Harvard requires AI literacy for freshmen

Writing class is where Harvard decided to make AI unavoidable. Not by telling students to use chatbots, and not by banning them harder, but by making every freshman in Expos complete an AI literacy module first. That is a bigger move than it sounds. Expos is the one academic experience nearly every Harvard student shares, so putting AI there turns the topic from elective curiosity into basic campus literacy. (thecrimson.com) ### Why does Expos matter so much? Expos is Harvard College’s required first-year expository writing course — basically the default writing gateway for the undergraduate experience. The writing program says a course meeting this requirement has been universal at Harvard since 1872, and more than 90 percent of students take Expos in their first year. So if you want to give almost an entire class year the same baseline on a new tool, this is the obvious place to do it. (writingprogram.fas.harvard.edu) ### What exactly changed? This spring, all Harvard freshmen enrolled in expository writing courses were required to complete a three-part AI training module. That is the new part. The program had tested the material in select courses starting in fall 2024, but this semester was the first time it became mandatory across the full Expos program. The training was developed inside the Harvard College Writing Program rather than dropped in from some outside ed-tech vendor. (thecrimson.com) ### What’s in the module? The first section explains how large language models work. The second looks at AI in education, including examples from Harvard teaching projects like custom physics chatbots and the CS50 AI course assistant. The third moves into the messier stuff — environmental costs, data privacy, copyright, and disinformation. That mix tells you the goal is not “here’s how t(thecrimson.com)e it can go wrong.” (thecrimson.com) ### Is Harvard telling students to use AI? Not exactly. The people who built the module framed it as a shared baseline, not a pro-AI campaign. The idea was to give students enough understanding to make their own judgments and to save individual instructors from having to spend class time explaining what an LLM even is. That is a subtle but important distinction — Harvard is normalizing AI knowledge, not prescribing one correct attitude toward the tools. (thecrimson.com) ### Why put this in writing class? Because writing is where the collision is most obvious. Generative AI can summarize, outline, mimic style, and produce passable prose on demand — which means it sits right on top of the skills Expos is supposed to teach. If students are going to argue, revise, cite, and think carefully about language, they also need to understand what happens when a mac(thecrimson.com)uestions show up fastest. This last point is an inference from how Expos is structured and what the module covers. (writingprogram.fas.harvard.edu) ### Why does the privacy piece matter? Because students often treat chatbots like harmless study aids when they are really data systems with terms of service, retention rules, and opaque downstream uses. One part of Harvard’s module has students read those terms closely. That is smart. The real literacy test is not just knowing that ChatGPT exists — it is knowing what you hand over when you paste in notes, drafts, or personal information. (thecrimson.com) ### Is this a one-off or a signal? It looks like a signal. Harvard is using a required, first-year course to establish AI fluency as part of general academic formation. That is different from offering an AI elective or a one-time workshop. It says the institution thinks students now need a common vocabulary for these systems the same way they need a common vocabulary for sources, arguments, and evidence. (thecrimson.com) ### Bottom line? Harvard’s move is less about teaching freshmen to love AI than teaching them not to be naive about it. And by putting that lesson inside required writing instruction, Harvard is making a broader claim — AI literacy now belongs with the basics.