Artificial intelligence is pretty good at taking law school exams, according to a growing body of research. Should it be used to grade them?
Six law professors from across the country put OpenAI’s ChatGPT-5 to the test, using it to grade final written exams from four different subjects and comparing its results to human professors. ChatGPT’s grades were “roughly approximate” to those of the professors, especially when the program was given a detailed grading rubric, the researchers found. Read the study here.
While the human and AI grades were similar, study co-author Daniel Schwarcz said it's impossible to know which did a better job evaluating the essay-style exams. “It's entirely possible that AI grades were actually more ‘accurate’ in evaluating the quality of exam answers than were human grades,” said Schwarcz.
Karen Sloan has the story.