What Students Think About AI and What That Means for Teachers
Two studies published in quick succession this week together reveal an interesting pattern. One looks at how students use AI in their learning process, the other at how teachers deploy AI in ways that actually enhance learning. Two separate perspectives, but together a complete picture.
Four ways in which students interact with AI
Researchers Bente Cerpentier, Cune van Asselt and Paul Ketelaar published the study "Between Tool and Taboo" in Examens, based on interviews with thirteen master's students at Radboud University. They distinguish four personas: the critical user who consciously employs AI as a supplement, the convenience seeker who is mainly driven by efficiency, the social follower who goes along with what the group does, and the ambivalent doubter who is torn between personal growth and peer pressure.
Notably, the students are perfectly able to identify the risks of AI, such as unreliable output, plagiarism, and dependency. However, in practice, this knowledge carries little weight. Usage is often justified by framing it as "supportive", and peer pressure proves to be a stronger driver than the risks themselves. The researchers advocate for institution-wide policy, AI literacy for both students and teachers, and a learning culture in which risks and opportunities are openly discussed.
The right prompt starts with the learning objective, not with AI
Where the Radboud study starts with the student, Hans Visser starts with the teacher. In a post about the opening of the ZAAM Education Academy in Amsterdam, he explains that a teacher who knows the learning objective, the learning phase, and the student's own step, automatically knows which prompt is appropriate. His example: a student is learning to write an argumentative essay. The structure is already familiar to them; selecting and weighing arguments is the step they need to take themselves. AI is therefore not tasked with writing sentences, but with asking questions about those arguments and pointing out the weak spots.
Visser refers to Van Damme's (2026) scoping paper on AI in education, which draws the same conclusion: good use of AI in the classroom requires knowledge of how learning works, combined with knowledge of how AI works. Standalone AI skills are not enough.
Two sides of the same lesson
If you place these two studies side by side, a logical chain emerges. Students know what the risks are, but need guidance to act accordingly. That guidance does not arise by itself, but comes from a teacher who has a clear learning objective and deploys AI purposefully, as Visser demonstrates.
This is precisely what we are focusing on at EduGPT. The teacher retains control over the learning objective and learning phase, AI asks questions and provides challenges instead of taking over the work, and the conversation about risks and boundaries is part of the tool, not separate from it. We will also address this theme in the webinar we are organising together with the Flemish AI Academy on 30 September about critical thinking and AI.
Sources
Cerpentier, B., van Asselt, C. & Ketelaar, P. (2026). Between Tool and Taboo: Insights into Students' Perspectives on Generative AI in Higher Education. Examens.
Visser, H. (2026). LinkedIn post about the opening of the ZAAM Education Academy.
Van Damme, D. (2026). Artificial Intelligence in Education. A scoping paper.