Comparative effects of direct written corrective feedback and AI-generated reformulation on second language learners’ depth of processing and grammatical accuracy

This study employs depth of processing (DoP; Leow, 2015, 2020) to examine the comparative effects of unfocused direct corrective feedback (direct CF) and AI-generated reformulation on second language (L2) learners’ feedback processing and their grammatical accuracy in writing. Thirty-two first-year non-English majors at a private university were divided into two groups: one receiving direct CF (DF group) and the other receiving AI-generated reformulation (AGR group) on their narrative essays. While examining the respective feedback, the participants engaged in written languaging. Learners’ three writing texts (i.e., the first draft, a revision without accessing feedback, and a new piece of writing) were analyzed to assess the accuracy of the target grammatical feature (i.e., the past counterfactual conditional). Language-related episodes were identified in the learners’ written verbalization data to examine their DoP of the feedback on the target structure. The results indicate that the DF group exhibited deeper processing than the AGR group. The analysis of their written products indicates that both groups improved the accuracy of the target feature in their revisions. However, only the DF group showed a significant improvement from their first drafts to their new pieces of writing, while the AGR group did not show such significant improvement.