- Education
- 21/05/2026
TU/e surveys on AI: knowledge and ethical use lag behind
Students are using AI on a large scale, but often lack the necessary knowledge and ethical and critical-thinking skills. That is the conclusion of two TU/e surveys on AI in education. At the same time, lecturers could make better and more frequent use of AI. Starting in September, the university will introduce AI labels for all assessments.
The surveys were intended as a “temperature check within the organization,” says AI & Education Portfolio Lead Tom van Woensel. The results did not surprise him, but mainly confirmed the impression he already had. “My gut feeling was already that staff are less advanced in their use of AI than students; that also comes out of the surveys.”
The same applies to the wide variety of tools students and staff use and the fact that students often do not know what the rules surrounding AI are. “On the other hand, staff do not always know either, but we are still developing those guidelines.” 36 percent of the 407 staff members surveyed do not know whether there are any formal guidelines for AI use.
Hallucinations
Both students and staff also appear to lack knowledge about the technology itself, for example about how AI works, its limitations, and the fact that it can hallucinate. According to Van Woensel, ethical use and the critical evaluation of output score relatively low as well.
These shortcomings cause students in particular to become too dependent on AI, the professor says. Students often seem skilled at generating text and code, but because they are not sufficiently aware of the system’s flaws and find it difficult to verify the output, they may start seeing hallucinations produced by the system as facts. This issue could arise for 59 percent of students.
Staff members, in turn, could make more and better use of AI in their courses, Van Woensel says. Above all, however, he believes they should think more carefully about assessment methods in relation to AI use. “For example, whether they still have the right form of assessment given the fact that AI now exists.”
AI labels
To provide clarity for both lecturers and students, the university will introduce AI labels for every form of assessment starting in September. Lecturers will explicitly indicate whether and in what way students are allowed to use AI during exams and assignments.
Van Woensel hopes the labels will make lecturers more aware of AI in their teaching and encourage them to adapt their assessment methods accordingly. Ultimately, students should no longer be able to misuse AI or commit fraud with it, simply because the assessment methods will no longer allow room for that.
The university wants to embrace the use of AI, but in a responsible way. Because the survey shows that there is still much to gain in terms of knowledge and critical evaluation, Van Woensel wants to quickly introduce learning modules on AI literacy and workshops on critical evaluation.
Then there are the tools students and lecturers choose to use. Van Woensel: “You see that all kinds of AI tools being used around the world are also being used by students and staff.” These include both paid and free versions. For example, 71 percent of students use the free version of ChatGPT, which according to Van Woensel poses major risks in terms of data privacy and intellectual property.
Own tool
One possible solution would be for the university to provide its own AI tool. That could also reduce inequality between members of the TU/e community using free tools and those using paid tools, since according to Van Woensel the latter produce much better results.
As for which tool the university could potentially offer, he has not yet made up his mind. What is certain, however, is that its quality must be high: “people will always choose the tool with the best output.”
This article was translated using AI-assisted tools and reviewed by an editor.
Discussion