Exploring AI to automate EFL corrective written feedback in the first language
Rob Hirschel, Kayoko Horai · Technology in Language Teaching & Learning · 2025
With the advent of ChatGPT and other AI programs using large language models, it is now relatively easy to provide automated written corrective feedback in a student’s native language. This paper reports on an exploratory study using a ChatGPT-powered plugin currently in development for the popular Moodle learning management system. The classroom intervention on which this study is based had four main steps: a) individual or collective brainstorming and vocabulary search (5 minutes), b) subsequent free-writing activity in an online browser (10 minutes), c) reading the ChatGPT feedback in L1 Japanese (5 minutes), d) completing paper error logs to process feedback (5 minutes). Data was collected from students’ written submissions as well as the ChatGPT-generated feedback, students’ error logs, and surveys administered to students after each activity and at the end of the semester. Analysis suggested that both students and teachers appreciated the grammatical feedback, found it to be relatively accurate, and largely comprehensible. There is room, however, for improvements in accuracy and presentation. Pedagogical implications are discussed.