Frequent generative AI use is associated with higher cheating rates among students
A commentary from the WeChat account 不懂经 reports that a UC Berkeley study of 95,513 undergraduates across 20 US public research universities found that 26% of students who used generative AI daily admitted using it to cheat, compared with 7% of those who used it about once a month. The commentary also summarizes research by Georgetown University neuroscientist Adam Green’s team, which analyzed more than 370,000 US college application essays. It reports that post-ChatGPT essays were more fluent and received higher creativity scores from human evaluators, but were also more similar in their underlying ideas.
The report highlights a distinction between using generative AI to deepen understanding and using it to replace the learning process. It also illustrates how improved language can coincide with less diverse ideas, although the reported association between frequent AI use and cheating does not establish causation.
每天都用AI的大学生里，有26%承认用它作过弊。每月才用一次的学生里，这个比例只有7%。这个数字来自一项覆盖9.5万名本科生、发表在《Science》上的研究。它戳破了一个我们深信不疑的说法："多用AI，就能跟上时代。"用得越多，未必越强。有些人越用越强，有些人越用越空，可从外表上看，你根本分不出谁是谁。20......
本文来自微信公众号： 不懂经 ，作者：不懂经也叔的Rust
每天都用AI的大学生里，有26%承认用它作过弊。每月才用一次的学生里，这个比例只有7%。
这个数字来自一项覆盖9.5万名本科生、发表在《Science》上的研究。它戳破了一个我们深信不疑的说法："多用AI，就能跟上时代。"
#University of California, Berkeley#Georgetown University#Adam Green#ChatGPT