by Kate Cammack, CfT Co-Director
Many of us use this late-May window to reflect and think ahead to how we might adjust our fall courses, before mentally transitioning to summer activities. One thing that will change all of our fall syllabi is the presence of our newly approved university AI policy (in its working form). With that comes an invitation for us to revisit and update our own course AI policies for our fall syllabi.
A Berkeley researcher analyzed 3000+ course syllabi and found that instructors have seemed to “warm” to AI over the past 3-5 years. Here’s Beth McMurtrie’s summary.
A brief tour through his figures:
Course AI policies are becoming more common in syllabi, though this varies across discipline. Business, humanities, and social science courses are more likely to include AI policies in their syllabi; physical sciences are less likely. Notably, across disciplines in this study, ~40% of syllabi still didn’t contain AI policies.
The permissiveness of course AI policies varies across disciplines. In this study, humanities, life sciences & social sciences were least permissive; business, arts, and math/CS were more permissive.
The framing of course AI policies has evolved relatively quickly. Over time, fewer syllabus policies mention academic integrity and more policies mention AI’s impact on learning.
Permitted uses of AI depends heavily on the task/purpose. Nearly 80% of instructors restricted AI use for drafting/revising work, but a similar percentage permitted AI for editing/proofreading. Interestingly, instructors were almost evenly split about whether they permitted or restricted AI use in ideation/planning work.
More instructors are exploring AI-based tasks. Overall, 11% of syllabi mentioned at least one AI-based task, such as an output verification activity or AI-integrated assignment. Again, this ranged across disciplines.
One caveat is that these syllabi were pulled from a large university in Texas. Before you say, “But big universities are entirely different from Sewanee!”, I think the data remain worth considering in the context of broader, overall patterns that point to disciplinary differences, student assumptions, faculty understanding of AI, and other “big-picture” issues that are worthy of consideration. If you’re interested in more context, the methods begin on page 12.
A few take-home considerations:
- Adopt a course AI policy, if you don’t already have one.
- In your coursework, consider how/where AI might augment students’ learning experience and where it poses risks to their learning and skill-building. Then, set clear rationale and guidelines for AI use on those specific tasks.
- The disciplinary issues that you face in your teaching are different from those faced by your colleagues in different departments. Remember that we’re all approaching these issues with different perspectives and seeing different concerns and opportunities.
- Our pedagogical choices can impact labor markets. As stated on page 29:
- “The task-based framework highlights a potential feedback loop between AI in education and AI in production. If AI displaces skill-building tasks in educational settings, students may graduate with weaker capabilities in precisely the areas where AI is most capable. This erosion of human skill would further shift comparative advantage toward AI, potentially accelerating automation in those tasks.”
- “Consider writing as an example. If students practice writing less because AI can draft text, they may enter the labor market with weaker writing skills. Weaker human writing skills make AI-generated writing relatively more attractive to employers, reinforcing the displacement of human labor in writing tasks. The same logic applies to coding and other domains where AI capabilities are strong.”
- “This feedback loop suggests that decisions made in classrooms today may shape the trajectory of automation tomorrow. Instructor responses to AI are not merely pedagogical choices but upstream determinants of skill supply that feed into labor market dynamics.”
The full paper is worth a read for more details, nuance, and thoughtful consideration of these data.
Full citation: Chirikov, I (2026). How Instructors Regulate AI in College: Evidence from 31,000 Course Syllabi. CSHE Higher Education Working Paper Series, Vol. 26-1. Retrieved from: escholarship.org/uc/item/9c51s3gs
AI was not used for any aspect of this blog post.