New Cardiff University Study: AI Still Cannot Grade Assignments as Reliably as Human Teachers

Summary

A new study from Cardiff University and the University of Melbourne found that generative AI is not yet able to reliably grade students' written assignments like human teachers. Researchers used two versions of ChatGPT to score 50 undergraduate biology essays, finding that AI scores varied widely, could not accurately predict human scores, and tended to lower high scores and raise low scores, compressing scores toward the middle.