Can universities trust AI to mark student essays fairly?
AI like ChatGPT can't fairly mark essays, often giving higher grades and missing key criteria.
Australian universities are exploring AI to help mark student work. The technology isn't ready yet.
A study by the University of Melbourne and Cardiff University tested ChatGPT on 50 essays. Some marks matched human grades.
ChatGPT struggled with individual criteria. It couldn't tell if content fit.
It couldn't judge if the evidence was strong enough. ChatGPT gave higher marks than human markers did.
This shows how AI works. It predicts patterns instead of judging quality.
Universities using AI for marking require human oversight. Deakin University bans uploading student work to public AI tools like ChatGPT.
This is because of privacy concerns. Some universities are rethinking grades.
They're moving away from marks. Instead they check if students meet defined learning goals.
- 50
- Essays tested
- ChatGPT
- AI tested
Why it mattersAI marking alone could hurt student learning and damage teacher relationships. It could also cause teachers to miss problems in their own teaching.
AustraliaAustralian universities are navigating when and how to use AI in assessment; students should know whether their work is marked by a human or supported by a machine.
