Generative AI in educational processes: ChatGPT-4 in providing feedback to students’ written responses

Abstract

This article examined the use of ChatGPT-4 in offering written feedback to students regarding their open-ended responses in written exams. Proper feedback is crucial for learning, helping students to understand their strengths, identify areas for improvement, and devise strategies and practices for future learning. However, crafting detailed and systematic feedback is a time-intensive task for teachers. Consequently, there is growing interest in educational circles to leverage generative AI and Large Language Models (LLMs) like ChatGPT for facilitating feedback provision as part of adaptive learning environments. For this article, ChatGPT-4 was employed to generate feedback for university students’ written exams. It familiarized itself with evaluation guidelines, three short articles as learning materials in English and related 54 student responses, which varied in length from 24 to 256 words in English. Then it evaluated these responses and provided each student with personalized feedback that was on average 64 words long. The findings suggest that ChatGPT-4 has the potential for providing systematic and constructive feedback on students’ written exam responses. ChatGPT-4 need to be instructed with precisely crafted prompts to ensure the feedback is precise and consistent and aligns with teacher’s and educational institution’s objectives to support students in learning.