After just starting a grad program after 12 years in industry. I'd have to disagree. While a large fraction of homework is busy work designed to give the illusion of challenge and rigor - tests simply estimate whether someone has memorized the material sufficiently for a short 1 hour exam.

In CS, a ~4-20 hour project is vastly more representative of how well someone understands the material and could apply it in a real world setting than a 40 minute multiple choice exam. At the advanced levels this is true for fields such as Physics, English, History or any others.

Maybe we should ask ourselves how to give better assignments in a class that aren't simply busy work?

Many engineering programs have their most challenging courses set up as semester long projects.

In chemical engineering the final boss is the process design class, a project where you are asked to produce a chemical substance with desired properties at scale without losing money. Almost everything you learned during the program has to be used to pull it off. Programming, numerical methods, CAD, Transport phenomena, kinetics, physical chemistry, thermodynamics. It really is the best all around test for a chemical engineer.

While this is feasible for the senior year, I am not sure if you can convert for example calculus 1 into a semester long project.

Calculus 1 is an interesting subject as there certainly is a degree of memorization required (you can't re-derive the derivative of x^n every time it comes up in your career). There is a similar to intro to Organic Chemistry, Algorithms and DataStructures, intro to programming etc. But the goal is to build detailed understanding of these methods more so than memorization.

On the other hand we live in a world where access to derivative rules is trivial. I'd imagine in 1800 mathematicians would assume that you would need to have multiplication tables to be productive and not reduced to pen and paper their entire career.

I wonder if there is an opportunity to push more challenging material into the earlier classes and make them more project like.

I am currently in the last few days before submitting my Chemical Engineering Design Project (I'm designing a packed bed methanol reactor), and yes I can confirm it is absolutely fucking brutal and hands down the hardest thing I've done in my life so far

I get what you're saying but I also disagree with it as a generalization, and say it would depend on the subject. For theoretical subjects, an exam is about the only way to test your understanding. Memorization is not going to help you solve math problems.

I was a physics undergrad who hopped into a few grad classes, and to be honest I was terrible at homework and great at exams (mostly due to some youthful obstinance on putting the time in on homework). At the time I believed that the exams showed who really knew the material and who applied time to solve the problem. With some time past I see that the larger/tougher problem sets were where the real challenge was.

I recall a few unique problem sets from Graduate QM such as

- Derive from first principles the color of the sky.

- Prove that charge must be Quantized if there is one magnetic mono-pole in the universe.

The exam questions were far simpler than the theory questions asked in the problem sets. The work for the first question easily totals > 20 hours of pen and paper time.

> The work for the first question easily totals > 20 hours of pen and paper time.

I guess grad students generally take less coursework than undergrads, but how could a professor expect students to have 20+ hours on hand to solve a single question, given other demands on a student's time?

Grad students usually take 1-2 classes at a time, and the problem sets are spread out over 2 weeks.

A problem like the above would be given as a single problem for students to solve over 2 weeks.

I had an undergraduate lab where I spent 20 hours per week on the lab write ups.

That's what motivated me to switch away from a physics major.

> Memorization is not going to help you solve math problems.

On the contrary, memorization is the way most people I know got through most of their math classes, at least through calculus and linear algebra. You memorize the steps by rote repetition without really learning why they work, then the test is mostly an exercise in guessing which steps and formulas you should apply to the given problem.

Is that really memorization? Memorizing multiplication tables is one thing. Practicing the techniques over and over isn't memorization imo. In grad level maths, you are solving proofs pretty much, you can't just memorize facts in a textbook to do that.

It's memorization insofar as you can do all of that practice and become proficient at solving math problems without really knowing what they mean or why the steps work. You're regurgitating what you were taught, not making connections and using your understanding.

You used math as an example of a subject where tests are used to check understanding. I disagree, because most people that I know who did well in math did so by being good human computers, not by understanding anything.

I expect that doesn't continue to be true at the grad level, but most people don't get that far.

I’m someone who crammed their way through 4 years of computer engineering exams at a challenging university. It’s possible. It’s hard and the worst few weeks of life before exams, but it’s possible.

Cramming is not memorization. It's not optimal studying, sure, but you've still learned something.

In my experience there's little long term retention from cramming.

Can confirm. There's 0 retention. Maybe if I kept cramming over an extended period of time I could retain it. Typically though I stop after taking the exam so within about a week or two things I thought I understood disappeared.

>tests simply estimate whether someone has memorized the material sufficiently for a short 1 hour exam.

I feel a deep sadness reading this. Is your computer science curriculum more accurately described as a software engineering curriculum?

