‘The purpose of a liberal education is to dare to think your own thoughts’: The Dartmouth’s investigation into provost’s AI use prompts campus-wide conversation - The Dartmouth

In the days since an investigation from The Dartmouth revealed several of Provost Santiago Schnell’s publications since January were detected as entirely “AI-written” by artificial intelligence detector Pangram, faculty members, students and the Dartmouth community have raised concerns about Schnell’s leadership and AI practices at the College.

The Dartmouth chapter of the American Association of University Professors is circulating a petition that calls for the Board of Trustees to remove Schnell from his position as provost, according to a copy of the petition obtained and reviewed by The Dartmouth.

As of Wednesday evening, the petition has more than 220 signatures from faculty, staff, students, alumni and other members of the Dartmouth community. Schnell did not respond to a request for comment by time of publication.

The Sept. 21 investigation by The Dartmouth and the Aug. 26 Semafor investigation that first identified Schnell’s Aug. 9 op-ed in The Washington Post as “100% AI-written” both used Pangram, which was found to have a “near-zero error rate” by an October 2025 University of Chicago audit.

On Sept. 25, College President Sian Leah Beilock announced an independent committee had been created to conduct a “rigorous, objective review” of Schnell’s AI use in his “academic publications and public writing.” The committee will be led by former provost and computer science professor David Kotz ’86. Earlier that day, Schnell apologized for not consistently disclosing his use of AI in a letter to the editor published in The Dartmouth.

“In retrospect, I now realize I should always explicitly disclose use of AI in my public writing and be attentive to the expectations of AI use set by publications,” Schnell wrote. “I am sorry that this was not my standard practice and, moving forward, it will be.”

The Dartmouth Student Government discussed the investigation at their Sept. 27 meeting but decided not to issue a statement until the committee’s investigation is concluded.

In interviews with The Dartmouth, four professors and two students raised concerns about whether the College applies consistent standards to students, faculty members and administrators as norms around generative AI continue to develop.

Biological sciences department chair Magdalena Bezanilla wrote in an email statement to The Dartmouth that she believes any use of generative AI should be disclosed alongside the extent of its involvement in producing a piece of work.

“At this point, I think many people are likely using generative AI to some extent,” Bezanilla wrote. “So if AI is used in any way it should be disclosed and the extent to which it is used should also be disclosed.”

Schnell disclosed his AI use for “language refinement” and “review” of code in the academic article “From Signals to Measurands: A Measurement-Science Roadmap for Reproducible Analytical Biochemistry” and described a collaborative human-AI writing process in “Who Wrote This? Turing, Total Variation and the Mathematics of AI-Text Detection,” The Dartmouth previously reported. “From Signals to Measurands” was 92% AI-written and “Who Wrote This?” was 99% AI-written, according to Pangram tests. Schnell did not disclose AI use in the seven other works he published this year, all of which were reviewed by The Dartmouth.

Economics professor Paul Novosad said there is “certainly not a shared understanding” of “responsible” AI use among Dartmouth faculty and administrators. He added that he believes context matters when determining whether AI use is appropriate.

“If you are using AI to do writing in a context where the audience would have a very strong expectation that they’re hearing your words and your voice, I think that’s not acceptable in the classroom,” Novosad said. “I don’t think it’s acceptable out of the classroom.”

Novosad distinguished between contexts in which readers have a “very strong expectation” that they are hearing someone’s own “words and voice” — such as an opinion piece, personal reflection or student essay — from more functional uses of AI, such as generating portions of a syllabus or multiple-choice options on an exam.

English and creative writing professor Jeff Sharlet said The Dartmouth’s findings raised “serious” questions warranting further investigation into Schnell’s AI use, but added that his concerns did not depend on the accuracy of AI-detection tools.

Instead, Sharlet pointed to Schnell telling The Dartmouth that he used ChatGPT to “refine a few arguments and check for grammatical errors” in his Aug. 9 Washington Post opinion piece that was flagged as “100% AI-written” by Pangram. The Washington Post requires guest writers to confirm that submissions were “not created or manipulated with artificial intelligence or editing software,” a spokesperson for the Washington Post told Semafor.

“We can agree to disagree about AI,” Sharlet said. “We cannot agree to disagree about if I write for a publication, and it has a policy.”

Sharlet said in the days after the investigation was published, he struggled to talk to his students about AI use in his classes.

“I was in the position of going before my students and saying, ‘I don’t really know what to tell you. I’ve asked you to abide by certain things, and leadership apparently doesn’t,’” Sharlet said.

He added that a lack of consequences for Schnell could cause students to struggle to maintain their “integrity” amidst growing AI use the message that “‘I guess it doesn’t matter.’”

“It’s going to be a big signal to, ‘Oh, I guess that’s how it is. If you’re powerful enough, the rules don’t apply to you,” Sharlet said.

