The format

A short opening, five lightning talks of seven minutes each from colleagues in different disciplines, then refreshments with a panel discussion and questions from the floor, 35 minutes of conversation in small groups, and a summary. The speakers were not told what to talk about; each was asked only to speak to the perspective that matters most from where they sit.

The speakers

• Berglind Einarsdóttir, lawyer and founder of the consultancy Bentt, who chaired the Law Faculty’s conference on law and artificial intelligence in Harpa last year.
• Haukur Ingi Jónasson, Department of Engineering, co-director of the MPM program in project management.
• Leó Sakamoto Kolbeinsson, third-year student in computer science, and so the one speaker who belongs to the group the rest of us were talking about.
• Luca Aceto, Department of Computer Science, member of Academia Europaea and for many years president of the European Association for Theoretical Computer Science.
• Magnús Smári Smárason, who leads the adoption of artificial intelligence at the University of Akureyri.

Three tasks

The opening put three questions to the room. They are the frame the rest of this work hangs on, so here they are, lightly abridged from the speech.

Protect real learning. AI makes it easy to obtain an answer without doing the thinking. What worries me most is not cheating. It is that when the effort disappears from learning, so does the thing that makes the mind grow. A muscle does not strengthen unless it meets resistance, and the mind is not so different. The danger is not that students learn the wrong things but that they never develop the depth that education is supposed to produce. And it is insidious: we see impeccable-looking solutions while the growth that was meant to accompany them has never taken place.

Enhance learning through AI. Used well, it can act as a guide, a coach, a sparring partner in thought, and it can give feedback instantly. The largest opportunity is that it can adapt to each individual and meet very different learners where they are. But this does not happen by itself. A chatbot is not built to offer guided resistance; it is built to remove it. That is what makes it useful everywhere else and what makes it hard to teach with. The tools exist; the pedagogy, the implementation and the experience have a long way to go.

Rethink what we teach, and how. If our graduates will work alongside ever more capable AI systems, then what should distinguish a university-educated person in the age of artificial intelligence? What knowledge must a person still hold in their own head? Which abilities become more valuable than before? And what, in truth, is the purpose of a university in a changed world?

None of us has final answers. As academics we are used to relying on solid evidence and avoiding premature conclusions, and we should not abandon that. But at this pace of change we cannot wait for certainty either. The task is to combine scholarly rigor with a willingness to experiment: try things, learn from them, and correct course together.

That was the opening. The speakers took it in five quite different directions, which was the point.

What the groups were asked

The small-group discussions worked from two questions, and each group was asked to write one sentence in answer to each:

1. Which of the things we ask students to do are difficult on purpose, and which are just difficult? Name one you would change before term starts.
2. What should our graduates still be able to do unaided, ten years into their careers, and why that, specifically?

The first question is, in effect, the distinction between desirable difficulties and merely difficult ones that learning research has drawn for decades. It is the right place for a university to start.

What comes next

The follow-up is a series of orientation sessions for departments and instructors, which will be the subject of a later dispatch here.