1. A Talk That Asks What We Lose When AI Stands In for People
Imagine walking into a room where the conversation is about something both deeply human and increasingly technological: what happens when we replace people with computer simulations in research? That is the question at the heart of a talk by Eun Cheol Choi, a Ph.D. candidate in Communication at the University of Southern California. His presentation, titled Proxies for People: Social Networks, Misinformation, and Simulated Respondents, invites us to think carefully about a practice that is becoming more common across many fields—using large language models as stand-ins for human beings. It sounds efficient, even futuristic. Why gather thousands of participants when an AI can simulate their opinions, attitudes, and behaviors? But Choi cautions that this convenience comes with hidden costs. He is not here to simply praise or condemn AI. Instead, he asks a subtler, more important question: when we swap a real person for a simulated proxy, what survives the transition, and what quietly disappears? The answer matters not just for academic research, but for how we understand misinformation, trust, and the social ties that shape our beliefs. This talk is designed for anyone curious about the intersection of communication, computing, and human behavior—and it promises to be as accessible as it is thought-provoking. For those attending in person, the setting will be warm and welcoming: light refreshments will be served, and ASL-English interpreters have been requested, making the event inclusive to a broader audience. At its core, Choi’s work is about people—how we connect, how we are influenced, and how we can responsibly use technology without losing sight of our humanity.
2. From Fact-Checking to Cutting-Edge AI Research: The Journey of Eun Cheol Choi
Choi’s path to this research is as interdisciplinary as his questions. He is currently finishing his doctorate in Communication at USC, where he also earned a master’s degree in computer science. That combination is rare and powerful. He is, by training, a computational social scientist—someone who moves fluidly between the messy unpredictability of human behavior and the precise logic of algorithms. Before arriving at USC, he studied communication at Seoul National University, earning both his master’s and bachelor’s degrees there. It was during that time that he worked at the SNU FactCheck Center, South Korea’s largest fact-checking platform at the time. This hands-on experience gave him a front-row seat to the real-world challenge of misinformation—not as an abstract concept, but as a daily battle to separate truth from falsehood in the public sphere. That experience left a lasting impression. It shaped his interest in why people believe and share false information, and how the structure of our social networks makes us more or less vulnerable. Since 2022, Choi has worked as a research assistant at the USC Information Sciences Institute, contributing to projects funded by DARPA and the National Science Foundation. He has also served as a teaching assistant for courses on data science and social networks, helping students navigate the same interdisciplinary terrain he explores in his research. His work has been published in respected venues, including the journal Social Networks and major conferences such as the International Conference on Machine Learning, the International AAAI Conference on Web and Social Media, and the ACM Web Conference. But his research is not just about publishing papers. He has built practical tools to help fact-checkers identify recurring misinformation using large language models—a clear sign that he cares about translating academic insight into real-world impact. At this talk, Choi looks forward to meeting new colleagues and exchanging ideas with faculty and students from communication, computing, and cognitive science. His journey shows that understanding misinformation requires more than one lens; it demands a willingness to cross boundaries and ask uncomfortable questions.
3. The Evidence: How Social Networks Shape Belief in Misinformation
To understand why Choi is so concerned about AI simulations, it helps to first understand what he has learned about misinformation in the real world. His research starts with a simple but powerful observation: no one believes misinformation in a vacuum. The people we surround ourselves with, the communities we belong to, and the information pathways we rely on all play a role in shaping our susceptibility to false claims. Choi and his collaborators conducted a survey of U.S. adults to examine this phenomenon closely. Their findings reveal that susceptibility to misinformation is influenced by two things working together: individual attitudes and the structure of social networks. In other words, it is not just about whether a person has certain beliefs or personality traits; it is also about how they are connected to others. Are the people around them diverse or homogenous? Do they encounter information through tightly knit groups or through loose, wide-ranging connections? Are they exposed to repeated misinformation from multiple directions, or do they have access to corrective voices? These structural features matter enormously. They can amplify a person’s existing biases or provide buffers against falsehood. This insight aligns with a broader shift in how researchers think about misinformation. It is not simply a problem of ignorance or irrationality; it is a social phenomenon, embedded in the fabric of relationships and communication channels. Choi’s work brings this social dimension to the forefront, showing that our networks are not neutral backdrops to our beliefs—they are active forces in shaping them. This is a crucial point because it means that interventions cannot focus only on individuals. We must also consider the environments in which people live, share, and communicate. And here is where the concern about AI proxies begins: if researchers use AI to simulate human respondents, and those simulations fail to capture the relational structures that shape real behavior, then the conclusions drawn from such research may be misleading. The problem is not that AI is useless; it is that AI may be giving us a distorted picture of reality while appearing convincingly accurate.
