Exams as a prerequisite for a degree are useless. Exams as part of a degree must be abolished. And degrees must be abolished. Ultimately the only metrics that really matter are money and fame. See also: motivation.
The only thing exams should matter for is as a screening tool to select people with specific abilities that you care about as an employer or principal investigator. If:
  • you have no idea about what the content of specific exams are (and you don't because they are all ad-hoc university secrets)
  • or don't have a way to machine learn what grades correlate with your desired performance (you don't because where's the data?)
then exams are useless for your purposes. then might as well just go by interviews (basically what all employers do already, though not PIs). Degrees are too course grained to mean anything to anybody. Employers and PIs likely only care about very few specific subjects.
Once the question of an exam has been formulated, the usefulness of the problem is already been completely destroyed, because formulating the problem that matters is the most important part of things. And any problem with an answer, is useless to put effort into: give answers.
Furthermore, preventing people from searching for answers while answering an exam, AKA preventing "cheating", also makes absolutely no sense. In the real world, we want people to find answers as quickly as possible! We should be teaching people how to "cheat"! What we should teach them instead is what a fucking license is, and what you have to do to comply with it.
And if you pass the exam, you pass the course, without any further time requirements.
And those exams must be applied by professional test application companies to ensure no cheating and to factor out the anti-cheat work, while still making the tests available to people anywhere.
A quote from Richard Feynman present in the book Surely You're Joking, Mr. Feynman chapter O Americano, Outra Vez!:
You cannot get educated by this self-propagating system in which people study to pass exams, and teach others to pass exams, but nobody knows anything.
You learn something by doing it yourself, by asking questions, by thinking, and by experimenting.
The only metric that matters is "to feel that you've satisfied youre curiosity". When one studies for that, it can take a lot more time to actually learn everything, because it is sometimes not as clear when you should stop. But it is the only way to go deeper.
A person's understanding is the most illiquid asset that exists, to judge that based only on standardized exams, is a certain way to fail to identify top talent.
This is Ciro Santilli's name for the idea that we should not have structured degrees at university that require entry exams, only tests that anybondy could take, likely for free, and then they would just have proof that they know the stuff for e.g. teachers that care about a subject while selecting students to work with them in research.
We just need control rooms where someone can watch students for cheating. Multiple different exams can be taken in the same room of course, students just have to sign up in advance. The exams should happen regularly depending on demand. E.g. extremelly common subjects should happen every month, and highly specialized ones every 6 months or 1 year.
Questions should be always taken from an open question pool which also contains answers, thus allowing anyone to effectively study for it.
How many questions can you actually come up with about a given non research subject, right?
We then make an API available, so that students can grant access to specific results to anyone they choose, or even make the results public for anyone to see. This way the people that care about the exams can just machine learn what exams correlate with their desired performance.
See: exam.
Same remarks as Section "Exam".