ChatGPT for assessment feedback on a complex programming assessment

This innovative practice paper explores the integration of ChatGPT into the feedback process for university programming courses, addressing the growing demand for efficient and consistent grading and feedback. The research involved a cohort of postgraduate students enrolled in a compulsory Java programming module. ChatGPT was utilized to generate feedback, which was then compared with traditional feedback methods. The results demonstrated that ChatGPT could produce quick, accurate, and human-like responses, significantly reducing the workload for faculty. Feedback generation was generally well-narrated and detailed, although occasional technical errors required manual adjustments. A survey conducted with the student cohort revealed that 75% of respondents preferred ChatGPT-generated feedback over traditional methods, appreciating its detailed and constructive nature. However, some students expressed a need for more specific guidance, such as code snippets. Lecturers also noted that while the AI feedback was efficient, it sometimes lacked the empathetic tone of human-generated comments. Despite these challenges, the overall response from both students and faculty was positive, highlighting the potential of ChatGPT to enhance the feedback process in programming education.