Teaching is a slow process of becoming everything you hate

But stupidity is relative and stupid children are more common. The heuristic can be reworded as adult parents are more knowledgeable and capable than children.

Put the average child in the position of the average parent and you will see the difference.

If the real objective is learning goals/outcome rather than (or in addition to) a working temperature circuit, then that objective has to be somehow encoded into the requirements. Or else, sometimes all the stated requirements will be met without that unstated one being hit.

This is difficult because, for instance, the possibility of cheating means that the person who says they performed the assignment might have contracted it off to someone else and learned nothing.

Someone who already has all the required knowledge can also just spin out the assignment without learning anything.

Basically, learning is a state change in the pupil; if you want to validate that some state change occurred, you have to have a way of measuring the state before and after and calculating a difference.

i always personally preferred project work. i was never great at exams (although i did get a whole hell of a lot better at them). project work always better suited my obsessive personality and desire to really polish things. projects feel creative, homeworks and exam prep... don't. (although learning how to take exams meant learning how to make good cheat sheets and memorizing them well, so in a way it became creative)

that said for most lower division material projects are unsuitable, for that stuff the system i saw i liked the most i first saw online for an undergrad intro ai course at mit. it was pretty simple, the course had a handful of carefully designed uncurved but not tricky half exams units that were given throughout the term. the final was two half exam spaces for any units you wanted to try again, if you did well all semester, you didn't have to show up for the final, if you messed up, it's your chance to retake the specific units you wish to improve. goal: demonstrate you learned all the techniques in the course, that's it.

sometimes it felt like putting more weight on homeworks was for student comfort and to reduce stress on exams for everyone, sadly sometimes i think it had the opposite effect of producing lazier exam design and more reliance on curves. i once took a course which had no official notes, fairly weak lectures and the claim "i teach at a level above the assigned textbook." no, he didn't, he wasted everyone's time.

i once went to see a professor after the fact to go over the final, i told him explicitly i just wanted to understand the things i got wrong but he kept returning points even after multiple statements that i didn't care. this made me very sad to think that he probably sat for hours with people arguing over points rather than discussing material.

overall it felt like some professors (or maybe their students) spent hundreds of hours designing amazing courses and some spent less than ten. those in the former camp were often prickly in terms of their specific asks, but obviously in those cases it didn't matter as the care and craftsmanship that went into the course design justified any particularity. it was the waste of time courses that were the worst (even if they did sometimes come with generous mea culpa grading).

One of my favorite teachers in high school had the following policies:

* There will be a short quiz every week covering recent material.

* Homework is optional for any student who got an A on the last quiz (due to the length of the quizzes, that essentially meant 100%).

* Anyone with an A average in the class so far and an A on the last quiz is permitted to sleep in class.

It worked great. Nobody's time was being wasted on busy work, nor were people recklessly left behind.

This works great, and it is still how evaluation is done when the stakes are higher: PhD defenses, executive hiring... even getting hired as an entry-level engineer at Google requires about five hours of what is basically oral examination.

But society is not willing to pay that kind of price for the earlier levels of evaluation. We want "scalable" systems. Unfortunately those same evaluations are often treated as more sensitive than they really are. For example, if you're comparing two students I'd argue that 3.0 vs 3.5 GPA gives you at least some signal, whereas 3.5 vs 3.6 GPA gives you basically no signal at all (maybe the 3.6 student took easier courses, maybe they were more lucky with cutoffs, etc.). And yet the distinction sometimes matters e.g. to graduate programs.

In well-designed systems, the GPA cutoff is set relatively low and more sensitive methods are used to select the best students from the pool. Often this includes an interview with a professor, which is also a form of oral exam.