Berkeley Professor Shifts Homework Grading Amid AI Concerns

University of California, Berkeley mathematics professor Richard Borcherds has shifted homework grading toward completion rather than correctness, saying he can no longer reliably determine whether submitted work reflects students’ own understanding or artificial intelligence tools. He describes the approach as a stopgap, with no better alternative currently available to him. Borcherds says AI use on homework has become widespread and that the technology’s improving answers have made it harder to distinguish generated work from students’ work. He has argued that grading for correctness can encourage students to rely on AI and disadvantage those who complete assignments themselves. The reports do not provide class-level measurements of AI use to independently assess how prevalent it is in his course. Under the revised approach, students receive credit mainly for completing assignments, not for getting answers right. Borcherds has also said he relies primarily on exams administered in person, where students cannot use electronic devices. He previously described online exams as vulnerable to AI use and said average scores rise substantially during them, though no figures were...