Student experiences with static versus adaptive AI feedback on self-explanations in computer science courses

Feedback is a critical component of formative assessment yet remains underutilized in higher education STEM courses due to large class sizes and instructor workload. Emerging artificial intelligence (AI) technologies are reshaping STEM education by enabling scalable, personalized feedback and promoting active learning. This study explores the instructional potential of AI-supported feedback in addressing these challenges, with a particular focus on self-explanations. It reports on two iterations of a classroom response system in an undergraduate computer science (CS) course across two semesters. The first iteration provided static feedback based on instructor-prepared sample responses, while the second incorporated AI-generated adaptive feedback powered by natural language processing. Using a comparative case study design, the authors investigated differences in student self-efficacy, engagement, and perceptions of system usability. Both implementations sustained student engagement and participation in CS courses, demonstrating the system’s effectiveness in fostering active learning. Findings highlight the potential of AI-driven feedback to enhance scalability and interactivity, deepen conceptual understanding, and reduce instructor workload. These results underscore the promise of AI-driven feedback systems for enhancing interactivity and learning in STEM courses, particularly in CS contexts with growing enrollments. Future work should explore question frequency, question formats, instructor’s perspectives and pedagogy with feedback tools, and improve AI accuracy to reduce over-reliance.