Trying a show your prompts policy instead of banning ChatGPT outright
this year instead of a flat ban or a free for all i’m trying something in between. students can use chatgpt for brainstorming and outlining but they have to submit their prompt history alongside the final draft. no prompts submitted means the assignment isnt accepted for grading. i tested this with a small summer school group and honestly the prompt logs were more revealing than i expected, you can see exactly where a student got stuck and leaned on the tool versus where they were just using it to organize their own ideas. still working out how much weight to put on the prompts themselves versus the final product. anyone tried something similar for a full semester? worried this becomes a paperwork nightmare by november.
5 Replies
Join the discussion.
Log In to Reply
honestly same lol we tried the prompt log thing in my department too and half the kids just typed something vague like help me write an essay about theme and called it a prompt. might need clearer expectations on what counts as documentation before this actually works.
reminds me a bit of the cell phone ban debates, the rule that sounds clean on paper gets messy the second real students interact with it. genuine question though, do you read every single prompt log or just spot check? because reading forty kids worth of chatgpt transcripts every week sounds like a second job.
@Ian Takahashi spot check mostly, i skim all of them but only dig deep if something in the final draft doesnt match the students usual voice. @Ashley Lemieux thats a fair point, i think i need an actual rubric for what a documented prompt looks like instead of just asking for logs.
this. 100% this. the vague prompt problem is universal, we hit the same wall. a simple rubric fixed it for us within two weeks.
did something close to this last year. built a simple google form for prompt submission so it wasnt just pasted text in a doc, way less chaotic to grade. by the end of the semester i had actual data on which units drove the most ai reliance, turned out to be thesis statements and conclusions, every single time.