AI vs. Human Feedback in EFL Writing: Iranian Learners’ Perceptions

Abstract

This research investigates the perceptions of Iranian EFL learners regarding AI-driven corrective feedback, specifically through Grammarly, in comparison to traditional human feedback for enhancing writing skills.  Utilizing an embedded mixed-methods design, the study involved 120 participants who completed quantitative surveys, alongside qualitative interviews with 15 students from each feedback group. A questionnaire by Huang and Renandya (2020) assessed attitudes towards AI feedback, focusing on effectiveness, usability, and emotional responses. Additionally, a questionnaire adapted from Ushba Rasool et al. (2023) gathered insights on students' perceptions of written corrective feedback. Quantitative results showed a significant improvement in learners’ attitudes toward AI feedback (Z = –5.72, p < .001), with mean perception scores rising from 34.19 to 42.40 (approx. effect size r = .76). In the human feedback group, metalinguistic feedback perceptions increased (Z = –5.58, p < .001), while negative perceptions decreased (Z = –3.82, p < .001). Overall, AI feedback was valued for immediacy and motivation, while human feedback remained preferred for depth and clarity.