Abstract
This thesis addresses how system design decisions shape equity and access in modern technological systems. As digital platforms and AI-driven tools become more embedded in everyday decision-making, they increasingly influence how people access resources, opportunities, and outcomes. These systems are often presented as efficient and objective, but they reflect the assumptions, data, and priorities built into their design. As a result, technological systems can either reinforce existing inequalities or help reduce them, depending on how they are developed. While my STS research examines how automated systems can unintentionally reproduce inequality through embedded assumptions and biased performance, my technical project shows how thoughtful system design can instead improve access, reduce waste, and support more equitable resource distribution.
My technical project focuses on the development of Agora, a web-based peer-to-peer rental platform designed specifically for University of Virginia students. The goal of this project is to address the inefficiency of short-term ownership on college campuses, where students often purchase items for temporary use and discard them shortly after. Agora allows students to list items they are willing to lend and enables others to browse, request, and coordinate rentals within a campus-restricted environment. The platform includes core features such as user authentication through university email, item listings with descriptions and photos, search and filtering capabilities, a rental request workflow, messaging between users, and a review system to support trust and accountability. By focusing on short-term access rather than ownership, this project aims to reduce unnecessary consumption, lower costs for students, and improve the circulation of underutilized goods within the campus community. This reflects how intentional design choices can be used to expand access to resources and create more equitable systems within a specific community.
My STS research paper examines how accent variation becomes a structural disadvantage in AI-assisted interviewing systems. I argue that this disadvantage emerges through the interaction of technical design choices and existing social hierarchies, rather than from isolated transcription errors alone. Using research on automatic speech recognition (ASR), linguistic bias, and algorithmic fairness, the paper shows that these systems often perform unevenly across accents and embed assumptions about standardized English as the norm for professional communication. Because automated interviewing platforms rely on transcribed speech to evaluate candidates and operate at scale as early-stage filters, these disparities can systematically shape hiring outcomes. The analysis also highlights the limits of purely technical solutions, showing that improving model accuracy alone cannot fully address the broader institutional norms and social assumptions that define what counts as clear or professional speech. This demonstrates how system design can reinforce inequality when underlying assumptions and social contexts are not critically examined.
Developing both projects together highlights how technical design and social context are deeply interconnected. My STS research emphasizes that technologies are not neutral systems, but sociotechnical systems shaped by human assumptions, institutional norms, and power structures. This perspective informed my approach to designing Agora, where considerations of access, trust, and fairness were incorporated into core features rather than treated as secondary concerns. At the same time, working on a technical system reinforced the idea that even small design decisions can directly shape user experiences and outcomes. Together, these projects show that addressing real-world problems requires not only functional technical solutions, but also critical attention to the social impacts of those systems and the ways they structure access and opportunity.
