Remaking grade the
Generative AI is rewriting the standards for business school assessment; here EDHEC’s Peter Daly and Emmanuelle Deglaire consider how large language models such as ChatGPT might help improve the way faculty assesses students’ work
that had already been graded by experienced professors – and fed them into a custom AI tool based on ChatGPT. These assignments were primarily from law and management courses and involved structured essay-type responses graded with detailed rubrics. We then compared the grades and feedback generated by the AI tool with the original human assessments. The results were encouraging. In more than 80 per cent of the cases, the grades produced by AI were within a one-point margin of those given by the professors on a 20-point scale. While this variance might seem significant to those outside academia, for educators familiar with the inherent subjectivity in grading, such a margin is considered entirely acceptable. It is not uncommon for two experienced faculty members to differ slightly in their evaluations, particularly when grading involves complex or open-ended responses. Encouraged by this alignment, we decided to take the research a step further and test the system in real classroom conditions. During the 2023/24 academic year, Emmanuelle Deglaire used an AI-based tool in her law courses to grade 600 student assignments. The tool she used, which was developed in partnership with an external provider, was tailored specifically to her curriculum and
I s traditional paper grading on its way out? At EDHEC Business School, this question has driven a research project we’ve been working on for the past two years. Our research began with a straightforward hypothesis: given that AI can mimic human language and reasoning fairly well, then it’s reasonable to imagine it might be able to evaluate written assignments with a degree of accuracy comparable to that of human graders. We also wanted to consider whether students and faculty would be open to such a shift. From controlled testing to real classrooms The project started as a controlled experiment. We selected anonymised student papers from previous academic years – papers
34 Ambition • ISSUE 6 • 2025
Made with FlippingBook - Share PDF online