Hacker News Reader

Comment by bradley13 | Hacker News Reader

I'm a prof. The easy solution is to give very little weight to homework. The assignments should be designed to teach concepts. Concepts that are then tested in a controlled environment, where cheating is not possible.

Remove the incentive to cheat, and save yourself the time trying to catch it (and punish it, despite an uncooperative administration).

View on HN

Problem is, homework is also what most closely matches real work. Tests are good for controlling the environment (we need them too!) but some students will just naturally be better at homework. Also, incentivizing students to put time and effort into the homework builds work ethic which will serve them well in the working world.

Personally, the best classes I had all had rigorous homework assignments. I would learn much more from them than studying for the tests. In fact, doing the homeworks would generally cover more than would be possible to test.

Plus I'm just glad to have built things like a DNS server, and inode filesysyem. Many small games and web servers. Database applications, shells, and compilers just to name a few. These are all things that give me confidence as a programmer.

> Problem is, homework is also what most closely matches real work.

...I wonder if this is true. "Real work" sometimes requires you to perform in the moment. You can't always just pull an all-nighter to get it done if you're slow.

Also, the best way to do better on a test is to do homework to prepare for the test. And teachers can still assign homework to help students learn, they just can't necessarily use the results for assessment.

I guess tests are like being on call, supporting outages and the like, while homework more closely mirrors feature development. We also had a number of group assignments and labs which required on-demand work but were more project based than tests.

My main point is that while testing knowledge in small bits is useful for grading, it's less where I got value as a learner.

Moreover, getting stuck in homework assignments and learning how to get unstuck is what, I think, propels you forward the most.

The biggest value I received in my undergraduate courses was from the classes where I built large projects. Building a compiler from scratch, piece by awful piece. Implementing a ray tracer, from splines to shaders. Even if I don't use compilers or graphics in my day-to-day work, it was my experience of building those larger-scale systems and working through the problems that benefited me.

As a teacher I just don't know how to replicate that experience in a world where half will skip the work, and all I'm able to do is a two-hour sanity check that they learned some concepts. How do I test for "the experience you developed spending four hours making that register allocator behave properly, to the point where you're intimately familiar with every failure mode"?

>How do I test for "the experience you developed spending four hours making that register allocator behave properly, to the point where you're intimately familiar with every failure mode"?

ask them to list and describe failure modes on your exam.

They’ll ask someone else to share a list. It won’t mean anything. To give an analogy: imagine we taught new drivers to drive from lectures and a written exam, without ever requiring them to log any hours in a car. It doesn’t work there, and it doesn’t work well anywhere else.

I really enjoyed how they did it in Tübingen, where I did my masters. You usually had to achieve a certain score on the homework (usually 50-75% of the overall points) in order to be qualified to take the exam. Additionally, if you did really well you got a few bonus points on the exam (something around 80/90%+). This incentivised you to take the homework seriously, especially if you want to get the bonus points. But you still have the written exam at the end as the "controlled environment" and something you have to prepare for, so you can not just listen to the slides once and then forget it after doing the single exercise

The grade system we implemented was a little program: you can pass the class if you get a passing grade on your HW and tests (weighted average); if you fail either HW or exams, you get the lower of the two.

This solution is heavily dependent on the material.

An intro to CS class where every answer is a few lines of pseudocode might be able to effectively teach the material by forcing everything on to the test.

A software engineering class that just test your knowledge of memorizing patterns and makes the project where you actually implement real software incorporating the patterns worth nothing would be useless to actually teaching students anything they can use after graduation.

Agreed. Generations of students had their degree classifications determined by a small number of final exams under exam conditions. Why did we move away from that, opening up all of these routes to score degree grading in non exam conditions?

Just my personal theory, but maybe universities wanted to avoid being considered just pay-to-test institutions that couldn't justify their insane and continuously rising costs when the only thing that really mattered for a degree was passing an exam. Someone spending thousands of dollars just to skip every class but then show up and pass the exam test is not a good look for the university and could devalue the entire thing from the perspective of the public and potential future students. But if someone is required to go to every class nobody can easily claim the classes were pointless or bogus for passing exams and getting a degree because nobody can do that. Increases in overall schooling costs can also be explained away as more invested in classes whether it was true or not. And it makes people who were forced to attend classes whether they wanted to or needed to look disfavorably upon the potential for future students to be allowed to simply pass an exam by completely itself, in a sort of bucket of crabs situation where people think "I had to go through all this bogus stuff for my piece of paper, these younger kids should have to too!"

