The obvious way to square this circle is to go back to how things used to be: less emphasis on coursework and more on old-fashioned tests in an examination environment with just pen and paper.
You can cheat on your homework all you like, but you'll completely fail the exams. On the other hand, students who use LLMs to augment their learning will do fine
It’s one dimension, I agree. Though, for essay-based courses I’m a little skeptical about writing under time pressure.
My thinking here is that you might want to also introduce more “viva voce” (verbal defense) style tests. These have been expensive to administer but my hunch is that you can scale these with AI administering the first round, random human review or participation. This feels like a better endpoint than essays under time pressure (but also maybe you can elect one or the other?)
My other angle is to break completely from the past. I want to see more project work. If AI makes it easier to fake or attain any given level of knowledge, we need to ask more of students, ie ask them to demonstrate mastery by building real things. More open-ended projects are harder to grade (use AI to scale grading), but more fun and engaging for students.
This doesn’t just refer to STEM of course; why not ask students to write a play, compose a symphony, etc?
My core thesis is that most students would work harder and engage with things that interest them, and the fundamental problem with Taylorized education is that one-size-fits-all is only interesting to some.
I studied in a French Lycée in Spain. We did dissertations, brutal 3h live writing tests. There was no problem with it other than the difficulty. Kids didn’t come out screwed up or impaired.
Exactly, “write an essay under time pressure” used to just be called… an exam.
It’s normal and was done for centuries (and still done in many countries) for a reason, it works and it’s equally hard for everyone so it’s not unfair.
(3 hours straight is brutal though)
I agree it’s an option. I did some essay exams at Uni too. I just don’t think we should assume that the old cost/benefit calculus still applies and this is the best way to test e.g. English Lit or Philosophy.
> You can cheat on your homework all you like, but you'll completely fail the exams
This might sound principled, but we need to recognize that school administrators are incentivised to have as few kids as possible fail their exams; and consequently, so are the teachers. Either exams will change, or the teaching will change.
People keep bringing this up as though it’s a counter argument.
Yes, things will have to change. That’s the idea.
This is definitely what's going to happen. But there is still a problem - people (not just children) are fundamentally lazy and put things off. People always leave assignments until the last minute. I used to smoothly transition from "I've got plenty of time, no need to start" to "there's not enough time to do it, there's no point starting".
They will 100% just use AI for the whole year and then panic and fail the exam when the time comes.
So I think not only will we see more invigilated exams, but they'll become more frequent and shorter. Which I would say is a good thing anyway. I always hated learning a whole year of stuff for a 3 hour exam.
In what countries were the exams only once a year? When I grew up in the 90s in Sweden we would have tests and exams frequently, usually at the end of each module. This continued all the way through university. I think we had 3 separate exams for the first math course (which lasted a quarter of a year, so roughly one exam per month).
(Though they didn't give formal grades for the first several years of elementary school, which I'm not sure was a good idea.)
In the UK. In high school (age ~11 to 16) the only exams that mattered were SATs at ages 7, 11 and 14 (though I just checked and apparently they've scrapped the 14 year old ones for some reason), and GCSEs at 16. After that you have A-levels for two years (age 17 and 18) where IIRC it was just one big group of exams at the end of each year, and then university where I guess maybe it varies but at least at Cambridge it was one big group of 3-hour long exams at the end of each year.
Though Cambridge does have "tutorials" which are 1:2 tutoring sessions where you probably couldn't completely rely on AI.
Your way definitely sounds better to me.
In Poland, that's how it worked at unis. We had a lot of subjects where we had only one big exam at the end and your grade depended on it.
Before uni though, we would get a grade or two every week, be it from short tests, classwork, homework or exams.
It’s a good point. I could see a world where exams get cheaper to administer.
I also think we can use project work as an ongoing assessment.