AI’s Corrective Feedback vs. Traditional Recasts: which fosters better grammar accuracy?

This paper discusses how effective ChatGPT, developed by OpenAI, is likely to be compared to traditional corrective recasts in teaching English grammar to engineering undergraduates. With the rise of digital communication, a generation of learners for whom traditional methods seem to be less effective has emerged. This research contrasts the effects of ChatGPT's corrective feedback versus conventional teacher-led corrective recasts on grammatical proficiency among students, especially in subject-verb agreement and accuracy in tenses. 90 engineering students were divided into two groups ᅳ one with the traditional approach to instruction and another using ChatGPT to practice grammar. Pre- and post-assessment tests were administered to the participants to measure improvements in grammar proficiency. Statistical analysis, using paired t-tests, indicated that both groups improved significantly, with the ChatGPT-assisted group showing more improvement in proficiency compared to the traditional approach. Another independent sample t-test confirmed that the AI-led feedback group had a significantly larger mean improvement in grammar proficiency compared to the traditional instruction group. The results of this study demonstrate that ChatGPT provides more effective corrective feedback to engineering students with varying levels of proficiency in subject-verb agreement and tenses usage, suggesting potential benefits for language education through integration with AI.