But I have zero qualifications for any of these opinions other than having been forced to attend a number of nearly worthless university classes that were pointless in the face of just reading the course book. Of course I also had some classes that were worth way more than just the book material, but probably half the classes I had to take I felt were dubiously useful to start with, not to mention the absolutely terrible actual class, and forcing people to attend to get a passing grade was just there to prevent 95% of the class being empty every week.

Lectures were originally created as a way to share a single book. Since printing was expensive, a reader would stand at a lectern and read the material aloud. You might say "why don't we have all the students share a single book, that worked fine" but the truth is it was a crummy solution to a resource constraint problem.

Similarly, exam-only courses are an excellent way to teach huge numbers of students without the costly hassle of grading homework assignments. But that doesn't mean they're an optimal solution.

From my experience as a student and now a professor, nothing can possibly compare to the benefit you get from hands-on learning. You get so much more understanding from the 20 hours you spent making a complex system work, than you will ever get in three 3-6 hours of in-class instruction you'd get during the same timeframe. And I have no good idea how to test for all the skills you learn from "writing a tiny OS kernel from scratch in C" or "building a compiler" or "implementing a complicated cryptographic protocol and then realizing an attack on it." I do observe that my students who do the homework tend to do much better on the exams, but I'm concerned that the incentive to take shortcuts will be much too high if homework isn't required.

The easy bad solution: this also gives little incentive to do homework and thus actually learn, instead perpetuating the bad practice of cramming learning before one big "controlled environment" exam.

From my own university experience . . . admittedly, many years ago . . . for me, the assignments were a distraction from learning. Regardless, it's not the professor's job to force students to learn; it's the student's job to learn if they want to.

That's a separate issue, but then the solution is not to have assignments at all to avoid distraction, not have them without grading.

The other dichotomy is also strange, especially since the only person with an actual job is the professor - it is in fact his direct responsibility to structure the course, including incentives, to achieve the best learning.

The assignments just teach different things. Arguably the most important things, because they are what is most required to succeed at a job. The theory matters at work too, but actually writing code that works tends to be what matters most.

> for me, the assignments were a distraction from learning.

How? I can vaguely see it if the assignments were truly out of step with the material, but I cannot understand how relevant assignments are a distraction. In mathematics, and in all CS courses I have taught, assignments are at the core of what you are learning, and the best feedback (to yourself and your teacher) that you have a working knowledge of the material.

If I already understood the material, assignments became busywork that took time away from learning other things. I was not there for grades; I was there to learn. For this reason, I rarely did the assignments and I willingly took the hit to my grade. It is important to understand that not everybody learns in the same way. I have no problem with assignments as a method of learning. I have a problem with being "forced" to do them when I don't find them terribly helpful.

No, that works well IME. If it's worth something towards the final grade, even 1%, most students will do it. It can be hard to persuade some of my students not to spend multiple hours attempting to get 0.1% more of the course grade by doing another quiz attempt when they've already achieved 90% - I think they're better off moving on to the next thing.

Make the homework compulsory but valid 0 points.

Then many will submit blank papers, or things so low effort that they're as good as good as.

Still needs to be passing...

At the university level, should you really take students by the hand like that and trick them into learning? You have to learn to learn at some point.

How is your argument relevant to cheating? And why are the universities not at the university level since they mandate this practice?

You were saying that assignments are needed otherwise students won't do them and then fail the exam. That's what I was answering to. And it's relevant to cheating on assignments because the thread was about not grading assignments to avoid people cheating on assignments.

I don't understand your second question. Currently, university assignments are graded to hold students by the hand. I'm saying they shouldn't do that.

> And it's relevant to cheating on assignments because the thread was about

The thread was abou that, but your argument wasn't, it suggested an alternative reason for rejecting assignments, so would be true regardless of cheating, thus not related to this thread

It's related in the sense I think stopping cheating should take priority over taking students by the hand.

But I guess I let through that yes, I don't think we should take students by the hand at all and you're right that's a whole different debate.

The incentive to cheat is the money paid to participate in the undergraduate program plus the career expected to follow from it. Remove those and you will have no students. But what you describe wouldn't remove the incentive to cheat, just make cheating a different and presumably more difficult process. But as long as it's easier to cheat than not to cheat, rational students will continue to cheat.

That’s assuming the administration will allow you to administer the exams onsite which is increasingly not the case. Online students bring in more money.

Reply on news.ycombinator.com