You'd get a stack of 120 blue books to grade in a week's time a few times a quarter.
The grading was entirely just checking if the student used a set of key words and had a certain length. This was a near universal method across the University for blue book exams.
Honestly, an LLM would be a better grader than most stressed out grad students.
Everyone has been phoning it in for a few centuries now
No issues to me in using LLM for suggestive grading assuming we have some evidence on its grading rubric and paper trail to audit for appeals to human review - ie human teacher is responsible not LLM
Any audits will be quickly farmed out to yet another AI for review, is my guess.
I'd imagine some system like YTs appeals system, where everyone is maximally unhappy.
One anecdote from my SO's time as a grader was that pre med students were the worst. They would just wear you down to get the best possible grade, appealing literally every missed point ad nauseum. Most profs would give in eventually in the undergrad classes and not deal with them. Of course further emboldening them.
No other major was like that, only those dealing with the future hellscape that was US healthcare.
I'd imagine that, yes, eventually your appeals in the AI future will end up at a prof, but delayed to hell and back. Even paying $200k+ won't matter.
You'd get a stack of 120 blue books to grade in a week's time a few times a quarter.
The grading was entirely just checking if the student used a set of key words and had a certain length. This was a near universal method across the University for blue book exams.
Honestly, an LLM would be a better grader than most stressed out grad students.
Everyone has been phoning it in for a few centuries now