First student fully supervised by AI
by Bas Haring, Leiden University, the Netherlands. At the end of the summer, just before the start of the new academic year, Alicia Cai graduated – a student of Chinese origin who studied Media…
by Bas Haring, Leiden University, the Netherlands. At the end of the summer, just before the start of the new academic year, Alicia Cai graduated – a student of Chinese origin who studied Media…
by Bas Haring, Leiden University, the Netherlands.
At the end of the summer, just before the start of the new academic year, Alicia Cai graduated – a student of Chinese origin who studied Media Technology at Leiden University. Nothing remarkable in itself, were it not for the fact that Alicia is the first student ever whose graduation project was fully guided by AI. Except for the first two weeks, when I was her supervisor. In those weeks, I noticed that Alicia made an exceptional amount of progress, unlike anything I’d ever seen before. I asked her how that was possible, and she told me that she was constantly chatting with ChatGPT – and similar tools – about her research: receiving suggestions for research questions, critical comments about possible methods, and so on. All day long. I started to doubt my added value as a supervisor, and I made a list of tasks that a graduation supervisor typically has. It didn’t seem impossible to outsource a good part of that list to AI, and I then suggested to Alicia that she hand over the entire guidance to AI.
The goal of her research project was twofold: on the one hand, she would work on her original research theme, but on the other hand, she would also study what it was like to be supervised by AI. Alicia agreed, and she chose to be guided by a duo of ChatGPT and Claude.
Since then, I have only seen her again at her graduation presentation. (which I had to assess, along with her thesis, as I formally remained the supervisor.) In the meantime, I could follow her progress only through a shared document. We had no direct contact, but we agreed that we could both always stop the experiment at any given point.
Alicia has successfully graduated, and the guidance by AI has worked in that sense. But there are a number of remarks to be made. I would like to summarise the lessons that Alicia has learned as follows:
Overall, these lessons are quite positive, although the latter disadvantage weighs particularly heavily. The experiment has had a considerable impact (in the Netherlands): newspapers and other media reported on it, usually positively, but sometimes negatively.
I would like to make one point for clarification. This experiment is not intended to suggest that AI can take over the supervision of all students for an entire graduation project. The intention is to show that some of our academic work will probably change in nature; with this experiment, we were able to discover fairly quickly where suspected possibilities and impossibilities of AI can be found.

Bas Haring, Leiden University, the Netherlands