Proxies for People: Social Networks, Misinformation, a Cognitive Science Speaker Series Presentation

Speaker: Eun Cheol Choi, a Ph.D. candidate in Communication at the University of Southern California (USC)

Title: Proxies for People: Social Networks, Misinformation

Short Bio: Eun Cheol Choi is a Ph.D. candidate in Communication at the University of Southern California (USC), where he also earned an MS in computer science. A computational social scientist trained across communication and computing, Choi studies misinformation and social networks, with a particular focus on the social-scientific application and evaluation of AI systems.

His research spans two connected threads: how the structure of people's social networks shapes their belief in and sharing of misinformation, and how faithfully AI systems can reproduce human beliefs, attitudes, and social structure when used as proxies for human research participants. He has also built tools that help fact-checkers identify recurring misinformation using large language models. His work has appeared in the journal Social Networks and at venues including the International Conference on Machine Learning (ICML), the International AAAI Conference on Web and Social Media (ICWSM), and the ACM Web Conference.

Since 2022, Choi has been a research assistant at the USC Information Sciences Institute, contributing to DARPA- and NSF-supported projects across communication and computer science, and he has served as a teaching assistant for Data Science for Communication and Social Networks at USC. Before USC, he earned his MA and BA in communication from Seoul National University, where he worked at the SNU FactCheck Center, South Korea's largest fact-checking platform at the time. Choi looks forward to meeting RIT's faculty and community and exchanging ideas about misinformation, social networks, and the responsible use of AI in social research.

Abstract: Abstract: Proxies for People: Social Networks, Misinformation, and Simulated Respondents Across disciplines, researchers increasingly use large language models (LLMs) as proxies for human judgments. Rather than claiming that AI proxies are simply beneficial or detrimental, Eun Cheol Choi's presentation asks what survives, and what quietly disappears, when a human subject is replaced by a simulated proxy ("silicon sample").

Building on research at the intersection of communication and computer science, Choi utilizes misinformation and social networks as a testing ground. Initial findings from a survey of U.S. adults indicate that susceptibility to misinformation is influenced by both individual attitudes and the structure of social networks. Subsequent analysis of LLMs prompted to simulate these respondents reveals that current models tend to overemphasize individual-level tendencies and have difficulty replicating the relational structures critical to human behavior. Simulations that appear accurate on the surface may, therefore, distort the relationships most relevant to researchers.

In response, Choi advocates a more rigorous evaluation standard. He proposes assessing AI-simulated responses across multiple dimensions of fidelity, following criteria social scientists use to disentangle complex relationships among cognition, behavior, and social context.

The presentation also addresses ongoing efforts to enhance simulation accuracy and examines fairness challenges that emerge when models represent certain populations more faithfully than others. This presentation offers both a cautionary perspective on using AI as a substitute for human participants and a constructive path forward, with significant implications for the responsible use of generative AI across the social sciences. Choi welcomes conversation with colleagues in communication, computing, and cognitive science.

ASL-English interpreters have been requested. Light refreshments will be provided.