能拿高分的技能，正是 AI 最擅长伪造的

An experiment at Bocconi University with over 1,000 freshmen shows that GPT-4o significantly boosted grades on a business assignment. A short lesson on causal reasoning didn't improve traditional scores but pushed students toward more diverse and unusual solutions. Whether AI use also improved actual learning remains unanswered. There was no follow-up test without ChatGPT. ChatGPT significantly improved student work, while a teaching intervention encouraged more diverse and unusual solutions. Whether AI also improves learning remains an open question. What makes a good student paper, and which parts of that can AI improve? A randomized experiment with 1,053 freshmen at Bocconi University found that GPT-4o helped students earn significantly better grades on a business assignment. A short lesson on causal reasoning didn't raise traditional scores, but it pushed students toward more diverse solutions and deeper thinking about causes, assumptions, and how their proposals would actually work. In November 2025, 13 sections of an introductory management course were randomly split into four groups: control, causal reasoning lesson, GPT-4o access, or both. Students had to write marketing recommendations for the university's merchandise shop in up to 180 words. The lesson covered coherent causal logic, falsifiability, and how a proposed action might lead to a desired outcome. Ad Students with GPT-4o scored nearly a full point higher on a 1-to-5 scale. Their answers contained about two more ideas on average, were more logically coherent, and more closely matched the recommendations of three subject-matter experts. Even after the researchers controlled for argumentation quality, number of ideas, idea diversity, and text properties, a measurable GPT-4o advantage remained. The authors attribute this to higher content quality, not greater student knowledge. Ad The causal reasoning lesson didn't raise traditional scores. Work from those students actually scored slightly worse on average, but they more often explained why a proposed action should work and under what conditions it might fail. They also generated more diverse ideas that diverged from what their peers wrote. Combining the lesson with GPT-4o didn't add any further boost to traditional scores. On causal reasoning markers, the two approaches complemented each other, and the greater idea diversity from the lesson group held up. Ad More ideas, more coherent arguments, and greater idea diversity within a single answer all correlated with higher scores. But stronger falsifiability, more detailed explanations of how proposed actions would work, and greater divergence from other students' ideas correlated with lower scores. That doesn't mean originality was penalized across the board. For this assignment, the traditional score mainly rewarded well-structured answers that stayed within the expected solution space. The authors conclude that diversity and originality need to be explicitly built into grading criteria if they're supposed to count. Ad But current grading systems measure polish, structure, and completeness, but not learning and understanding. That makes it easy to use AI as a cheating tool. Ad