Abstract
Introduction
As artificial intelligence becomes more integrated into everyday life, people increasingly
rely on AI systems to access information, make decisions, and interpret complex data. However,
large language models often lack transparency, making it difficult to understand how responses
are generated or to verify their accuracy. In complex environments such as universities, where
large networks of people, courses, and research outputs are interconnected, finding relevant
information efficiently can also be challenging. My technical project and STS research both
address issues of trust, transparency, and accessibility in AI systems. My technical project
developed a University of Virginia Graph Retrieval Augmented Generation (GraphRAG) system
designed to help students navigate academic resources using a structured knowledge graph.
My STS research examined the societal implications of AI generated media, particularly
concerning authenticity, misinformation, and declining trust in digital content. Together, these
projects explore how engineering solutions can improve trust in AI systems while addressing
broader societal risks and ethical concerns.
STS Project
In my STS research, I examined how AI generated media is reshaping digital
communication and creating risks related to misinformation and authenticity. As generative AI
models improve, they are increasingly capable of producing realistic images, videos, and voice
recordings that are difficult to distinguish from human created content. This raises concerns
about fraud, misinformation, and declining trust in online platforms. Voice cloning technologies
and AI generated content have already contributed to impersonation scams and misleading
information. These developments highlight the need for reliable detection tools and
transparency mechanisms to help users evaluate digital content. My research concluded that
improving transparency and developing detection systems can reduce misinformation and help
maintain trust in digital environments.
Technical Project
The technical project aimed to produce a prototype Graph Retrieval Augmented
Generation system for the University of Virginia designed to help users navigate academic
resources. A knowledge graph was constructed containing courses, departments, professors,
and their research interests and publications, along with the relationships between them. When
a user submits a query, the system retrieves relevant information from the graph and passes it
to a large language model to generate a grounded response. A key feature of the system is
transparency, as users can view a visual representation of the retrieved graph and node
information to better understand how the AI generated its response. The system also
incorporates personalization using transcript data and conversation history. Testing
demonstrated strong factual accuracy, and user feedback indicated that the system was
perceived as helpful and trustworthy.
Conclusion
Working on both the STS and technical components reinforced the importance of trust
and transparency in artificial intelligence systems. My STS research highlights the risks
associated with AI generated content, while the technical project focused on building a
transparent and trustworthy AI tool. Together, these projects demonstrate that engineers must
consider both technical performance and societal impact when developing AI systems. By
combining sociotechnical analysis with technical development, this work contributes to
improving trust, accessibility, and responsible use of artificial intelligence in academic
environments.
