An Investigation into ChatGPT-Based Instructional Feedback on Programming Projects

Providing high-quality instructional feedback is essential for enhancing learning and motivation. However, the large number of learners and time constraints of instructors can hinder this process. Automating instructional feedback through online learning environments presents a potential solution, but current systems may not adequately address complex learning assignments. Artificial intelligence technologies like ChatGPT can offer an opportunity to deliver feedback efficiently, especially for complex learning tasks. Despite its potential, relatively few studies have examined the distinctions between ChatGPT-generated and expert-based feedback and learners' perceptions of these feedback sources. This study stands among the limited research efforts that seek to compare instructional feedback produced by ChatGPT and that provided by experts, shedding light on learners' subjective assessment of such feedback in the context of programming projects. The findings showed that experts consistently rated programming projects higher than ChatGPT. However, ChatGPT demonstrated strengths in providing more comprehensive textual feedback on the learners' work. Notably, the learners rated ChatGPT feedback more favorably than feedback from domain experts. The study's findings are thoroughly discussed, and future research avenues are outlined.