AI math feedback looked good. Did students learn more?
Good-looking exercises are not proof of better learning. Reviewers rated AI-made math practice more favorably than the feedback’s clarity and ability to spot errors. Later exam comparisons found no clear difference, and the study could not separate feedback from attendance or practice.
What the source reports
The paper looks at AI feedback on math work in two groups of college students. One group had 343 students. The other had 236. Reviewers rated 157 feedback notes from three tasks in the first group. They liked the practice tasks more than the feedback on errors, tone, and clarity. Four adjusted exam checks found no clear gap. The study could not tell if feedback, class time, or practice drove later scores. It does not show a gain for K-12 students.
Original source
- Title
- Evaluating LLM-Generated Feedback on Handwritten Mathematics: Feedback Quality and Subsequent Exam Performance
- Author
- Arne Vanhoyweghen, Vincent Holst, Melika Mobini, Lukas Van de Voorde, Tibo Vanleke, Andres Algaba, Sam Verboven, Marie-Anne Guerry, Filip Van Droogenbroeck, Vincent Ginis
- Publication
- arXiv
- Date
- Wednesday, October 21, 2026
Also covered by the other newsletter.