Since generative AI tools became widely available, almost every university has published some kind of guidance for students and teachers. Public discussion tends to sort these documents on a single line. Some universities are called strict, others permissive, and the debate is about where on that line a university should sit.
When I started reading the documents themselves, that line stopped making sense.
One document, two messages
A single guidance page can warn students that undisclosed AI use counts as academic misconduct, and a few paragraphs later offer workshops on how to use AI critically in their studies. Is that university strict or permissive? It is both. Putting it at the middle of a scale would hide exactly what makes it interesting.
Measuring restriction and support separately
That is why I built the Nordic University AI Policy Tracker around two independent scores.
- Restriction and Enforcement Index: prohibitions, penalties, misconduct framing, AI detection, mandatory disclosure and control over assessment.
- Pedagogical Integration and Support Index: permitted use, AI literacy, critical evaluation, teaching integration, support for students and staff, citation guidance, and attention to equity and accessibility.
The two scores are never combined. A policy can be high on both, low on both, or high on one and low on the other, and each of those combinations tells students something different about what their university expects.
Why it matters for learning
A rule tells students what they may not do. It does not teach them how to use AI well. Students who arrive with strong digital skills can work that out on their own. Students who do not depend much more on what their institution offers. If the pattern I found in European survey data, where confidence with digital tools shapes attitudes toward AI, also holds for students, then the support side of a policy is where inequality gets decided.
Where the project stands
The tracker is a pilot with one university in each Nordic country: Sweden, Norway, Denmark, Finland and Iceland. The pipeline collects the documents, processes them in Python and codes them in a way that a human can check. The pilot tests the whole process end to end before I scale it up to more than 30 European universities. I will write about the comparisons once the coding is validated.