Almost every week, I come across a new initiative on artificial intelligence in research and higher education. A declaration, a set of guidelines, a working group, a conference, or recommendations on teaching, research and assessment. International organisations, national agencies, universities and departments are all trying to respond to the rapid development of AI.
Many of these efforts are valuable, but I must admit that I sometimes find the sheer volume overwhelming. Are we collectively making things clearer, or are we adding to the confusion?

As a dean, I am very much part of this process. I participate in discussions, support new initiatives and contribute to decisions about how our university should respond to AI. So the question is also directed at myself: are we, as university leaders, sometimes adding to the uncertainty rather than reducing it?

There is broad agreement on many of the challenges: AI is transforming education and research, students need AI literacy, assessment methods must evolve, and academic integrity remains essential. But translating these principles into everyday academic practice is considerably more difficult than agreeing on them.
Different courses and research environments will reasonably adopt different approaches to AI. A course teaching the foundations of AI may even require students to solve problems without AI assistance. There is nothing contradictory about that, provided the expectations and their rationale are clear. What matters is not uniformity, but a coherent framework within which students and teachers can navigate these differences.
Such a framework should not have to be developed independently by every university. Students and researchers move between institutions, and many of the questions we face are shared. Yet universities and national agencies continue to develop their own guidance in parallel. Rather than multiplying frameworks, shouldn’t we work towards a common one, developed collectively, while leaving room for disciplinary and local differences?

There is another aspect that deserves attention: stability.
AI technology is developing at extraordinary speed. But must our educational rules and academic expectations change at the same pace? I would argue that they should not. Students need clear expectations throughout their education. Doctoral students and junior faculty need guidance and stability as they develop their academic independence. Teachers need time to adapt their methods and evaluate what works. A continuous flow of new recommendations risks creating uncertainty precisely where we should be providing direction.
This does not mean resisting change. We should distinguish between principles that remain stable and practices that evolve as we learn more. We should continue experimenting and learning, while giving ourselves time to implement existing decisions and evaluate what actually works.

For our part at SCI, we will prioritize building on existing initiatives, developing our guidance in dialogue with others, and giving students and colleagues the time and stability they need to put it into practice. Follow us at https://intra.kth.se/en/sci/ai-for-utbildning.

And yes, I realise the irony of adding yet another text about AI to the already overwhelming collection. Perhaps I should take my own advice.