Dartmouth President Xian-Lee Beilock is calling for greater use of artificial intelligence while emphasizing that university leaders must be transparent about how they use it. In a statement announcing Provost Mr. Schnell’s resignation, Beilock said leaders should model the “transparency and human judgment” expected from students and scholars, suggesting that disclosure is central to Dartmouth’s approach to AI.
Beilock has argued for incorporating more AI into academic work. In an Atlantic essay, she wrote that school leaders must be “held to the highest standards.”
In the announcement of Mr. Schnell’s resignation, Beilock wrote, “Each of us has different perspectives on when and how to incorporate AI into our work and lives. Moreover, keeping Dartmouth at the forefront of AI necessarily means continuing to lead in a fluid and rapidly changing environment. As leaders, it is critical that we model the transparency and human judgment we expect from our students and scholars.”
What does AI disclosure mean?
In a September 2026 paper about the AI-detection tool “Who Wrote This?”, Schnell argues that describing a work as either “written by a human” or “written by an AI” creates an overly simplistic binary. He suggests that future writing will often combine human judgment with AI assistance, and that such collaboration is not inherently problematic.
For Schnell, the essential issue is whether the human author accepts full responsibility for the final text, regardless of how it was produced, structured, or edited. In his paper, he describes his own process:
In September 2026, I asked OpenAI’s GPT-5.6 Pro whether it was possible to reliably identify essays written by an AI, and proposed connecting this problem to the Turing imitation game and the “reverse” Turing test triggered by AI text detection. GPT-5.6 Pro connected the question to binary hypothesis testing and statistical decision theory, suggested an initial structure and title, identified candidate literature, and showed generated portions of an early draft. I chose the problem, the diary, the audience, and the ultimate direction of the discussion. I decided which examples and results to keep front and center, including the decision to retain Liang’s work in the face of significant counter-evidence. I checked citations, rejected or revised claims I found inaccurate, corrected mathematical and conceptual errors, and restructured arguments.
In revising the manuscript, we rewrote most sections and added significant clarification. I occasionally asked GPT-5.6 Sol for language editing assistance with rewritten sections. I created the diagram and made the final decisions regarding mathematics, interpretation, sources, organization, and presentation. I take full responsibility for the completed manuscript.
Before submitting the manuscript for publication, I also asked GPT-5.6 Pro and Anthropic Claude Fable 5.1 to critically review it and identify any issues that the reviewers might raise. I evaluated the issues they identified and made fixes where warranted.
Despite the extensive AI assistance described in the paper, Schnell rejects the idea that an AI system should be listed as a co-author. He writes that “[co-authored] can cause another confusion. Mathematical tools have long made substantive contributions without authorship… The point is not that software has not contributed anything. That is, there is a difference between contributing and being an author.”
Source: arstechnica.com