Abstract
My technical project and STS research paper both examine information flow within and across knowledge networks, and how sociotechnical relationships impact these flows. In my technical project, I developed a large language model (LLM)-based AI agent and constructed a knowledge network of human and non-human actors to analyze how ideas diffuse across the network. To analyze social dynamics that impact knowledge flow and innovation, my STS research paper examines the network formed by Elmhurst hospital management and the New York City Government to frame an effective response to the COVID-19 pandemic in its initial months.
In my technical project, I developed an AI agent designed to analyze knowledge diffusion or the spread of ideas among people, fields, and organizations to drive innovation and progress, using a temporal scientific knowledge graph. Structural shifts in topics across communities and in community composition tend to be crucial predictors of changes in scientific direction. Hence, the agent analyzed the formation, splitting and merging of communities, such as research groups at an institution or scientists collaborating across the world, and the evolution of topics covered by these communities over time. Despite extensive research conducted in this area, there has not been an interactive system developed to date to analyze such trends. This agent takes in open-ended user queries as input and outputs results on community dynamics and topic migration over time, setting the stage for advances in the science of science through interactive knowledge graph analysis.
My STS research paper examines the collapse of Elmhurst Hospital’s during the early months of the COVID-19 pandemic. Using Actor-Network Theory (ANT), I identified human actors, such as medical personnel, patients and administrators, and non-human actors, such as medical equipment and hospital infrastructure, within Elmhurst’s network. I argued that Elmhurst’s breakdown was due to a lack of coordination within and between interconnected networks of people, tools and systems within and outside the hospital and a failure to align with the broader New York City hospital network . By framing the hospital as a sociotechnical network, my paper shows how degradations in network relationships limited knowledge flow and hindered an effective COVID-19 response.
These two projects were developed separately, however insights gained from the STS research will aid future related technical work. Understanding how failures in alignment among actors and network translation can stem knowledge flow, especially in crisis situations, emphasizes the importance of considering sociotechnical relationships in systems like the AI agent-based knowledge diffusion analysis system developed in the technical project. I can apply this idea to future projects in the same field by incorporating sociotechnical dynamics, for instance mentorship connections between authors or institution migration, into the knowledge diffusion analysis. This will help ensure that designed knowledge networks are robust and applicable to real-world relationships and situations. Integrating sociotechnical awareness into the design of technical systems can improve the effectiveness, adoption, and long-term stability of future projects in my field.
