Abstract
Large online platforms shape how people find and consume information, exchange services, and build trust. Yet these platforms are optimized for the highest return on investment, and as a result, they are often unavailable in smaller communities or fail to align with the users’ interests and environments. This mismatch is the general problem underlying both of my thesis projects. My technical project, HoosHelping, addresses this gap by building a community-specific task marketplace designed around the social context of a university town, Charlottesville. My STS research examines the other side of the same problem: how Google’s integration of artificial intelligence (AI) into search consolidates power within a dominant platform, extracting value from online businesses and content creators while reducing the diversity of information. These two projects investigate how large platforms fail the communities they claim they serve and what alternatives might look like. HoosHelping is a web platform designed to connect University of Virginia students and Charlottesville residents for local tasks such as errands and furniture assembly. The platform supports a complete task-based workflow in which users create tasks and request help or bid on tasks. The platform interface provides several features to increase trust between university-affiliated and non-university users such as automated university verification, one-click third-party background checks, and digital payment handling. We conducted a structured beta test with university students to evaluate the platform’s core functionality and workflows. The results showed that most of the platform’s workflows functioned correctly, and testers reported positive feedback regarding the concept and overall usefulness of the platform to the UVA community. The presence of verification badges on profiles, user reviews, and tasks filtering compared to informal alternatives suggests that the system’s design goals were directionally correct even in its prototype state. My STS research investigates the sociotechnical implications of AI-driven search, focusing on how artificial intelligence consolidates power within dominant platforms and changes the flow of information on the internet. Utilizing the framework of platform capitalism, I examine how Google leverages its control over digital infrastructure and data to strengthen its position and expand its control through features like AI Overviews and AI Mode. The evidence is clear: 20 to 50 percent of traffic is at risk from traditional search, and Search Engine Land reported that Google AI Overviews resulted in a 60 percent drop in click-through rates. Google is no longer an intermediary that indexes, organizes, and facilitates users’ access to information and content across the globe. With the release of AI Overviews and AI Mode, Google became the primary producer of answers and information, actively extracting value out of content created by publishers to enrich itself and keep users within its ecosystem, thus reshaping the relationship between users, platforms, and online businesses. Furthermore, generative AI remains unreliable, and the lack of transparency in how these systems generate responses and cite sources poses a risk that extends far beyond the decline of online traffic, threatening other ecosystems such as monocultures in information and research. Both projects contribute to the broader problem of platform dominance through different angles. HoosHelping demonstrates that community and context-driven design could increase trust and usefulness, but the short and limited nature of the evaluation period made it difficult to capture any long-term usage patterns. My STS research documents the harmful impact of AI-driven search but cannot yet capture the full long-term consequences of this shift. Given that both projects suffered from having a limited duration, future researchers should not only extend the timeframe of these studies but also investigate how community-driven platforms can scale beyond a small group of users.
