A professor at École Polytechnique de Louvain rethinks exam rules in the age of LLMs, allowing students to use chatbots under strict conditions. Students must disclose usage and justify outputs, with violations treated as cheating.

Most students opted out of chatbot use, citing preference or distrust. Those who avoided them scored higher on average. Only three students used AI, with mixed results. One forgot to use it, one used it minimally, and one became over-reliant, hurting performance.

The experiment reveals a generational skepticism toward AI tools in academic settings. Many students still value personal understanding over automated assistance. Educators may need new models for assessing AI-literate learning.