The only real solution to this (assuming that grading systems are all that valuable in terms of optimal educational outcomes) is to base grades on in-class essay writing (and for CS, in-class coding) where students don't have access to any outside assistance, and have to think quickly.
Using tools outside class to practice for the in-class work would make a lot of sense however, as students could then rapidly see what a decent essay or code body looks like.
Of course. I don't understand how essays written at home are acceptable. It's a good exercise to do them (if done properly) but they shouldn't count for the final grade.
Before AI there were professional essay-writers, or tutors, or parents. In a sense AI is progress because now everyone can cheat, instead of only the wealthy.
This is common in UK universities at least. You can have your entire grade for a year of dense reading at the MSc level be based on three hours of hand-numbing speedwriting. It makes sense while at the same time rewarding and punishing different personality types (and calligraphers).
The worst thing about this system is that there's not usually any chance to retake the exam. As far as I can see, there's no good reason for this, and all it does is heap pressure on students while providing a worse overall assessment of their ability.
Don't most UK universities let undergraduate students resit exams they've failed, within reason?
And a degree consists of a good number of courses, each of which will have its own exams - so if you do poorly in one or two exams, there are plenty of chances to bring your average back up.
> Don't most UK universities let undergraduate students resit exams they've failed, within reason?
I've attended 3. Two of them made students retake the entire year if they needed to resit an exam. The other one based students' entire grade on one set of final exams at the end of the 3 years and had no provision for resits at all.
I believe the first two are typical in the UK.
> And a degree consists of a good number of courses, each of which will have its own exams - so if you do poorly in one or two exams, there are plenty of chances to bring your average back up
That's true, but it's also true that a more lenient system would be more compassionate (student mental health is a huge topic for universities) whilst simultaneously being a better measure of student's true abilities.
> there's not usually any chance to retake the exam
Of course there is. One of my classmates milked this system to remain a student for about twice as long as the rest of us. He was good at manipulating the bureaucracy, not so good at physics; probably should have studied law or some sort of social science instead.
If you look at it another way, it's like a lifetime warranty. You can't trust an appliance vendor to provide quality, but you can trust that if they have a warranty they will meet the conditions where it will make financial sense to have one.
Having such an exam adds enormous extraneous pressure that does not reflect the reality of conducting research, but it is this pressure that guarantees (with some exceptions) that the student has been coerced into learning the material to a sufficient standard. There will always be the cramming psychopath who grabs a first after just a week of study and who retains nothing of it, but most students just have to learn intensely over the course of a year to pass through the filter. In some ways, this is also why it's common in the UK for employers to not care so much about the specific degree, just that you have one with good grades from a reputable institution.
> There will always be the cramming psychopath who grabs a first after just a week of study and who retains nothing of it, but most students just have to learn intensely over the course of a year to pass through the filter.
My concern is for the students who study intensely, learn as much as anyone else, but don't pass the filter. Universities in other European countries allow single units to be retaken without retaking an entire year, and it seems to me that this provides a fairer system without devaluing the qualification in any way.
I agree in principle, but there are so many special accommodations required (extra time for graded assignments, special assistance in the form of a reader, etc.) that would make this a nightmare.
I lament, though, that students are often either (A) so overburdened with distractions that they don't have time to learn, (B) so pressured to get high grades that they cannot afford to get a bad grade, and thus must cheat, or (C) just don't care. There are a few gems that are willing to put in the work to learn and appreciate a subject, though. I just wish that more students put in the work required to learn.
I have legitimately met college students in the last few years who had no frame of reference for how much time they should be putting in. Freshmen who think that spending 15 minutes a week looking over their notes is "studying a lot" are pretty common right now (in the USA).
Three hours of study for each hour spent in class is an average estimate of what it takes to absorb even moderately difficult material in courses in sci/tech.
Relevant quote: "Every truth has four corners. As a teacher I give you one corner, and it is for you to find the other three." (Confucius)
Which with napkin math can be shown to be bullshit. Students are spending approximately 1/3 of their time in classes, so how would they get time to sleep among the 4/3 required by studies and classes?
1:3 hours sounds super aspirational even for only the hardest of classes like organic chemistry and fluid dynamics.
But your napkin calc isn't addressing the bit where you only take 1-2 classes like that per semester and pad the rest out with things like literature and electives so you could have a life.
> students could then rapidly see what a decent essay or code body looks like.
You mean like how today they study well-regarded human works? I’m not sure how the ML adds anything except which words tend to appear close to each other across a large corpus. The moment where young writers believe that ML is creating anything new and they try to learn from that is when modern civilization starts to eat its own tail.
Given that a typical programmer on a typical day has online access, and given the creative lengths students will go to to maintain access to said resources during exams, it seems worth considering allowing online access for programming exams. To address the obvious problems with that approach, could we use AI to generate unique variations on the problems we want the students to solve?
Or we could stop asking students trivia and force them to think. Most of my graduate school exams were open book + notes but it hardly helped. You had to demonstrate knowledge by producing a novel solution to the problem most of the time. Not novel as in new research, but novel in the sense you weren't just regurgitating factoids from the books. I also had tons of open book and open note advanced math classes. The problems were made in such a way you had to make a connection with the material and link pieces together in non-obvious ways. In those classes I retained more information even to today.
Teaching, even at the collegiate level, has become "how can I do the least so I can do what I want". Tenured research professors generally make terrible teachers...perhaps we need professional teachers.
For me, the most "open" a test was, the more I worried about it. Expecting students to remember everything on the test meant that it had to be easy.
The hardest exam I ever took was my final exam for Linear Algebra. It was so "open" that people could use matlab on a shared computer that was also projected to the entire class. Turns out, the best students in the class didn't need help on a computer, so it turned into this kind of mind-fuck where you weren't sure if they got a different answer than you because they were wrong, or because you were. To add to the excitement, the questions all tied together and used the answer from the previous one, so if you missed one, you were guaranteed to miss every subsequent question.
This is like any open-book exam. Make the time limit reasonable such that if they know the stuff, they can get it done while referring to a few things in a source, but if they don't know it, there's no way they can look it all up in the available time.
You could have them write using a program that allows the teacher to play back their writing process in real time as well.
Theoretically they could copy it from offscreen but unless they’re really good at faking going back and changing things and getting stuck at the right points it would probably be fairly easy to tell.
Using tools outside class to practice for the in-class work would make a lot of sense however, as students could then rapidly see what a decent essay or code body looks like.