Vice Dean Jeroen Touwen on AI and education: ‘Thinking process shouldn’t be outsourced’

Published on 27 August 2026

One of the hot topics in higher education is the emergence of Generative AI (GenAI). Vice-Dean Jeroen Touwen shares his thoughts on the future impact of large language models on education and assessment.

How does the Faculty of Humanities view GenAI?

In our ‘language-focused’ faculty, a lot of attention is paid to written assignments. GenAI and Large Language Models (LLMs) therefore cause more consternation here than in other faculties. This past spring, the faculty members and researchers from the AI Expert Group, under the guidance of Maarten Bergwerff and Ferdinand Harmsen, gave this issue a great deal of thought and engaged in extensive discussions. They arrived at two key questions: How will we assess in the future what we have traditionally sought to assess through a term paper or essay? And: Will our students still learn to think critically and independently and to express themselves clearly?

The latter sounds as if the humanities core of education is under pressure.

I sometimes hear colleagues in the Faculty of Humanities say: given the advent of language models, we’ll have to revise our learning objectives. After all, the computer can write and formulate ideas for you very clearly. I think that’s the wrong approach. First: we still need to teach students to ask the right questions. Second: asking the right questions requires knowledge. Third: we have to arm ourselves against the disinformation, errors, and hallucinations produced by GenAI. It is absolutely essential that future academics contribute to the information society based on knowledge, insight, and, above all, critical thinking skills.

Can you elaborate on that?

The thinking process should not be outsourced to GenAI. After all, we want students to acquire interdisciplinary and conceptual skills from their field of specialisation, such as critical thinking and independent analysis. To achieve that, students need to grapple with issues at an abstract, conceptual level.

What’s different now compared to the past is that students no longer learn these skills by writing their own papers. That’s why I advocate for a re-evaluation of how we teach and assess to achieve our learning outcomes. This means that some things need to happen more in the classroom, while others can be done more often at home.

What does that look like in practice?

We need to make sure that the risks of AI are addressed in lectures. That’s one of the things the new core curriculum course does. I can also imagine a lecturer asking students to write down a problem statement with sub-questions or a hypothesis using pen and paper, without digital tools. The results can be discussed during class through peer review, in which students comment on each other’s notes. By assigning a grade or a pass/fail to this, you make it clear that independent thinking is a serious part of the course. Some lecturers decide that laptops should be kept in their bags for the entire lecture series, and students respond positively to this.

At the same time, you can also use AI to enhance the learning process. A lecturer can set up a chatbot to challenge students to ask the right questions based on the lecture material. Such a bot provides the right feedback and probes deeply—a bit like your older brother, your mother, or your roommate used to do. Try the Quizbot Sophie or the History Research Guide (see more here). Leiden is developing the LUCA platform specifically for this purpose (in a secure environment).

How can we ensure that students make smart use of GenAI outside of class as well?

You can ask students to submit a ‘log’ or ‘conversation history’ as an attachment to their essay that includes their discussion with the chatbot. Providing feedback on this helps them reflect on the value of AI. If students submit multiple versions of a paper, you can see how their thought process evolved by looking at the history of the document.

So, will assessment need to change in the future?

We’re actively working on best practices for AI-proof assessments: the solutions are shared on the ECOLe website. In addition, we’ve revised the assessment guidelines. Not everything needs to change: an exam in ANS is still rock-solid.

Can’t we just ban the use of AI?

Banning it doesn’t properly prepare students for their professional lives. Whether they pursue a career outside the university or in academic research, they need to learn how to work with AI.

Don’t you find it problematic that large data centres consume so much energy and water?

That is a concerning aspect of the AI revolution. The new faculty guidelines will state that we aim to provide support and alternatives for students with conscientious objections. And it will not be mandatory for instructors to use AI in their teaching.