alt Hacker NewsAI use leads UC Berkeley math professor to see 'no point' in grading homework
Comments
As an aside, the professor in question is Richard Borcherds, fields medalist. He is probably most well known for proving the “Monstrous Moonshine” conjecture of Conway and Norton that there is some sort of correspondence between modular functions and representations of the Monster Group. When the conjecture was made in the 1970s, it was purely because someone noticed that a certain number was the same as the cardinality of the monster group. It seemed a crazy coincidence, so John H Conway said (playfully) that it was “Moonshine”. Hence the name “Monstrous Moonshine”.
https://en.wikipedia.org/wiki/Monstrous_moonshine
He has a youtube channel with lectures in number theory, category theory and other things.
This is quite sad. Even if students would cheat off each other, someone had to do the homework. And even the collaboration of cheating on homework always served as a forcing function for me or others to think about the material. If not actively, at least in the form of a quasi background radiation. With AI that is completely gone, and without reward loops for incremental work, the truth is there is a large number of people (especially 18-20 year olds) who will not do well on the exams because they need a bit of a stick.
I am not saying this as a black and white statement. I just think there are different cohorts of students based on maturity and self motivation and we might be losing out on one cohort that is at neither extreme.
With some exceptions I think higher education should probably go back to being about academia and not necessarily teaching skills for industry. I think companies can now easily invest in their own extended training and grab people out of high school. Just a thought.
> Borcherds said he currently grades on completion if a student hands in their homework.
In high school (2006) my physics teacher gave credit for handing in the homework (he checked it and handed it back, but no grade was assigned). Didn't hand in homework, you got a zero.
He knew then that if you assign grades to hw, people will cheat.
If I'm paying for Berkeley, the professor have office hours for me to ask questions.
What is the point of grading homework for college students anyway? The objective of exams is to determine mastery of the subject. It makes sense to have a grade act as a certificate of mastery.
Grading homework is a measure of conscientiousness and perhaps a pedagogical tool to force learning but by college perhaps we should have moved out of the need for the latter.
One can simply measure conscientiousness through attendance grade and mastery through exam grade and then we have two grades the universities can certify students for that are no longer dimension reduced into a single number.
In-class exams, either written or oral. If you don't do homework, it is only on you. And with paper and pen, the next semester, yes.
When I studied, in Norway, we never got any marks or points for our homework. Our final grade was based 100% on our final exam (or mix of final exam + projects for the classes which had those), which was always a pen and paper type of exam. We did have mandatory homework we had to complete, in order take the final exam, but we never got any grades on those.
Later on, when I became a TA, there were obvious signs of cheating in the homework, i.e. people handing in the same homework with the same step-by-step solution, with the exact same errors.
In any case, cheating on homework wouldn't help their grade.
At the lower undergrad level, there was already no point in grading homework a decade ago because Wolfram Alpha existed.
At the graduate level maybe this problem is solved by doing Moore-style classes where the students present (on the chalkboard, not slides) and the professor critiques. Students could still have AI generate proofs and print them out or hand copy them, but maybe that's fine.
Upper undergrad level is tricky, because students are expected to prove things, but maybe don't have the maturity for full-on Moore-style.
We are still going to be interested in teaching people to think. We are still going to be intrinsically interested in mathematics. Many (most?) of us here have gotten immense pleasure and growth out of solving a problem, or understanding something well enough to explain. Even though the answer was known or someone else could explain it better. AI doesn't change that, really. So we have to find new ways to assess students. It doesn't mean anything fundamental about education. I don't even think it's such a fundamental challenge to math. Most of these things we are trying to prove are, to an engineer, either irrelevant or known empirically already. People want and need to try to understand. They will continue to.
Reminds me of a story from Freakonomics.
A teacher would ask his students to note 50 coinflips. And based on distribution he could immediately tell which one of them actually flipped the coin.
If I would be a math/physics teacher in a high school, in a class of 30 students, I would start every class by picking random 2 or 3 students and ask them to explain the solution to me. Or do a carbon copy of one of the assignments in front of the class. Something like that.
In my high school I had a physics teacher. Every class she had a 3 rudimentary questions about the recent subjects. She would call a random name and ask first question. If you fail to answer on the spot you get an F, and another random person will get a shot at the same question. If they fail they got and F as well. Of course by the 6th or 7th person someone would manage to find the right answer. If they got the right answer they'll get the next one. If someone answered all three - they got an A. Most of the times half of the class would get an F.
Borcherds has been working with Epoch AI since at least 2024:
https://epochai.substack.com/p/ai-math-chat-1-thinking-smart...
Now he issues the standard taking points about looms and horses and claims he could not have foreseen it? He is an AI shill, just like Tao.
