From the Daily Cal: A study released by UC Berkeley’s Center for Studies in Higher Education found a correlation between the release of ChatGPT in 2022 and an increased share of A grades for classes with take-home assignments at a selective public university in Texas. However, Berkeleytime data collected by The Daily Californian found no equivalent trends at UC Berkeley. Between 2019 and spring 2026, A’s at UC Berkeley consistently stayed between 30% and 35%, with a dip during the COVID-19 pandemic likely due to increased pass/no pass courses.
The study, authored by Igor Chirikov, a senior researcher and director of the Student Experience in the Research University Consortium at the Center for Studies in Higher Education, analyzed 500,000 grades from 2018-25 and found that the share of A grades increased by 13 percentage points at the Texas university. “The pattern shows grades rising mainly where students can use AI on unsupervised take-home work,” Chirikov said in an email. “This strongly suggests that AI is substituting for student effort rather than improving underlying skills.”
While the study found that grades have been increasing in classes with lots of take-home assignments, grades across UC Berkeley have remained relatively stable. The Daily Cal scraped more than 30,000 course grades beginning in the summer of 2019 from Berkeleytime to assess grade trends at UC Berkeley compared to the university in the study...
Josh Hug, a professor in the electrical engineering and computer sciences department, has found that the amount of AI-generated work submitted by students has increased considerably, but said grades have decreased. In EECS and computer science courses, some classes’ failing grades have more than doubled in the spring 2026 semester. Professors attributed the increase to an increased reliance on AI, a lack of math preparedness and understaffing within classes.
“The reason is that they’re still getting full scores on the assignments, but they’re doing much worse on exams,” Hug said. To adapt to AI, Hug has been incorporating frequent in-class quizzes, oral check-ins and “diagnostic questions” on exams. Hug reuses questions from past exams and compares the performance between the two groups of students. “I have found that students do far worse on these problems than they used to, which is my biggest piece of evidence that they don’t quite know what they’re doing in the way they used to,” Hug said.
...History professor Mark Brilliant has run into similar challenges, with students especially struggling with writing...


