Examining the Efficacy of ChatGPT and Human-Derived Corrective Feedback in Addressing Grammatical Errors in Saudi EFL Students' Compositions

Abstract

Employing an explanatory mixed-methods approach, this study compared the affordances of ChatGPT’s written corrective feedback (WCF) and human corrective feedback in fostering linguistic accuracy among Saudi EFL learners. Participants were 53 undergraduates at the University of Jeddah, divided into an experimental group (n = 26) and a control group (n = 27). Over eight weeks, both groups engaged in two 1.5-hour writing sessions per week. The experimental group received WCF from ChatGPT, while the control group received human WCF. Pretests and posttests assessed participants’ grammatical competence before and after the intervention. Error-flagging accuracy was evaluated by a panel of experienced university teachers of English. Semi-structured interviews with six experimental-group participants explored perceptions of ChatGPT. Quantitative findings indicated that ChatGPT was a more effective source of WCF than human feedback, as the experimental group significantly outperformed the control group in the posttest. However, human feedback was slightly more accurate in error-flagging. Qualitative results revealed that most students valued ChatGPT for its precise error identification, though some considered human feedback clearer in explanation. The study concludes that ChatGPT can serve as a valuable supplementary tool for corrective feedback in Saudi EFL writing.