Beyond the AI feedback paradox: How metacognitive re-flection transforms AI corrections into L2 learning

Main Article Content

Phạm Huy

Abstract

Despite its importance in L2 writing classes, there has been difficulty in providing prompt and personalised feedback in large classes due to teachers' workload. On the other hand, generative artificial intelligence technologies, including ChatGPT, allow for immediate and consistent feedback on students' writing; however, its efficiency might lead to the emergence of the "AI Feedback Paradox," which can undermine the cognitive processes of deep processing and thereby prevent L2 acquisition. The quasi-experimental research aimed at investigating the possible solution of the problem, namely the possibility of overcoming the AI Feedback Paradox by adding structured metacognitive reflection to the AI-generated feedback. Ninety low-intermediate level Vietnamese EFL learners divided into three groups based on the writing classes received either AI-generated corrective feedback, teacher-generated metalinguistic feedback or AI-generated corrective feedback accompanied with metacognitive reflection log during a 12-week writing class. Students' grammatical knowledge was tested through Grammatical Error Identification Test, while the nature of revision was marked as deep or surface revision. The findings revealed significant superiority of the third condition over the first one concerning the increase of grammatical knowledge and the proportion of deep revisions performed by participants; the second condition proved to be between those two. Thus, the usefulness of the AI-generated feedback depends on the level of cognitive demands imposed on learners by the process.

Downloads

Download data is not yet available.

Article Details

How to Cite
Huy, P. (2026). Beyond the AI feedback paradox: How metacognitive re-flection transforms AI corrections into L2 learning. Research Studies in English Language Teaching and Learning, 4(4), 823–846. https://doi.org/10.62583/rseltl.v4i4.150
Section
Articles