Retrieval Augmented Generation Application on Digital Backbone/ChatGPT as an Actant: Rethinking Human-AI Relationships Through Actor-Network Theory

My technical and STS research projects both examine the growing influence of large
language models on human systems. In my technical project, I worked on improving generative models by integrating long-term memory using Retrieval Augmented Generation. The goal was to overcome limitations in contextual awareness and make the system more helpful in real-time use. My STS project focused on the societal role of these models by analyzing ChatGPT through Actor Network Theory. While one project focused on building the system and the other on critiquing it, both explored how these tools are becoming more than passive technologies.
The technical project was a summer assignment for a biomedical venture capital firm. I created a user-ready RAG application using Amazon Web Services such as Bedrock, S3, and Service Catalog. The application allowed generative models to map and recall relevant data from cloud storage in a way that improved consistency and utility. While the system did not possess an internal model of the world, its ability to remember and respond in context made it feel closer to human reasoning. This technical work helped me see how even basic infrastructure changes could reshape user expectations and raise new questions about how these tools are used.
In my STS research paper I used Actor Network Theory to explore how language models like ChatGPT are not just tools but active participants in how people think learn and communicate. ANT helps explain how technology interacts with human actors rather than being controlled entirely by them. My paper argued that ChatGPT and similar systems now help structure information filter content and guide decisions. Their ability to remember context and
engage with users gives them agency within knowledge systems. This perspective makes it necessary to rethink their role and potential influence in society.
Working on both projects at the same time gave me a better understanding of how technical choices affect social outcomes. While building the RAG application I began to think more critically about how systems like it shape user behavior. Reading about ANT helped me understand why users treat these systems like intelligent partners. Even if the models are not conscious their interactions often feel personalized and persuasive. This dual perspective has shaped how I now approach design choices especially when considering ethics usability and long-term impact.

School of Engineering and Applied Science

Bachelor of Science in Computer Science

Technical Advisor: Rosanne Vrugtman

STS Advisor: Kathryn Webb-DeStefano

Technical Team Members: Emmett Hannam