Abstract
Both of my thesis projects began with the same observation, though I did not see the
connection at first. UVA students rely on a growing stack of tools that the University does not
really govern. Facebook Marketplace and GroupMe carry most of the informal exchange of
belongings on Grounds, and ChatGPT has become a routine study aid across the curriculum.
These tools work well enough that students keep using them, but because they sit outside any
formal campus structure, the norms around them get made up in real time: what counts as a fair
trade, what counts as an acceptable use, who gets to decide.
The stakes are practical. Research on basic-needs insecurity finds that about one in three
undergraduates deals with financial pressure that affects their academic performance, and
campus move-out generates hundreds of pounds of solid waste per student each year, much of it
functional goods that students abandon because they have nowhere to store or sell them. UVA’s
2023 Generative AI in Teaching and Learning Task Force surveyed over 500 students and found
that 77 percent said their instructors had not made AI policies clear. The two problems look very
different, but they share an underlying issue: student needs are moving faster than the structures
meant to support them.
My technical thesis, written with Christopher Dai, Mark Do, and Su Jin Shin, presents
Agora, a web rental platform built for UVA students. The problem is familiar to most people on
Grounds: an international student buys a winter coat she uses for one semester, a class project
needs a camera nobody wants to buy outright, someone’s roommate has the exact kitchen
appliance you need but there is no way to find that out. Open marketplaces like Facebook
Marketplace handle permanent sales, not short-term rentals, and they put students in contact with
strangers. GroupMe threads bury lending posts within hours. HoosReuse collects donations at
year-end but does not let students earn from items they still own. Agora restricts accounts to
verified @virginia.edu addresses, builds the rental flow around start and end dates, and adds a
trust layer of condition tags, required photos, in-app messaging, and post-rental reviews. We
built it as a web application over four two-week sprints with automated testing throughout, then
evaluated it through three rounds of usability sessions, a survey of UVA students, and semistructured interviews.
The pilot suggests that a campus-scoped rental interface reduces friction
and increases perceived trust relative to the informal alternatives, while surfacing real limits
around offline payment, dispute resolution, and equitable participation.
My STS research paper asks how students and instructors at UVA reinterpret effort and
understanding in courses that permit limited generative AI use. I draw on two STS frameworks,
mutual shaping and the social construction of technology, and work through four kinds of
publicly available evidence: the Provost’s FAQ on AI, the 2023 Task Force’s survey of 504
students and 181 faculty, syllabus statements compiled by the Center for Teaching Excellence,
and Cavalier Daily reporting from 2023 to 2025. My argument is that UVA’s decision to leave
AI policy to individual instructors is not an absence of policy but a deliberate choice to let the
meaning of AI stay unsettled.
Faculty then fill that space differently depending on what their
disciplines treat as the real intellectual work: accuracy in law, independent demonstration in
psychology, accountability under raised expectations in public policy. Students read these
competing rules across their course load and have had far less formal input into how the rules are
set than faculty have. The response most institutions have converged on is to shift assessment
from product to process through drafts, reflections, and oral presentations. That strategy helps,
but it does not resolve the problem, because process artifacts can also be generated by AI. What
counts as legitimate intellectual work stays open.
The two projects do not sit together neatly, and I want to be honest about that. One builds
infrastructure; the other examines why infrastructure is not always the right response. Read side
by side, though, both describe the same situation. Agora assumes that a campus-scoped platform
with institutional verification can organize a practice the University is not going to organize
itself, and the pilot suggests that is roughly right for short-term rentals. The STS paper looks at a
case where UVA has chosen the opposite approach and shows what that choice has cost students.
Working on both at once made clear that building and deferring are each choices, not defaults,
and that universities have not yet built good habits for making that choice deliberately
