Harvard College Adds Required AI Module to Expository Writing Curriculum

All Harvard College freshmen enrolled in expository writing courses were required to complete a three-part training module on artificial intelligence this semester, the first time the training has been mandated across the program.

The module — which covers how large language models work, their impact on education, and broader questions about AI and society — was rolled out across all Expos sections this spring, after being piloted in select courses beginning in fall 2024. More than 90 percent of Harvard College students take an Expos course in their first year to fulfill the College’s writing requirement.

Associate Dean of Undergraduate Education Gillian B. Pierce ’88 said in a Tuesday interview that the module was designed to give every student a shared baseline on generative AI — and to relieve faculty of having to teach the basics themselves.

“Faculty may have some sense that, ‘Oh, I need to teach students what is an LLM,’” Pierce said. “You can kind of assume now that students will come knowing that.”

The training was developed by the Harvard College Writing Program, which houses Expos. Writing Center director Jane Rosenzweig and preceptor Tad Davies, who led the module’s design, said in an interview Wednesday that the goal was not to take a position on AI but to equip students to enter ongoing debates themselves.

“There’s no agenda for what students should take away from this, except what we all agree on, which is that when you have information, you can make better judgments about what you want to do with the technology,” Rosenzweig said.

The first section walks students through the mechanics of large language models, while the second examines AI’s effects on education, including published research from Harvard faculty on custom chatbots in physics courses and the AI course assistant for Computer Science 50: “Introduction to Computer Science.” The third takes up questions of environmental impact, data privacy, copyright, and disinformation.

Rosenzweig said the module close-reads the terms of service for popular chatbots in the privacy section — an exercise she said her own students have found illuminating.

“This is a work in progress,” Rosenzweig said. “As society’s conversation about AI changes, expect the module to keep up with that as much as possible.”

The training was also required this semester in Expos 40, a public speaking elective taken primarily by juniors and seniors.

A small number of freshmen fulfill the writing requirement by taking Humanities 10 — a yearlong course capped at 90 students — instead of Expos. Classics professor David F. Elmer ’98, who serves as course head for Hum 10, wrote in an email that the course does not include the Expos module but offered its own training on AI.

Davies said the module was intended as a starting point rather than a comprehensive treatment, and that disciplinary specifics would need to be addressed in later coursework.

“There’s going to be disciplinary differences in the way that the models can engage in certain kinds of materials very effectively and less effectively in others,” Davies said. “So students are going to need to engage their classes and their professors in this ongoing conversation.”

The module was designed by Expos leadership, with revisions each semester to keep pace with the technology. Rosenzweig said the team re-tests the lessons every term, since outputs from image generators and chatbots used as teaching examples can change or disappear as the underlying models are updated.

Rosenzweig confirmed that the module has been shared with Dean of the College David J. Deming, Dean of Undergraduate Education Amanda Claybaugh, and the FAS Faculty Advisory Committee on Artificial Intelligence.

The training arrives as College administrators turn increasing attention to how Harvard should adapt its curriculum to generative AI. Deming made AI one of his top priorities for the academic year, and Claybaugh hosted a Generative AI summit in January to solicit student input on the College’s AI policy.

In a guest lecture in Economics 50 earlier this month, Deming urged students to learn how to use AI — while warning against letting it displace the core work of learning.

Pierce said she hoped the Expos module would mark a first step toward a more coordinated, University-wide approach.

“This is a collective project for the University,” Pierce said. “Our motto is Veritas. So how do we also educate students about what it means to be human, and what are our responsibilities there?”

“Those are big questions, and it’s going to have to involve everyone,” she added.

—Staff writer Sebastian B. Connolly can be reached at [email protected] or on Signal at sbc.23. Follow him on X @SebastianC4784.

—Staff writer Summer E. Rose can be reached at [email protected] or on Signal at ser.85. Follow her on X @summerellenrose.