Memorization should be virtually irrelevant on most computer science exams. Proofs should be core to computer science exams; the ability to reason is the most fundamental skill to all scientists, especially for fields which are tightly coupled to mathematics.

> Is your computer science curriculum more accurately described as a software engineering curriculum?

Given that most CS students want to go into software engineering, it would surprise me if this isn't the case for most CS curriculums. In my experience CS students don't generally want to be scientists, so most CS classes are more application-oriented than proof-oriented.

Schools are starting to provide separate software engineering programs, but we're not all the way there yet.

Graded assignments are useful to give feedback to students. And more importantly they force students to work regularly and not wait for the last minute to study.

Thus we see the problem that universities are admitting students who aren't ready for tertiary education.

I think it is mostly the latter. At least -- I rarely got useful feedback other than a little x (best case it would be on the error, more likely on the questions).

Personally, when grading I keep a file of all my feedback so I can easily copy-paste it into their feedback files (since everything is digital nowadays). For a given assignment, usually only a handful of mistakes are made (repeated by each student). If anything, having the file makes my grading more consistent -- same points for the same error.

I'm under the impression that this is a not-unpopular system, but try as I might, I cannot get anyone else to adopt it.

Personally I always preferred quizzes for that. I've always been a very strong autodidact though, there are probably people who prefer getting dragged through things by homework.

This is tough because it creates a strong incentive for them to make bad long-term decisions. Think of it from the perspective of a student: you're taking 6 other courses, all of them very demanding with graded assignments, except for this one class where the assignments are not graded. You have a limited budget of time over the week, and time is getting short. Do you: a) work really hard on your ungraded assignment and turn in your best effort for no impact on your grade or b) tell yourself that you'll make up the work at a later point in time, and then focus on your other graded assignments to make sure you optimize those grade. Then you will focus on the other course later on during spring break or something.

Sure everyone says they'll do a but really, this sets a lot of students up for a trap. They think they will have time to make all of this up later, but really what will happen is they will just fall behind in the class. The assignments from other courses keep piling up, so the free time never really materializes. In fact, the same scenario repeats: the student will forego a second assignment, having already done so once before. Then the deferred responsibilities pile up and you end up with a student who is failing your course (even though on paper the grade is undetermined (kind of like a wave function), in all actuality it's just waiting to collapse to a grade of F at test time.

Look at it this way: it's like a reinforcement learning problem. If your reward schedule is that you only give a reward to the agent when it achieves the end goal, sometimes training that agent takes a very long time; if the search space is too large, then the agent can go any which way and will take a long time to reach that goal. That's ungraded assignments.

Instead, if you give the agent little rewards along the way when it makes some significant progress, then the agent can converge to the goal state much faster, in a way that avoid a lot of unpleasantness for everyone. I don't like giving Fs, and they don't like receiving Fs. I feel like if I give an F that's really more on me than them. Part of my job is not just to put course content into student brains, but to also shape their ability to manage their time and juggle a variety of projects. It's the kind of thing I spend many semesters (4) instilling in my students and grades are one of the effective tools I use to do so.

You may say just do away with all grades and we can talk about that. There are different models we could use. But as long as others are using grades it's kind of a baked in assumption at this point. Very hard to change that kind of system.

It does sound like a pointless arms race (between different courses)

I majored in Law but took a couple CS courses on the side so I saw the contrast between traditions in different departments. CS courses had a constant stream of non-trivial graded homework. Even if I knew the materials it took me quite some time to complete them. Law courses usually one essay that counts for ~15-25% (or less frequently, a mid-term test), and the rest is the final exam.

Obviously, both methods work (I guess). But if you're already in an environment where courses give out lots of graded assignments, your concerns definitely make sense.

When I was a college student, I wasn't diligent enough to always do ungraded assignments. I'd do reading, but for actual questions I'd only do them maybe 50% of the time when preparing for a test on the material. And out of my peers I felt like even doing assigned reading and trying to do ungraded homework put me well ahead of the pack.

I think it's a maturity thing. Probably until I was ~24 I just didn't have the executive function to be able to do things like that. It seems beneficial to have graded assignments as a forcing function especially given some college students are literally teenagers.

Also, I took an accelerated math curriculum as a freshman where I went from never having written a proof/knowing how to prove something rigorously, to pretty good at it. The feedback from the assigned homework was absolutely crucial in helping me learn these skills. It's easy to follow a proof from the answer section, but since there are usually several ways to prove something, it doesn't always help just to see an answer, plus you don't know what kind of divergences/hand waves are acceptable or not without feedback.

Professors grade the assignments to make sure that students do them.

There are many other options for evaluating the students, but not many to force them to learn something.