禁用AI後成績暴跌50%——學術界是否面臨作弊危機？
布朗大學的一項研究發現，當監考考試取代了允許AI使用的開卷考試後，學生平均成績從96%暴跌至48%，這表明普遍存在的學術不端行為正在挑戰傳統的評估模式。
a brown university investigation revealed that student grades crashed from a historical high of 96% to a record low of 48.6% after an economics professor moved the final from an unsupervised take-home format to an in-person, proctored setting. the professor noted that students utilized the same obscure mathematical proofs as chatgpt rather than standard methods, and data from 26,000 students in china and 500,000 grades at a us university confirm that ai usage inflates homework scores while degrading actual exam performance. these findings expose a growing disconnect between ai-assisted assignments and genuine learning, forcing institutions to rethink grading policies and assessment integrity.