I have been an adjunct CS prof on and off for decades. I stopped in 2020 because teaching over Zoom sucked.
If were to teach a course today: I’d assign homework as usual, and have it be worth zero toward a student’s grade. The point is to learn, do it or not, that’s your choice. I would base the grade on in-person exams and maybe an interview. That’s it. No room for AI to boost grades.
Homework is useful as a form of revision, if idiots think they can cheat to a better score the university just needs to take away that incentive: 0 marks for homework, all marks on the exam and a viva.
He should perhaps ask students if they want it marked and if they do still provide feedback, but be clear it won’t count for grades as cheating is too easy.
Lazy students will do no work all year then fail. Clever motivated students will continue to do well.
More people will fail till they realise they need to do the work, but that’s just the way it is, it’s the only way to be sure (along with vivas) that students know the subject.
Many professors there didn't use homework as part of your grades when I attended. I think the reasoning at the time was students would just copy each other's work anyway or look up solutions online etc
But if you didn't do the homework / learn the material you'd be screwed on the mid-terms / final which was 100% of your grade.
On the flip side, AI's can now grade pure math problems with the accuracy and rigor of a postdoc. And unlike PhD tutors, they're universally accessible to all.
It's the start of a golden age for those of us who are curious autodidacts. If you're working through the exercises of a math textbook, you can now instantly submit your (attempted) theorem-proof to the nearest robot, and see all your misconceptions and shortfalls of rigor laid bare.
I never graded homework or anything else except the final exam.
In the lab I assist students with implementing and understanding code. Since they know they are not graded, there is no incentive for them to use AI. Sometimes we do try AI for the lulz but it makes as much sense as looking at the solutions (which I do provide already). As I frequently tell them, "AI can replace me teaching you, it cannot replace you understanding the material".
I've been enrolled in an online CS program provided by a respectable university for a while. I get a lot of joy from continuous learning and being in a university program helps me stay focused, plus I'll get credentials to prove it eventually.
Having said that, higher education as we know it is facing extinction in less than a decade. The materials I'm provided are a rambling stream of consciousness that are difficult to make sense of. Only truly elite universities that educate students efficiently deserve to survive.
An odd thought keeps running through my mind.
We don't yet know if AI has replaced mathematicians.
But if AI can do homework and exam problems, then it means AI has replaced students.
Just grade it like law school (in the US). Homework means nothing, and only the final exam counts. The final exam is administered in a room where you have no access to AI.
When I was in College (as a Chem major) homework was not graded. It was there for you to do to prepare for exams. I'm assuming that this is different for pure math majors though.
Then don't grade the homework! I have in-person quizzes in my class, and the students are motivated to do the homework anyways because it prepares them for the quiz.
Yes, there is no point in grading homework or essays. This has been largely true in India and China anyway, and the US will have to move to similar model of standardized tests and in person exams. Universities are anyway proxies for IQ filtering. It is perhaps impolite to say so explicitly in the west, but people will have to confront that.
Why would you grade math homework? I taught "intro math"[1] through algebra and calculus to linear algebra at UF and ASU over four years as a paid graduate TA. This was the middle '80s. I assigned homework from the text every class. At the beginning of every class we (the students and me) worked through any particular problem any student wanted. That was 15min, max. Then I lectured and assigned more homework. I graded my own 3 in-class closed-book exams and final exam.
Based on the classes the Math Department gave me (very satisfying) and the student evaluations this seemed to work more or less fine. I was a brutal grader though, some of those entry level classes had bimodal distributions with 'F' being the low mode.
I also showed up for office hours and did my best to help people learn by digging through their questions.
I don't see why this methodology wouldn't work fine today. I wouldn't care now how a student learned. Show competency on the human only test and we're good.
I had one student who achieved a solid 'F' come by office hours after the course to tell me how much he enjoyed my teaching. Whatever!
Anyway, teaching that way, with the positive interactions with students interested in learning, was one of the greatest experiences of my life.
[1] An absolute pedagogical shitshow that wasted everyone's time for zero gain.
If they don't mark the homework, it might be useful to redistribute that time to a viva about the home work.
It would give a more accurate impression of the kids understanding, incentive understanding it, be more pleasant interaction for all involved and build the kid up to Vivas and interviews.
Would have been great for me because at the age of fifteen I lost complete interest in school grades. Just interested in learning.
I see this as nothing else but an opportunity. Just start asking students about their homework. Not being able to answer about the homework should be punished with seize of the studying...
Can't we install AI on the student's phone, then let them work at home while the AI watches them, and then let the phone grade the student?
