Abstract
Technical abstract
This paper presents HallMate, a mobile first roommate matching app designed to help the roommate search process for University of Virginia students. Current approaches such as Instagram pages, Facebook groups, university housing portals, and commercial apps are often unstructured, outdated, difficult to navigate, or limited by paywalls, making roommate search inefficient and stressful. HallMate addresses these problems through structured lifestyle and preference profiles, a swipe based discovery interface, secure in app messaging, and a dual verification identity model that supports both current UVA students and incoming first year students without institutional credentials. The system was built using React Native with Expo and Firebase, with attention to privacy, reliability, and accessibility. Compatibility scoring compares a user’s stated preferences against other users’ actual lifestyle attributes to produce ranked matches. The platform was evaluated through beta testing with six UVA students across onboarding, verification, matching, messaging, privacy, and edge-case workflows. Testing revealed several issues in privacy controls, UI layout, and data persistence, which were subsequently corrected through backend and interface revisions. The results suggest that a structured, mobile first platform can offer a more usable, secure, and efficient roommate matching experience than existing informal methods, while also establishing a foundation for future work in advanced matching logic, broader deployment, and longitudinal evaluation.
STS abstract
This paper analyzes generative artificial intelligence as a sociotechnical system through the Social Construction of Technology framework. Rather than treating AI as a neutral technical tool, the paper argues that its design, deployment, and social effects are shaped by the values and incentives of the institutions that build and use it. The central claim is that generative AI tends to reinforce statistical centrality by privileging outputs that are familiar, legible, low-risk, and institutionally convenient, thereby narrowing creativity, expression, and cultural variation over time. Using SCOT concepts including interpretive flexibility, relevant social groups, and closure, the paper examines three cases: hiring and screening, social media recommendation systems, and generative assistance in writing and thinking. Across these cases, the analysis shows that organizations tend to stabilize AI around priorities such as efficiency, scalability, predictability, monetization, and safety, while users are pressured to adapt their behavior to what these systems reward. The result is not simply bias or factual error, but a broader process in which AI systems shape norms of visibility, merit, communication, and thought. The paper concludes that the most important effects of generative AI are organizational and cultural before they are purely technical, and that changing its trajectory requires not only better models but also institutional changes in where these systems are used and whose values are allowed to govern them.