Everest Nelson-Holder ’29 said Schnell’s use of AI has made him “skeptical” of the College’s approach to AI, adding that he believes there should be more “clear” and “effective” policies outlining how students and employees are allowed to use AI for work and assignments.

Clear documentation of policies is “something that students want in a lot of their classes, so they know when they can use [AI], when they can’t use it,” Nelson-Holder said. “Professors and other responsible employees in the College should also have similar guidance that they need to abide by to act ethically.”

While the College’s Guidelines on using Generative Artificial Intelligence (GenAI) for Coursework gives professors the “authority” to determine how AI is used in their courses, use of AI tools in submitted work “requires appropriate acknowledgment” from the student. On Sept. 29, interim dean of faculty of Arts and Sciences John Carey emailed additional “guidelines on use of AI detection in teaching” to faculty, which authorized them to use Pangram to detect de-identified student work, according to the email obtained and reviewed by The Dartmouth.

Avery Stern ’27 said she thought Schnell’s use of AI was “hypocritical.”

“If we’re doing an assignment in a class that says you can’t use AI and we use AI, we should get in trouble for that,” Stern said. “The same goes for you when you write for the Washington Post. It’s just like, ‘Come on, guys, we’re doing our job; you do yours.’”

Stern said she believes it would be “more respectable” for Schnell to step down instead of being fired, adding that his keeping his position as provost makes Dartmouth seem “like a joke.”

“This feels like an opportunity for the school to get ahead of further issues by having a clear policy,” she added.

Government professor Russell Muirhead said The Dartmouth’s investigation raises a broader question about the expectations Dartmouth places on members of its intellectual community.

“I think what’s at stake are the very basic standards that we use to define what our community is about,” he said.

Using AI to produce writing without disclosure “displays intellectual laziness” and can become both intellectually and ethically problematic, Muirhead added.

“It’s bad because if you don’t reveal to everyone that that’s what you’re doing, it’s also deceptive,” he said. “I ask my students not to outsource their writing to AI. I think that undermines the point of their education.”

Nelson-Holder said he viewed it as “hypocritical” for College leaders to use AI in ways students might be prohibited from doing.

“If I were to submit something for a class that said you should not use AI and then I were to use AI, I would get at least an academic [honor code violation], if not suspension,” he said.

According to the Committee on Standards, being found guilty of violating the Academic Honor Principle, which may include unauthorized use of AI on assignments, can result in consequences ranging from a failing grade on an assignment to suspension or “permanent separation” from the College.

Nelson-Holder said the episode has made him “a little skeptical” about Dartmouth’s broader approach to artificial intelligence.

The College has pushed for more AI integration into the classroom over the past few years. In a Sept. 5 article in The Atlantic called “Why I Want More AI at Dartmouth,” Beilock wrote that universities which “fail” to teach students on how to use AI “productively” will “cosign themselves into irrelevance.” In December 2025, the College announced a partnership with Anthropic and Amazon Web Services to strengthen “technological integration” at Dartmouth.

“Dartmouth has come out as being pro-AI, and in some ways I think that’s valuable and necessary for the future,” Nelson-Holder said. “ But I also think that if even our most responsible employees of the College are using AI in irresponsible ways, then the direction of Dartmouth College AI policy is definitely something we need to reconsider.”

Bezanilla wrote that AI could “propel and increase the speed” of aspects of research and prove “transformative for education,” though she added that it should not become a substitute for “human-generated writing in scholarly work.”

Muirhead similarly said AI could prove valuable in fields including scientific research and drug discovery, but warned against becoming dependent on technology capable of doing intellectual work for its users.

“The purpose of a liberal education is to dare to think your own thoughts,” Muirhead said, adding that he believes that generative AI “should never be used to compose public statements” from College administrators.

“To be a leader, you need followers that you have to inspire and guide,” he said. “The most powerful forms of persuasion are going to come from authentic human speech.”

As the independent committee conducts its formal review, interviewees also emphasized the importance of how the College responds to the allegations.

Bezanilla said a formal review is “absolutely warranted” and should determine whether the honor code was violated.

Novosad said “due process is good” but the College should move “quickly.”

“It’s really important for everyone’s trust for the committee to move quickly and for clear statements to be made about this soon rather than letting it drag on for months,” he said.

In an email statement, College spokesperson Jana Barnello declined to comment beyond Schnell’s opinion piece in The Dartmouth and Beilock’s email to campus on Friday.

Haley S. Rodriguez ’29 is a reporter with roots in New York City. She is studying history and biology and enjoys long-distance running, reading and sailing.

Jared Gonzalez Arce '29 is a news reporter from California majoring in history. Out of The D, Jared is in surf club and loves music.