More than 60 Percent of Harvard Faculty Say AI Harms Their Courses

Almost four years after the public release of ChatGPT, nearly two-thirds of professors who responded to The Crimson’s annual Faculty of Arts and Sciences survey said they believe the prevalence of artificial intelligence has had a negative effect on their courses.

Sixty-four percent of faculty respondents said AI had a “somewhat negative” or “very negative” effect on their courses this year — a sharp uptick from last year’s survey, when 42 percent said the same. The figures point to growing concerns about tools that allow students to complete assignments without fully absorbing content, even as professors adjust their courses to keep up with the technology.

Course-level policies are now commonplace: Only 3.5 percent of respondents said they have no explicit policy on AI tools in their courses, down from 10 percent last year. The percentage of respondents who said they “entirely permit” the use of AI tools in their courses was also roughly halved, from 8 percent to 4 percent.

The faculty concerns come as Harvard coordinates broader curricular responses to AI. Harvard College added a required AI module to the first-year Expository Writing course this spring and phased in an AI chatbot to advise incoming students on course planning. The FAS also announced plans this spring to offer access to Anthropic’s Claude, while ending its institution-wide ChatGPT Edu pilot.

The Crimson distributed the survey to more than 1,400 faculty members across the FAS’ three academic divisions and the School of Engineering and Applied Sciences from April 29 to May 22. Respondents were tenure- and non-tenure-track faculty pulled from the school’s public masthead. The Crimson received 463 responses to the emailed survey, 292 complete and 171 partially complete.

In the months since the survey, Harvard leaders have ramped up their messaging on AI. Harvard President Alan M. Garber ’76 lauded AI tools’ potential to advance research across disciplines at a Memorial Church address earlier this month. College Dean David J. Deming named AI policies as a priority and suggested that faculty integrate AI into writing- and project-heavy classes.

FAS spokesperson James M. Chisholm said in a statement that Hoekstra “recognizes the importance of engagement on the topic of AI,” adding that she appointed Physics professor Christopher Stubbs as her senior adviser on AI.

This is the fifth in a series of articles on the survey. Previous iterations focused on faculty governance and budget cuts, viewpoint diversity efforts, campus antisemitism, and national politics.

Shifting AI Policies

Although the FAS released guidelines on AI use in 2023, it has left broad latitude for professors to create their own rules. The Office of Undergraduate Education’s current guidance requires all faculty to inform students of their policies governing generative AI use and post those policies on their Canvas sites, while leaving the substance of the policies largely to individual instruction.

A Crimson analysis of the College’s most popular courses last year found that all but two of the 20 courses — none of which had policies before fall 2022 — had adopted policies regulating AI use by fall 2025. Since then, the OUE’s website has moved from encouraging to requiring faculty to delineate how AI will be used in their classes.

Those guidelines have ranged from explicit bans to gently curbed use. 34 percent of respondents this year said they permit the use of AI tools in their courses with “strong restrictions” and 35 percent with “some restrictions” — up from last year, when 29 percent allowed use with “strong restrictions” and about a third with “some restrictions.”

More faculty also entirely prohibited AI use — 24 percent, compared with 20 percent last year —while 4 percent allowed unrestricted AI use, down from 8 percent last year.

AI policies have varied greatly between divisions. In a random sample of 600 classes for fall 2026, more than 50 percent of science and engineering classes allowed AI use, but only 35 percent of Social Sciences classes did and 27 percent of Arts and Humanities classes.

The survey results appear inconsistent with The Crimson’s analysis: 79 percent of science and engineering faculty, 84 percent of social science faculty, and 63 percent of arts and humanities faculty said they permit some AI use in their courses.

Faculty are also reported widespread encounters with AI-generated work. Nearly 90 percent of respondents said they had received student coursework that they knew or believe was produced using AI. (Sixty-four percent said they felt somewhat or very confident in their ability to differentiate student-created work from AI-produced assignments.)

Of faculty who said they “frequently” received work they knew or believed to be completed with AI, 48 percent said the technology had a very negative effect on their courses, while only 23 percent said they entirely prohibit AI use for their students.

The College classifies submissions of AI-generated work without attribution as honor code violations, but formal referrals have remained rare. Last year, roughly 10 percent of respondents said they had referred students to the Honor Council or Administrative Board for unauthorized AI use; that figure climbed slightly to 12 percent this year.

Formal referrals were more common among professors who said they “frequently” encountered AI-produced work. Twenty-five percent of faculty “frequently” receiving AI-produced work referred students to the Honor Council, while only about five percent of faculty who “occasionally” received AI-produced work made the same call.

Faculty hesitate to bring cases to the Honor Council partly because AI use is “virtually impossible to prove,” Mary D. Lewis, the History department’s Director of Undergraduate Studies, told The Crimson earlier this year.

Faculty have also developed creative ways to enforce AI policies. On a take-home final for Government 1759: “Behavior Insights and Public Policy: Nudging for Good,” course staff embedded hidden instructions to catch students who’d copied the question into an AI tool.

Some have adapted assignments entirely, shifting to in-person blue-book exams or requiring students to turn in Google docs with version histories to make AI use easier to identify.

While specific policies may vary from class to class, the College has rolled out some broader changes for students, including AI chatbot advisors and an AI module for first-year writing.

Those changes were spearheaded by College leaders who have been especially eager to tackle the issue. College Dean David J. Deming has remained optimistic about AI use, encouraging students to understand how to use AI without outsourcing all skill-building and floating “AI acceptance, or even encouragement” in writing-based courses.

Deming told students in a Wednesday email that he hoped to roll out a “mission-aligned AI use policy at the College by the end of the academic year.”

Academic Rigor

Faculty frustration with AI on this year’s survey was paired with skepticism of students’ commitment to their coursework.

Roughly 70 percent of respondents “somewhat” or “strongly” agreed that Harvard students do not sufficiently prioritize their coursework, in line both with last year’s results and a 2025 report that found students spent more effort chasing extracurriculars than engaging with their classes.

That report also recommended discouraging phone use in class. Faculty expressed broad support for keeping students offline during class on this year’s survey — 86 percent “somewhat” or “strongly” supported limiting students’ use of electronic devices and the internet during class, up from 74 percent on last year’s survey.

Those efforts are taking shape outside the classroom as well. Deming announced at Convocation this year that the College will provide students with phone pouches to encourage in-person, tech-free interactions.

Concerns about academic rigor crescendoed last fall, when the Office of Undergraduate Education released a report showing that more than 60 percent of grades awarded to Harvard undergraduates are A’s and pitching reforms to the grading system. (86 percent of this year’s respondents “strongly” or “somewhat” agreed that grade inflation is a problem at Harvard.)

Faculty settled on a potential solution this spring, when they voted to adopt a 20 percent cap on A grades in the College.

Though that cap will formally take effect next year, many classes will impose the restrictions beginning this fall. In a June email to faculty, Dean of Undergraduate Education Amanda Claybaugh wrote that imposing the cap early would allow them to avoid “arbitrary” differentiation in grading.

“Things like that may well happen if our grading remains as compressed as it currently tends to be. To guard against arbitrary distinctions, we’ll need to change the way we grade,” she wrote.

—Staff writer Abigail S. Gerstein can be reached at [email protected] and on Signal at abbysg.97. Follow her on X @abbysgerstein.

—Staff writer Amann S. Mahajan can be reached at [email protected] and on Signal at amannsm.38. Follow her on X @amannmahajan.