As an educator with a Masters in this topic I am going to submit myself for possible pitchforks.
Exams have been shit for ages.
I will easily and immediately concede some exams by some professors some of the time are good. But generally? Most exams are crap.
Math Reasoning: If an exam is supposed to provide a snapshot into student learning, then the exam results should demonstrate that learning as a number.
Counter example: Consider an exam thst is cumulative. It contains 10 questions chosen by the professor. A student in the class has perfect knowledge of half the course material, and is perfectly impossible of getting the other half right (confidently in correct). This means that if the exam is to represent the students learning that student should score a 50. But if the professor chooses 10 questions outside this students knowledge set that student gets a zero.
Before you say "it's a fail either way" many mathematics programs have a passing grade on exams at 30% or higher so be careful with grade assumptions as well.
I'll post more later but these are time consuming and... Well I teach lol, I have limited free time.
Suffice to say I will get around to the AI part at some point.