"So you thought AI was a gift from heaven? Ha, it's your worst nightmare! Let this be a lesson."
Oh no, they might have to reduce class sizes, implement project-based learning, and provide human feedback instead
I dont recall being graded on math homework in college. Its the recommended practice for exams.
Part of the purpose of rigorous education is to make sure that students are competent for the academic labor force, but that's only half of the story. The other purpose is to preserve existing knowledge and prevent its loss to the force of entropy eating away at texts, files and neurons. If there's no preservation of knowledge in the academic form anymore we have no choice but to rely on AI, which has a massive downside:
In the event that the cascading metal mining yield declines aren't solved by new tech and we do lose AI, we are now in the state that we will probably also lose calculus, liberalism and the ability to design a steam engine.
Grading homework has never made sense anyway. Only grade work done in class.
In-class exams are anyway a proxy for grading homeworks, because those who don't do them will get lower grades.
Blame the game, not the player. Grading is about securing status in a primate social dominance hierarchy. Fuck it - math should transcend that as much as our bodies allow.
I never had a good math teacher in college that graded homework. The good ones told you which problems to do, and they told you that you'd be doomed on the exams if you didn't do them. They also only assigned the problems that had answers in the book.
Making homework a part of your grade is like making attendance a part of your grade. It's something that teachers who aren't capable of teaching do, so they can give students who haven't learned a thing reasonable looking grades. Students have always cheated when homework was graded, it's just that AI makes it so flagrant that it can't be conveniently ignored.
I'm overestimating people, though. There are clearly teachers who are still pretending like students don't cheat when it's like five clicks now. It was much, much harder when tons were still cheating half a century ago. You had to actually meet each other. Admitting that you're collecting a check for nothing is tough, especially because now you're going to have to work.
Who am I kidding? Teachers will just start talking about how grading is bad or racist or something, or saying that using AI is actually better preparation for the future. Hell, they should have done that half a century ago: in real life, being able to cheat well is the best skill.
A few thoughts on this:
- Terry Tao said it best, I think: We're no longer in the heroic age of mathematics. Meaning, I think, that math is now a group project, not just a single person. Many disciplines have become this already, math was just lagging a bit there.
- Other commentors have pointed out that this professor likely hasn't graded anything in decades. I'd bet that he likely hasn't changed tests all that much either. And knowing how extensive the test banks in greek life are, I'd bet then too that many of the undergrad students haven't done any homework or test prep for decades either. Cheating is by no means a new problem. Granted, this is pure horse-crap speculation on me here.
- AI isn't going anywhere, and it seems that this Prof is kinda already throwing in the towel per students. It strikes me as mostly cowardly and asinine and just deaf to anyone beyond his nose. Like, my dude, talk to your students. They are ostensibly adults. They aren't cattle. Figure out a system, together, that actually works.
- My own opinion is that the age of cheap higher education is over. 500+ person freshman lectures at 8:00am? Yeah not with AI. I think it's gonna be more like 4-10 person study sections with a proctor; the Oxbridge system. Yes, that introduces a lot of challenges in the proctor-student relationship. But it's also very expensive. You're gonna need to up your budgets a lot. And unfortunately, that means that this era of mass higher education is also ending. I fear that only the rich, again, are going to go to college.
No worries ~10 more years and it won’t make sense to educate people past a certain point anyway
It’ll be like you’re studying under Aristotle while AI is doing cold fusion next to you - incomprehensible on any level.
Maybe forcing them to turn in the homework hand written would help. They’d probably at least learn something by osmosis if they had to write down what the LLM spit out.
This was always the case. The university system (ignoring research and other functions, since that's mostly decoupled from the type of class in TFA) serves a dual purpose as a teaching system and a credentialing system. How does grading homework (different from correcting homework -- I'll bend over backward to help students understand the material better, and learning from your struggles and mistakes is one of the best ways to make that happen) benefit either?
- There's the paternalistic argument that the adults we send to college aren't real adults and need carrots and sticks to learn properly. Making homework part of the grade helps them help themselves. This is a value judgement on what's more important (squeezing knowledge into people at any cost vs giving people agency and accountability), but I have suggestions as to exactly where people can stuff that argument. IMO, letting fully fledged adults limp through the system without the motivation or wherewithal to do a little hard work in their own self interest isn't doing them any favours.