4. The Hidden Danger of “Silicon Samples”: Why AI Simulations Fall Short
Choi’s initial findings on simulated respondents reveal that current large language models have a particular weakness: they tend to overemphasize individual-level tendencies while struggling to replicate relational structures. To understand why this matters, consider what happens when a researcher wants to study how people are influenced by their social circles. If the AI model simulates each person as an isolated decision-maker, driven primarily by internal traits and opinions, it will miss the subtle, powerful effects of social context. The simulation may look perfect on the surface—the responses may sound human, the averages may match survey data—but the underlying relationships are distorted. Choi refers to these simulated participants as “silicon samples,” a playful but pointed term that reminds us they are not really people. They are approximations, built from patterns in language and data. And approximations can be dangerous when we forget they are approximations. The deeper issue is that many AI models are trained to predict what an individual would say or do based on their own characteristics, not to model how that individual is shaped by their connections to others. This creates a blind spot. In human life, relationships are not optional extras; they are foundational. Our beliefs are formed in conversation, reinforced by community, challenged by disagreement, and shaped by trust. If a simulation cannot capture those dynamics, then any study relying on that simulation will miss something essential. Choi’s research shows that seemingly accurate simulations can therefore distort the very relationships that matter most to researchers. This is especially troubling in the context of misinformation, where social influence is often the key driver of belief and sharing. A simulated respondent might accurately reproduce a person’s individual susceptibility to a false claim, but fail to show how that susceptibility is amplified or dampened by their network. The result is research that is not just incomplete, but potentially misleading—and misleading conclusions can lead to ineffective or even harmful policies and interventions.
5. A Path Forward: Rigorous Evaluation and Fairness in AI Research
So what should researchers do? Choi does not recommend abandoning AI simulations altogether. Instead, he advocates a more rigorous evaluation standard—one that treats AI-generated responses with the same skepticism and care we would apply to any research instrument. He proposes that AI-simulated responses should be assessed across multiple dimensions of fidelity, following criteria that social scientists already use to disentangle complex relationships among cognition, behavior, and social context. This means asking not just “Does the AI sound human?” but deeper questions: Does it accurately capture how people’s beliefs change in response to social influence? Does it reproduce the structural patterns of real communities? Does it preserve the diversity of perspectives and experiences found in actual populations? These are difficult questions, but they are necessary. Choi also acknowledges that current AI models are not equally good at representing everyone. Fairness challenges emerge when models represent certain populations more faithfully than others—often those already well represented in the data. This can lead to a troubling situation where research based on AI simulations silently overrepresents some voices and underrepresents others, reinforcing existing inequalities under a veneer of objectivity. To address this, Choi calls for careful attention to who is being simulated and who is being left out. His presentation offers both a cautionary perspective on using AI as a substitute for human participants and a constructive path forward. He is not opposed to innovation; he simply wants it to be responsible. By developing clearer standards for evaluating AI simulations, researchers can harness the power of these tools without falling into the trap of mistaking resemblance for truth. This work has significant implications across the social sciences, from psychology to political science to communication, and it comes at a moment when generative AI is rapidly becoming a standard tool in research. Choi’s proposed framework is a call for discipline and humility—a reminder that good science requires verifying our tools as carefully as we verify our theories.
6. An Invitation to Think Together About the Future of Research
Ultimately, Choi’s talk is an invitation—not just to hear about his research, but to join a conversation. He is eager to connect with colleagues in communication, computing, and cognitive science, and to explore the shared challenges and opportunities that AI presents. The subject matter may be technical, but the underlying concerns are universal. How do we know what we know? How do we trust the tools we use to understand human behavior? And how do we ensure that the future of research remains grounded in the rich, complicated, and irreplaceable reality of human life? These questions are especially urgent in a time when misinformation is rampant, social trust is fragile, and technology is evolving faster than our ability to fully understand its consequences. Choi’s work offers a valuable perspective: we can embrace the possibilities of AI while remaining clear-eyed about its limitations. By focusing on social networks, he reminds us that human beliefs are shaped by connection, and that any tool we build to study those beliefs must honor that truth. The talk will be a space for interdisciplinary dialogue, welcoming questions and different viewpoints. Light refreshments will be available, and ASL-English interpreters have been requested, ensuring that the event is accessible to a wide audience. Whether you are a researcher, a student, or simply someone curious about the intersection of people and technology, there will be something in this presentation for you. Eun Cheol Choi brings not only expertise, but genuine enthusiasm for exchanging ideas and learning from others. His presentation on Proxies for People is more than a summary of research findings; it is a thoughtful, timely, and humanizing reflection on what we stand to gain—and what we risk losing—when we invite machines to stand in for ourselves. In the end, the most important message is hopeful: by asking hard questions now, we can shape a future where AI serves as a complement to human understanding, not a replacement for it.