention, supplemented by semi-structured interviews and technology acceptance questionnaires. The experimental group demonstrated significantly larger improvements in all CAF dimensions, with significant gains in fluency (d = 0.51-0.61), moderate effects for complexity (d = 0.45-0.49), and large effect sizes for accuracy (Cohen's d = 0.66-0.71). Students showed a 5.6-point decrease in errors per 100 words and an 18.2% increase in error-free T-units specifically. Although there were limitations in terms of cultural sensitivity (65% of participants) and enduring needs for human interaction (87.5% of participants), student acceptance was high (M = 3.97/5.0). Results demonstrate that AI-based feedback systems significantly enhance EFL writing when integrated with traditional teaching methods, providing scalable solutions for large university settings while requiring targeted implementation strategies and comprehensive teacher training in hybrid pedagogies.