- On the credentialing side of things, ongoing grading from something like homework helps smooth over the effects of a bad day of sleep or test anxiety. I'm sympathetic to the argument, but I think there are tools which are more accurate and less work for the students and the teachers. Start the semester with an ungraded (corrected if students want to take advantage of it, but not affecting their final grade) take-home test of the background knowledge expected for the course. This lets students decide if they're ready or should take an earlier course instead, do some selected reading on weak points during an otherwise relatively light workload in the semester, and learn the teacher's testing and grading style so that there's less anxiety leading up to a real test. Your grade in the course reflects your best demonstrated mastery on the final, which is arguably closer to what people expect from the credentialing aspect of things; nobody cares how well you executed epsilon-delta proofs during the week you studied them, they care whether you're ready for the next course or ultimate job using that skill. Allow a makeup/retake exam (new questions), maybe two, and take the maximum score. To the extent that your test has any noise, this inflates the overall score, but it mitigates the effect of a single bad day, anxiety, etc, and gives you an easy framework for handling legitimate life events. Still assign homework. Still correct it for students who want feedback. Don't let it influence the grade.
this is what comes for education, the same way as agents can solve home work, they can grade, curate what you learn, what you work on etc.
professors needs to be life coaches and motivators, less of a oral/written book reproducer. That part will be automated. Evals also will be in a locked room, do you home work, bi-weekly do the check, correct with agent what needs to be repeated. Measure if you on course/off course doing better/worse. If you are not doing good, escalate to your coach(professor).
I mean, sure, charisma can engage people into several classes, but also, lets be honest what % of professors really got charisma? If it helps lectures are great, the active work should be done with agents, individually and in the team. of human, and other peoples agents. That's the future of work.
You guys were having homework graded in university? I was taking classes where the prof would just be like "pages X thru Y odd/even" at the end of a lecture with a very "do it or don't do it, I don't give a shit" type of mentality. Your final mark was two tests, a midterm and final.
Ultimately this will impact TAs and adjuncts, who are the ones actually doing the grading.
I never liked grade based education.
Never promoted genuine learning but the ability to get grades.
Live tests based on homework assignments are the only test that works.
I don't see a point, either. Twenty years ago, the only grades we ever had in law school were finals.
Give homework, have TAs go over it with students, have a final, which is the only thing graded.
Classes are for bringing you toward mastery, isn't it? You should pass based on mastery, not effort. I don't give a shit if my pilot tries really hard to fly. I care if he successfully flies.
Profs need to learn how to use AI productively or someone else will.
I don't see the point of universities at all anymore. They were originally places you went because you couldn't learn a thing anywhere else. Now you can learn anything, anywhere.
Example: If you want to learn to do something with your hands, it normally needs to be taught in person, so you can try to do it, fail, the teacher tells you what you did wrong, shows you how, you do it again and it works. But this can now almost entirely be done with AI. A YouTube video shows you how. You practice it on camera. An AI watches the video, tells you where you went wrong, and shows you specific clips to improve. You do it again until you get it right.
The piece of paper you get from a University doesn't need to come from "A University". Any organization that can give you a test in person can see whether you do the thing right or not and give you a certification if you do.
So really all we need now is an AI Khan Academy and independent testing/certification.
This is the wrong approach.
Assume they use AI and grade them based on where they can take it form there.
Cue - Zen and the Art of Motorcycle Maintenance
I had two classes like this, zero grades over the semester, they were among my favorite. Though we didn't really ponder what we were doing there and the means for final grades, for the admins, were both interesting (final test too long and hard, you already know what your grade is)
The irony here is perhaps missed by many. He didn’t grade homework in the first place. He paid assistants to grade. This is common practice among tenured professors with grant money to spend and graduate students beholden to them (I’m not making a moral statement, just noting this is common practice).
Aside from the irony that Borcherds probably hasn’t graded a homework set in a while, instead let’s examine the purpose of homework. It helps you learn. You can read a chapter and think you understand the material, but doing homework makes you engage more with the material. Recently we’ve used the external motivator of course points to encourage this good habit. But this turns the relationship between teacher and student into a transaction- you do a problem, you get a point. And most young students really don’t care beyond that transaction. They are going to college to increase their earning potential, and that’s about it.
So if we’re not going to rethink the purely transactional relationship schools have with most students, then you probably should stop giving points and grading homework. You have to figure out a new transactional structure that can’t have your tuition made irrelevant by a $20 subscription. That’s possible to figure out. A lot of people have pointed out various equity risks but it’s really just a matter of balancing the AI risk with the equity risk at the weighting acceptable to your institution.
You could also attempt the much more challenging prospect of turning the transactional nature of college into some other relationship which fosters a student’s intrinsic motivation of learning for the sake of learning, but I bet your institution goes bankrupt in the process. So really you should just stop grading homework and put 100% of the grade weight into proctored assessments with some sort of firewall against AI tools. Anyone that faces an equity disadvantage because of this was probably also disadvantaged by the fact that you let rich kids with premium AI subscriptions get better grades on their homework.