AI vs Human: A quasi-experimental study on ChatGPT and teacher feedback for English as a foreign language young learners’ writings

2025-9-15
Türkoğlu, Sena
The present study aimed to investigate the efficiency of ChatGPT-generated feedback by comparing it to teacher feedback on fourth-grade primary school EFL students’ writings. A quasi-experimental mixed-method design was used, incorporating quantitative analyses of writing scores and qualitative analyses from questionnaires (Likert-scale items and open-ended questions) and semi-structured interviews. Over ten weeks, students completed five writing tasks and received feedback from either ChatGPT or teachers, both using a four-category rubric. Questionnaires and interviews examined student perceptions, while statistical analyses (repeated-measures ANOVA and independent samples t-tests) explored differences in writing performance and perception variables such as usefulness, motivation, and clarity of feedback. Interviews were transcribed and coded in MAXQDA to identify themes. Findings showed both feedback types positively influenced writing progress, but in different ways. Teacher feedback offered consistent improvement, contextual appropriateness, an affective bond, accurate suggestions, and promotion of higher-order thinking. ChatGPT feedback was valued for immediacy and constant support, enhancing student motivation. Questionnaire and interview results indicated that teacher feedback was the preferred type. There remains a need for more longitudinal and comprehensive studies on AI feedback in young learners’ writing. Recommendations for future implications include guiding educational practice, informing EdTech design, supporting AI developers, and fostering early development of autonomous learning.
Citation Formats
S. Türkoğlu, “AI vs Human: A quasi-experimental study on ChatGPT and teacher feedback for English as a foreign language young learners’ writings,” M.A. - Master of Arts, Middle East Technical University, 2025.