Morning all. I have some thoughts today on how AI is likely to affect teaching and assessment in higher education.
A few weeks ago, I was at an event hosted by RBC, which was kind of a fireside chat with Anthropic’s Head of North American Government Affairs, Brian Peters. This was shortly after Anthropic released this think piece on AI and economic growth that asked people to contemplate the prospect of 15% annual GDP growth thanks to the miracle of AI. Now, I happen to think this kind of stuff is utter horseshit for a variety of reasons, the main one being that the economy is still substantially made up of atoms and even magical AI can’t speed up the process of re-arrangement of atoms quickly enough to reach that level of growth so quickly.
But hey, these guys were nice enough to invite me to a fireside chat, so I played along. When it came time to ask a question, I said “look, we clearly aren’t in a world right now where 15% growth is going to happen; there are any number of studies that show basically no firm-level or economy-level productivity growth from AI, presumably because there are very few companies that have yet to actually re-build their business processes around AI. So, what is it that is going to change this situation? What’s the “click” moment when we move out of a world of 1% growth to one of 15%?”
His answer, roughly, was: “when AI companies manage to make the re-engineering of business processes plug-and-play. When changing business process is so easy, so frictionless that it’s push-of-a-button”.
And that’s when a lightbulb went on for me. In the techno-optimist version of AI, almost no one will need AI skills because AI will have made everything so easy that it will be a push-of-a-button. We won’t need to teach specific to AI technology any more than we need to teach people how to use a refrigerator or a laptop.
Will there be a need to train more people to design/make AI systems? Of course. We will need specific programs to grow the number of graduates who can do this. But that means going from – say – three percent of students who can do this up to five percent, and it means finding ways to teach the material better. But that, for the most part, is something more or less internal to engineering and computer science. Everybody else? They can probably go about their business as if AI isn’t happening.
(Now that’s the scenario for AI that the AI hucksters are pushing. Paradoxically, the case for more teaching about the use of AI grows as truth and AI hype diverge. If – and I suspect this is true – AI is not some kind of miracle technology but rather a merely useful and sometimes flawed one, then the need to provide training actually increases because people need to know how to use and coexist with AI. What that looks like will almost certainly differ quite a bit by field of study, but for the moment it largely means critical thinking, media literacy, and not taking any answer you get from AI at face value. Which is something almost everyone claims to do, but few bother to demonstrate or prove – meaning we could all improve in this area.)
So much for AI and teaching. What about AI and assessment? I have written before about the economics of assessment in the age of AI, but the short version is that AI is wreaking havoc on many traditional forms of assessment that, in many disciplines, are not only traditional measures of gauging thinking and measuring learning but are also very economically viable (the ratio of student writing time to professorial assessment time is effective). The simple-stupid solution – just move to “authentic” methods of assessment like much greater oral examinations of one form another – simply isn’t scalable under our current system. Something big is going to have to give, soon.
Why soon? Because I believe very firmly that AI is putting universities and colleges in something very close to mortal danger. If universities can’t reliably confirm that their degrees are certifying some degree of actual learning (as opposed to ability to query ChatGPT), why would employers continue to trust degrees? And, if degrees aren’t trusted, why would governments and students continue to pay for universities?
(There’s an equity argument here too: if students don’t think they are being assessed fairly according to their talents, why would they bother coming to universities? Look at the wave of youth movements around the world: they are all protesting a lack of social mobility and some variety of “corruption” which closes off opportunities to many. Universities are only useful and popular if they stand against those things and remain beacons of impartiality. This is why some of the loudest voices on campus for clarity on AI policies are students: they want to make sure everyone is playing by the rules.)
I want to be clear about this. If we don’t solve the assessment problem in a way that students and employers see as valid, then that’s literally the whole ball game. Universities that do not provide trusted credentials of value are simply not fit for purpose. They will disappear. And it won’t take long, either.
So, institutions are stuck between a rock and a hard place. The amount of time devoted to student assessment is going to have to go up, but that’s expensive. Paying for it in an age of scarce funds means giving up other things, like time for research, which no one is going to be keen to do (though I suspect some reduction in research is inevitable). Instead, what people will prefer to do is search for ways to make “authentic assessment” less time consuming. There are potential solutions along this line, but they require a pretty big rethink of how universities are organized.
The most interesting suggestion I heard recently – at a dinner HESA hosted in Kitchener-Waterloo a couple of weeks ago – was to get out of the business of doing summative assessment at the level of the individual course and leave it all until the end of a program. This would relieve folks of the need to do authentic assessment multiple times a year, and just do it in a high-stakes end-of-program fashion. This is actually the way a lot of university programs in Europe worked prior to the Bologna process (Italy especially), so it’s not entirely without precedent, but the effort required to de-center courses and re-center programs would be massive.
That said, I think smart universities need to be thinking hard about these kinds of options. As I said, if we fail as a sector to prove our ability to assess students, we’re done. There is no issue in higher education where the stakes are higher and we will all need to be creative to cope with this challenge.
These are precisely the kinds of questions and challenges we will explore together at AI-cademy 2026 at UBC in three weeks. If you are joining us, we look forward to hearing your perspectives, learning about what is happening at your institution, and engaging in collective strategic thinking with postsecondary leaders from across the country.
If you have not yet registered, only a few tickets remain. The program begins on November 8th with a special Teaching & Learning Showcase, with sessions and conversations continuing through November 10th. We hope to see you all there.
