The Faculty Senate debated the role of generative artificial intelligence (AI) in education, attempting to strike a balance between encouraging independent critical thinking and acknowledging the utility of AI during its Thursday meeting.
Vice Provost for Undergraduate Education Jay Hamilton presented to the Senate on the ways students currently use AI and how its use has impacted learning.
Hamilton noted that students mainly use AI to understand difficult concepts. The result of this use is that office hour attendance and in-class exam scores have dropped while take-home problem set scores have skyrocketed, according to Hamilton. He referenced The Daily’s Editorial Board article, “Stanford students would rather not think,” as evidence of the issue.
Hamilton presented several solutions devised by students enrolled in CS 283: Governing Artificial Intelligence: Law, Policy, and Institutions, a class that explores AI policy regulations. He read four papers from the class that proposed explicit AI policies in syllabi. Hamilton also suggested that syllabi adopt a “barbell” model, either embracing or prohibiting AI use.
Hamilton expressed the risks of policies that call for AI use in between these extremes. “Sort-of use just goes to full-on use,” he said.
Hamilton asserted that an ideal AI policy would teach students both “augmented writing and coding and unaided writing and coding,” as well as help students achieve the goals of a liberal arts education, including critical thinking skills and networking.
Faculty expressed uncertainty about the efficacy of more specific AI policies in syllabi. Finance professor Jonathan Berk noted that Hamilton neglected to discuss the concern of “externality,” the idea that if students feel as though their classmates are using AI, they will also use AI to “keep up.”
Hamilton and other faculty agreed that “externality” plays a significant role in the use of AI in educational settings.
Comparative literature professor David Palumbo-Liu took a critical stance on the use of AI in the classroom.
“Stanford should understand that the use of this [AI] creates really bad mental health issues,” Palumbo-Liu said.
To mitigate the use of AI in the humanities, Palumbo-Liu suggested that Stanford expand offerings of seminars over lectures or even introduce “tutorials, like the Oxford model.” This smaller setting would allow professors to change the structure of their assignments to be less conducive to AI use, Palumbo-Liu said. He pointed to the final exam of the comparative literature class he teaches, in which students meet individually with him and discuss their writing choices.
Several faculty members supported this proposal and debated how to expand seminar and tutorial offerings. Suggestions included implementing caps on majors such as computer science and hiring more faculty.
The Senate also discussed how to renew public support for higher education. SLAC vice president Kam Moler and political science professor Brandice Canes-Wrone presented on the topic, sharing their plans to conduct internal surveys and research on the public’s declining confidence in universities. Moler and Canes-Wrone also argued in favor of faculty outreach and messaging to rectify this decline.