Supporting student self-learning using generative AI

Yuki Ito, Qiang Ma

Abstract: With recent advancements in language-generating AI, students increasingly depend on AI for problem-solving in their coursework. However, over-reliance on AI risks depriving students of opportunities for self-learning, which is essential in education. One of the considerable reasons students turn to AI is their inability to solve problems independently due to a lack of understanding. To address this, we propose an interactive system that guides students struggling with problem-solving by giving them hints. When the system gets students’ answers, it assesses the answer by breaking down problems into small parts, then evaluating each part, and finally, explaining the grading rationale. Based on this assessment, the system offers targeted suggestions for improvement generated by language generation AI with pre-prepared templates. These hints guide students without providing direct answers, fostering class understanding and problem-solving skills.