We spend a lot of time discussing whether students should be allowed to use AI. I think we need to move that discussion forward. AI offers many opportunities for education and research, and banning it across the university is neither realistic nor desirable. Our students will encounter these tools in their professional lives. We have a responsibility to help them use them well.
At the same time, I do understand the concerns. If a student uses AI to formulate the problem, develop the argument and write the conclusions, what has that student learned? We cannot answer this by looking only at the final report.
We need to understand what happened along the way and what the student can now do.
This is also where the discussion becomes relevant to the university as a whole. Once AI can contribute to much of the work we ask students to produce, we need to reconsider what we teach, how students learn, what we assess and what a degree demonstrates. These are questions I am working on in my role as AI and education lead at TU/e.
A lot depends on how AI is used. You can ask a chatbot to explain a difficult concept, give you another example or ask questions until you can explain it yourself. You can ask it to challenge your argument and then work through the objections. You can also ask it to do the whole assignment. Both involve AI, but the learning can be very different.
For me, that distinction deserves much more attention. A student who produces a better answer with assistance may have learned something valuable. But we should also find out whether that understanding remains when the assistance is removed, or when the problem changes.
Recognising a convincing explanation is different from being able to develop and apply one yourself.
Learning takes effort. You need to try, get stuck sometimes, discuss with others and reconsider your reasoning. Writing can be part of that too. Putting an argument on paper often reveals that you have not thought it through sufficiently. You discover a missing step, an assumption you cannot defend or a conclusion that goes beyond the evidence. Working through those difficulties has educational value.
That does not mean every difficult or time-consuming task is useful. We should happily remove unnecessary obstacles. The challenge is to recognise which effort contributes to learning and which effort can be reduced. AI can help a student get past a confusing explanation and start practising. It can also provide a finished answer before the student has seriously considered the question. We need to design learning activities with that difference in mind.
We often say that students should critically evaluate AI output. Fully agree. But this requires knowledge. How do you recognise a wrong assumption or an important omission if you do not understand the basics yourself? Critical thinking develops through working with ideas, evidence and problems in a discipline.
In my own field, a route plan can look quite convincing. The sequence is clear, the map looks sensible and the explanation sounds professional. To judge whether it is useful, however, you need to understand the constraints: vehicle capacity, delivery time windows, working hours, congestion, and perhaps charging requirements. You also need to ask whether the problem has been formulated properly. A convincing solution to an incomplete problem can still be a poor decision.
The same applies to students evaluating an argument, a piece of code or a research design. We cannot expect them to judge everything from the outside without first developing some expertise on the inside. Foundational knowledge remains important, including when AI makes information easier to access.
The idea of the cloister and the starship is helpful here. Students need space for concentrated study, reflection and independent reasoning. They also need opportunities to explore what becomes possible with powerful tools. A programme should deliberately provide both. These activities can alternate throughout a degree: exploring with AI may reveal a gap that sends a student back to the foundations.
This has consequences for assessment. If we want to know whether a student has internalised a principle, we need a situation in which the student demonstrates that understanding independently. If we want to know whether they can solve a complex problem with modern tools, we should let them use those tools and examine the quality of their decisions. The conditions should follow from the competence we want to assess.
For a thesis, that means looking closely at the research process and the student’s understanding of the content. Intermediate presentations, discussions and an extended oral defence can make their reasoning visible. Why was this method chosen? What alternatives were considered? Which result is uncertain? What would change if an assumption no longer held?
The written work still matters as a way to communicate the research. But a polished text is insufficient evidence of individual competence. We can teach and assess academic writing in settings where that competence is visible, while giving the research process and demonstrated understanding more weight in thesis assessment. A student should be able to explain and defend the work they submit, including the parts developed with AI.
This is also why I am cautious about making AI detection central to the discussion. Even identifying AI assistance would leave us with the educational questions. What did the student contribute? What was learned? Were the relevant rules followed? Can the student take responsibility for the argument and the evidence? Those questions require engagement with the student and the work.
Individual lecturers cannot resolve all of this on their own. We need to look across entire programmes: where students develop foundations, where they practise using AI, and where they demonstrate independent competence. Students need clear expectations, and staff need time, training and support. A lecturer who redesigns an assessment also needs a workable way to supervise and evaluate it.
The same responsibility extends to research and university management. When we use AI to support an analysis, review a document or prepare a decision, we remain responsible for checking what it produces and understanding its limitations. Our own practices should reflect the standards we expect from students. It would be strange to ask them to be transparent and critical while leaving our own use unexamined.
Responsible use also has a societal dimension. Sustainability belongs in this discussion, including the energy, infrastructure and resources AI requires. We should teach students to consider whether a task benefits sufficiently from AI, which tool is appropriate and whether a simpler approach would do. Generating more material is not automatically useful. That judgement belongs alongside questions about accuracy and quality.
There is an institutional responsibility here as well. Universities make choices about which systems they provide and which forms of use they encourage. Environmental impact, privacy, access and dependence on suppliers should be part of those choices. We should not place the entire burden of responsible use on an individual student while leaving the university’s own decisions outside the discussion.
Changing all of this takes time. Staff need room to experiment, and we need to evaluate whether new approaches actually improve learning. Useful questions include whether students retain their understanding, can apply it in unfamiliar situations and can recognise when assistance is misleading. Better-looking assignments and faster completion are only part of that picture.
We will not have every answer before we start. I think we should be prepared to test approaches, discuss what works and revise what does not. That requires coordination across the university, as well as space for differences between disciplines. A single rule for every course is unlikely to capture what we are trying to achieve.
Adding an AI policy will therefore take us only so far. The work reaches into our curricula, teaching, assessment, research, staff development and the way we organise ourselves. Quite some homework for us too..
For me, re-thinking thinking means taking that work seriously. I want our students to benefit from AI and to develop the knowledge and independence to question it. They should be able to explain their choices, reflect on the consequences and take responsibility for the outcome. Building a university that supports this is a responsibility we share with them.
