AI in the lecture hall: How lecturers at TH Rosenheim use OneTutor didactically

An AI tutor does not deliver its value simply by being available, but by being deliberately integrated into teaching. In the first part of this series, three lecturers at TH Rosenheim explained why their university is using OneTutor as part of a research project. This second part shows what that looks like in the lecture hall in concrete terms: from group work with AI to learning how to prompt well and exam preparation.

Felix Höpfl, professor at the Faculty of Applied Health and Social Sciences, sums up his basic rule: “For me, it is very important that I think beforehand, in my teaching concept, about where I use OneTutor. So not simply saying, dear students, we now have a great tool, have fun.”

An example from his course Quality Management in Healthcare: In a group exercise, one group looks at a situation from the perspective of physicians, one from the perspective of nursing staff, and one from the perspective of management and investors. The groups then reflect the arguments they have developed with OneTutor and check whether they are coherent for the respective target group.

“With the verified arguments, it then goes into, for example, a fishbowl discussion. My students are much more confident in their argumentation because they have checked everything once again.” – Prof. Felix Höpfl, TH Rosenheim

He also uses OneTutor in the first-semester course on academic writing, after a brief introduction to how language models work and a compact prompting course. His conclusion: “Most students do not want to give it up anymore.”

Meike Töllner, professor of structural engineering, uses OneTutor in Reinforced Concrete Construction and Glass Construction. For her, a major didactic advantage lies in the fact that her students’ chats are visible to her: “I think that is a huge advantage compared to ChatGPT: The students’ chats are readable for me. This allows me to discuss with them that they are not really prompting properly and that, with a prompt that is too general, they can never get the right result from my materials.”

Wrong answers thus become learning opportunities. Töllner discusses with her students that AI is useful primarily where one can assess the answer, and not in areas that are beyond one’s own knowledge.

“With OneTutor, I discuss the AI itself with the students, so to speak. They should develop their media literacy, because they will use AI, and they should learn to do it properly. I now connect that with my module.” – Prof. Meike Töllner, TH Rosenheim

A practical side effect: Students find their way around the script significantly faster. In Reinforced Concrete Construction 2, Töllner has also added the script from Reinforced Concrete Construction 1, so that the chat automatically restores the connection to fundamentals from further back.

Holly Ott, professor of production management, observes usage especially during the exam period: “We see that students use the tool right before the exams and really appreciate the additional practice exercises.” What she particularly likes about the quiz format is the feedback: OneTutor explains for each answer why it was right or wrong and provides additional information.

Felix Höpfl tested the effect in a particularly critical scenario: repeat exams. At TH Rosenheim, most degree programs start only once a year. Anyone who postpones or does not pass an exam has to retake it in a semester in which the course is not being taught at all. Höpfl continued to make OneTutor, including his own deliberately difficult exam quiz, available to the students.

“For me, the result was very surprising: The grade average in the repeat exam was better than the grade average in the exam in the semester in which I taught myself.” – Prof. Felix Höpfl, TH Rosenheim

His initial skepticism that students might simply have every answer generated for them was not confirmed: “The students’ feedback was rather: No, I want to be able to do this myself.”

How much effort is it for lecturers?

Getting started is manageable. Meike Töllner: “It is done in one day. Uploading the materials takes half an hour.” Felix Höpfl needs 10 to 20 minutes per semester to maintain the teaching content, and invests somewhat more time in maintaining his quiz questions: “I think that with half a day of work, I prepared my semester with OneTutor.” And Holly Ott uses the automatically generated quiz questions as a basis, which she reviews and adjusts before release.

The lecturers also rate the collaboration with the OneTutor team positively. Höpfl: “I sent two emails to the team in the almost two years in which we have been working together. I experienced the tool itself as very mature.”

Conclusion

Whether group work with reflected arguments, prompting exercises using one’s own script, or exam preparation without an ongoing lecture: The lecturers at TH Rosenheim show that an AI tutor has the greatest impact when it is deliberately integrated into didactics. The effort required is low, and the effect for students is measurable. Or in Felix Höpfl’s words: The tool brings a lot of value to me and my students, and I enjoy it every day.