Giving Feedback at Scale: Our Journey into Fine-Tuning GPT for Education
We have all felt that moment of frustration: you put hours of thought into a complex response, only to receive a generic "Good job" or a cold numerical score. As a researcher at Acuity Insights, I have spent years looking at how we can make educational assessment more human, even when we are dealing with thousands of students.
My team and I recently presented our findings on this challenge at The 40th ACM/SIGAPP Symposium on Applied Computing (SAC '25) in Catania, Italy. We wanted to know: Can we actually teach an AI to give feedback that feels personal, supportive, and—most importantly—useful?
DOI: 10.1145/3672608.3707735