BERKELEY, CA — As generative artificial intelligence (AI) reshapes the classroom, a new study from the Center for Studies in Higher Education (CSHE) provides novel large-scale longitudinal evidence of how faculty are responding.
The working paper, “How Instructors Regulate AI in College: Evidence from 31,000 Course Syllabi,” authored by Igor Chirikov, Senior Researcher and the SERU Consortium Director, tracks the evolution of AI policies at a major public research university from 2021 to 2025. Using computational methods to analyze tens of thousands of course syllabi, the research identifies a significant shift: instructors are moving away from blanket bans toward “task-based” approaches. Instructors are increasingly differentiating where new technologies can support learning and where they might undermine skill formation.
Key Findings and Implications:
- Faculty are warming toward AI: Initial policies were predominantly restrictive, but instructors are shifting toward more permissive approaches over time. They increasingly differentiate policies by specific tasks, restricting AI for some activities while permitting it for others, and are beginning to introduce new AI-based assignments.
- Stark disciplinary differences: Business courses have moved most rapidly toward permissive policies and show the highest adoption of new AI-based tasks at 27% of courses. Humanities courses remain the most restrictive, reflecting different exposure of their learning tasks to AI capabilities.
- Course task composition predicts regulation: Courses requiring more tasks where AI capabilities are strong, such as writing and coding, are significantly more likely to have AI policies. It suggests that instructors respond to AI pressures not in a uniform way but depending on when AI capabilities overlap with required student practice.
- Implications for policy and labor markets: The study suggests that effective institutional AI policies require flexibility to accommodate disciplinary and task-level differences. It also highlights a potential feedback loop between education and labor markets: if AI displaces skill-building tasks, students may graduate with weaker skills in precisely the areas where AI is strongest, potentially further accelerating automation.
“The data shows a notable evolution in faculty response to AI,” says Chirikov. “Instructors are shifting from broad restrictions toward distinguishing which specific activities benefit from AI and which require independent work. The emergence of new AI-based assignments, where students learn to work with these tools critically, indicates that faculty are finding ways to incorporate AI as part of student learning, not just something to police.”
The study documents an important transition period in higher education in response to AI and offers a framework for understanding the impact of this technology on skill formation. How instructors continue to adapt their teaching in response to AI will shape what students learn in college and how prepared they are for AI-augmented workplaces.
The full working paper is available at: https://escholarship.org/uc/item/9c51s3gs