Deep Learning-Assisted Feedback in Academic Writing Effects on EFL Students’ Revision Skills

Abstract

Background: The increasing use of artificial intelligence in higher education has expanded the role of deep learning-assisted feedback in academic writing instruction, particularly in EFL contexts. Although prior studies have shown its potential to improve writing performance, limited research has examined how such feedback influences revision skills through student engagement and digital literacy, especially in regional higher education settings.

Purpose: This study investigated the effect of deep learning-assisted feedback on EFL students’ revision skills in academic writing, the revision types most influenced by the feedback, and its contribution to the improvement of writing quality.

Methods: A quantitative correlational design was employed involving 22 undergraduate students in an academic writing course at Universitas Muhammadiyah Enrekang, Indonesia. Data were collected through questionnaires, revision coding of students’ drafts, and writing quality scores before and after revision. The data were analyzed using Partial Least Squares Structural Equation Modeling (PLS-SEM), supported by descriptive statistics of revision types and writing improvement.

Results: The findings showed that deep learning-assisted feedback had a positive and significant direct effect on revision skills and an indirect effect through student engagement. Digital literacy also strengthened the relationship between feedback and engagement. Surface-level revision was the most frequent type of revision, followed by meaning-level and structural revision. In addition, students’ writing quality improved substantially after the revision process, as indicated by the increase from a mean pre-revision score of 64.64 to a mean post-revision score of 84.00.

Conclusion: Deep learning-assisted feedback contributed positively to EFL students’ revision skills and writing quality. Its effectiveness was enhanced by student engagement and digital literacy, suggesting that AI-assisted feedback can function as a meaningful pedagogical support in academic writing, particularly in emerging regional higher education contexts.