I've curious as to the extent that materials should be refreshed between semesters, changing the contexts and updating numbers to avoid students passing on solutions from previous semesters of the same course.

Is the burden of a few hours of instructor work to refresh an assignment wasteful, necessary or somewhere in between to avoid the academic integrity problems of assigning work whose solutions are floating around the student body from previous semesters? I'm motivated not just to run a fair course but to to run an effective one (cheaters who get away with it haven't really learned, have they?).

For better and (in my opinion) worse, the advent of chat GPT and similar AI models ensures that pretty good solutions are already available. Is the cat already out of the bag and there isn't a need to change the coursework semester to semester?

I should specify that I'm really referring to homeworks, which are done outside of class and with open notes and internet. When it comes to exams, with their increased weight and lack of open notes, I think the effort to make fresh problems every semester is well justified.

Of course, the answer will vary a bit based on the content of the class. Some math classes have "classic" problems (graph theory courses seem to have a bunch: e.g. bridges of kronenbourg), and for good reason. My course is an introductory math course and, as such, it has the students do a good amount of computation so they gain experiences necessary to build intuition (e.g. Bayes Rule, converting between numbers in different bases, predicate logic).

2 Answers 2


I don't think there's any need to change the homework questions from one semester to the next, unless you have new ideas and want to make the questions better.

Like it or not, cheating on math homework--and getting away with it--is easy, and copying the solutions from someone who took the course last semester is only one method out of many. Students who want to cheat can always use wolfram or math.stackexchange. Now they can use AI tools as well, but there are plenty of options even without that.

You can punish very blatant cases of copying (solutions that are wrong in an idiosyncratic way, and identical) if you want, but you still have to expect that not all students will do their own homework, and you won't catch most of them. The usual solution is to not give too much weight to homework when calculating course grades. Personally, I wouldn't ever weigh homework more than 30% in a typical college math course. (You don't want to weigh the homework too low either, or else no one will even look at it.)

When it comes to exams, it IS a good use of your time to take reasonable measures to make cheating difficult, because (a) it's worth more of their grade, and (b) you have some hope of catching the people who cheat. This forces students to actually learn in order to pass the exam, whether they did their homework honestly or not.


One scheme I heard of: Each student gets a problem with different numbers (so he cannot merely copy someone else's paper or copy something from the web). That has been tried in the past, but it is a nightmare for the TAs who are grading it.

But now there is a new version: Let it be graded by AI. Let the AI do the slave work.

  • 1
    The nice thing is services like WebAssign will partially randomize questions student by student, but it is the same problem overall. Commented May 24 at 19:59

You must log in to answer this question.

Not the answer you're looking for? Browse other questions tagged .