Key Takeaways
- Aria‑Vue L. Daugherty ’29 sustained a concussion after being struck by a black SUV while riding a rented city bike, which led her to rely on Harvard’s Disability Access Office (DAO) Peer Notetaker service.
- The DAO program pays eligible notetakers a $600 stipend per course and currently serves about 50 accommodated undergraduates and GSAS students each year, with 84 notetakers hired for 2025‑26.
- Students value the human element of peer notes—such as contextual framing, handling of graphics, and the inadvertent “thought process” left by the notetaker—over AI‑generated transcripts, which they worry may hallucinate or miss nuance.
- Some students experiment with AI tools (Otter.ai, custom apps) and see potential for cheaper, scalable note‑taking, but remain skeptical about quality, privacy, and the loss of a student‑income source.
- DAO senior director Kate Upatham stresses that AI transcription and human peer notetaking are complementary, not interchangeable, accommodations.
Incident and Immediate Aftermath
On March 20, Aria‑Vue L. Daugherty ’29 was “peddling on a rented city bike when a black SUV slammed into her.” She had just visited a friend at Georgetown and was preparing for her next Harvard College Debating Union tournament in Washington, D.C. The collision left her with a severe concussion, prompting a seven‑and‑a‑half‑hour stay in the emergency department at George Washington University Hospital. Daugherty recalled feeling disoriented afterward: “I really wasn’t gripping anything because I was recovering from the concussion,” she said, adding that her vision began to blur during the first round of debate, which forced her to seek medical attention. The episode underscored how quickly an everyday activity can precipitate a lasting academic disruption.
Recovery and Reliance on Peer Notetaker
Returning to campus the following Monday, Daugherty leaned heavily on Harvard’s Peer Notetaker service, administered by the Disability Access Office (DAO). Because she already held medical accommodations through the DAO, the service was readily accessible, but her concussion made it “even more necessary.” She explained that reviewing another student’s notes was “way more helpful than looking at my own notes” when her cognition was still impaired. The peer‑notetaker system thus became a critical bridge, allowing her to keep pace with coursework while she recuperated from the injury.
Structure and Scope of DAO Peer Notetaker Service
The DAO’s Peer Notetaker program provides a $600 stipend per eligible course to students who share lecture notes with peers who have approved accommodations. Roughly 50 undergraduates and Graduate School of Arts and Sciences (GSAS) students qualify for the accommodation each academic year, and for the 2025‑26 cycle the office hired 84 notetakers. This scale demonstrates the program’s capacity to support a sizable portion of Harvard’s disabled student population while also offering a modest income source for student workers. The stipend model incentivizes participation without creating a financial barrier for those seeking notes.
Human vs AI Note‑taking: Student Perspectives
Despite the proliferation of AI tools such as Claude, NotebookLM, and Otter.ai, many students feel that the DAO’s human‑driven notes retain an irreplaceable quality. Daugherty described herself as “kind of an extremist — I’m really scared of hallucinations,” saying she would “trust someone’s handwritten notes a lot more.” AI hallucinations—where large language models generate false or nonsensical information—have been documented; one test found that 79 percent of outputs from newer AI systems were hallucinations. This distrust pushes students toward the reliability they associate with a fellow classmate’s interpretation of lecture material.
Limitations of AI in Capturing Non‑textual Content
Eman A. Seyal ’26, who uses the Peer Notetaker service because of rheumatoid arthritis, highlighted a concrete shortcoming of AI: it often produces a raw transcript that “might simply print out a transcript of text which you have to parse though.” In contrast, “peer notes capture not only the content, but also the way the instructor frames key concepts, which is critical.” Seyal also pointed out that AI struggles with non‑textual information such as images, graphics, equations, and plots—elements that frequently appear in STEM and humanities courses and are essential for full comprehension.
Personal Connections and the Value of Human Thought Process
Michael Isayan ’29, a notetaker for Expos 20, initially did not know whether his efforts helped any specific student. After Daugherty approached him following class, she revealed that his notes had been supporting her recovery. Isayan reflected that while his handwriting might be “crappy,” the true worth of his contribution lay in the “inherently human thought process” behind it. He mused, “Even if AI produces the same work that you as a human being could produce, there is something lost in the sense that you did not think it through,” suggesting that a marginal comment or ambiguous phrase could prompt a reader to engage more deeply with the material—a nuance an AI summary might flatten.
AI Enthusiasm and Experimental Alternatives
Not all students dismiss AI’s potential. Lia T. Zheng ’27, president of the Harvard Undergraduate Society for Artificial Intelligence, served as a peer notetaker in Stat 110 and mused, “It’s just a gig, right? I don’t think there was anything necessarily special about me being human. I think that AI could likely be just as good as, if not better, at peer notetaking.” Ege Cakar ’27, who previously built a note‑taking app for YouTube lectures, argued that modern AI tools now “plug in” and “are probably going to give you a result that’s significantly better than the median notetaker.” He advocated for DAO adoption of AI to cut costs and expand access, while Dashiell A. Bhattacharyya ’27 proposed that AI could be tuned to individual note‑preferences if the system were given the right infrastructural support.
Institutional Stance: Complementarity of Services
Kate Upatham, Harvard’s senior director of disability resources, emphasized in an email that both the AI transcription tool currently offered by the DAO and the human Peer Notetaker service “are valuable in different ways, and they shouldn’t necessarily be seen as exchangeable.” This viewpoint acknowledges that AI can efficiently convert slides or recordings into text, yet it does not replicate the interpretive, contextual, and interpersonal dimensions that human notetakers provide. The office appears poised to maintain both offerings, allowing students to choose—or combine—based on their specific needs, course format, and comfort with technology.
Conclusion: Balancing Technology and Human Support
Aria‑Vue Daugherty’s accident and subsequent reliance on peer notes illuminate a broader tension at Harvard between embracing cutting‑edge AI solutions and preserving the indispensable human elements of academic accommodation. While AI promises speed, scalability, and lower costs, students repeatedly cite concerns about hallucinations, the loss of non‑textual cues, and the disappearance of a subtle, thought‑provoking “human touch” in notes. Meanwhile, the DAO’s stipend‑based peer notetaker program continues to furnish essential support, income for student workers, and a personalized layer of comprehension that AI has yet to fully emulate. As the university experiments with AI‑driven transcription, the prevailing sentiment suggests that the most effective accommodation strategy will likely involve a hybrid approach—leveraging AI for routine text conversion while retaining human notetakers for the nuanced, interpretive, and relational aspects of learning.
https://www.thecrimson.com/article/2026/4/25/artificial-intelligence-and-peer-notetaking/