ChatGPT & Classroom · Posted by Kyle Hebert ·

AI Grading Assistance – What Are You Actually Using?

6

We talk a lot about students using AI. what about teachers using AI for grading and feedback?

I’ve been experimenting with using AI to draft initial feedback on student work, then reviewing and personalizing it before sending. It’s cut my feedback-writing time by about half.

what I do: give the rubric and the student work to the AI, ask for specific feedback aligned to each rubric criterion, then edit the result heavily. the draft gives me a starting point and catches things I might miss at the end of a long marking session.

Am I alone in this? And is this something we should be transparent with students about?

5 replies

5 Replies

6

YES!! I've been doing exactly this for six weeks!! the feedback quality actually went UP because I'm not rushing through 90 essays at midnight. and I share it with my whole department immediately. the rubric-aligned feedback draft is SO much better than starting from scratch every time.

9

The transparency question is worth taking seriously. If we're using AI to draft feedback and students don't know, there's a reasonable argument that's inconsistent with asking students to be transparent about their AI use. probably worth having a department conversation about disclosure norms for teacher AI use as well.

4

the transparency symmetry issue is exactly what our department got stuck on. we expect students to disclose AI use, but we haven't disclosed our own AI use in drafting feedback. one of our grade 11s asked directly: "do you use AI to write comments?" hard question to answer honestly when we haven't had that department conversation yet.

5

worth noting that this approach may need modification for different grade levels. at the elementary level the feedback needs to be simpler and more personal - AI draft feedback often skews too formal for younger students. but the underlying workflow is sound for secondary.

3

I use it for rubric drafting, not grading. it's good at generating first-pass criteria from a learning objective, which I then revise. lets me think about assessment design faster without automating the actual judgment of student work.