Teacher, peer, or AI? What students prefer is not what helps them most
AI-generated feedback is cheap and always available. Before it replaces anything in a university course, we should know how it compares with feedback from teachers and from fellow students.
What we did
We ran a randomized field experiment with 90 bachelor students in a lecture course at the University of Zurich. Each student wrote an assignment and then received feedback from one of three sources: a teacher, a peer, or a large language model. All three used the same criteria and the same template.
Two design choices matter for reading the results. Students did not know which source their feedback came from. And every student gave feedback to a peer before receiving any, so the learning that comes from giving feedback could not explain differences between the groups.

What we found
Students judged teacher feedback as less fair and harder to accept than peer or AI feedback. Those who received teacher feedback were also less willing to revise their work. Ratings of usefulness did not differ between the three sources.
The quality of the revised work showed the opposite pattern. Teacher feedback led to the strongest improvement in scientific argumentation, clearly ahead of AI feedback (d = 0.82). For formal quality, both teacher and peer feedback outperformed AI feedback. AI feedback produced the smallest improvement overall.
Who the student is also mattered. Students with more productive attitudes toward feedback, measured with the SFLI, got more out of teacher feedback for their argumentation. Intrinsic motivation shaped how willing students were to revise after AI feedback.
What it means
How students perceive feedback and what it does for their work can point in different directions. Demanding feedback may feel less fair and still help the most. Peer and AI feedback can complement teacher feedback, but for higher-order skills such as argumentation they did not replace it in this study. Helping students build feedback literacy is one way to make demanding feedback land.
Weidlich, J., Gotsch, F., Schudel, K., Marusic-Würscher, C., Mazzarella, J., Bolten, H., Bütler, D., Luger, S., Wohlfender, B., & Maag Merki, K. (2025). Teacher, peer, or AI? Comparing effects of feedback sources in higher education. Computers and Education Open, 9, 100300. https://doi.org/10.1016/j.caeo.2025.100300