CSH Lecture Series on Network Inequality – with Indira Sen

"Multilingual and Multicultural Misrepresentation in LLM Simulations of People"

Social simulation presents an intriguing and potentially revolutionary use case for Large Language Models (LLMs). Some researchers suggest that human biases encoded in LLMs, due to the data they've been trained on, can be exploited to mimic people with greater fidelity. Promising results have been found in using LLMs to simulate survey respondents, annotators, and even more complex simulations of groups of people. However, under which circumstances are the biases in LLMs true mirrors or people, and when are they distortions or exaggerations? To shed light on this question, I will present three case studies --- 1) using LLMs to simulate data annotators, 2) using LLMs to generate training data for Machine Learning Models, and 3) using LLMs to simulate moral reasoning. Each of these studies shows the limitations of current LLM technology in faithfully and realistically simulating people due to an inability to represent minority and marginalized subgroups, especially for languages beyond English.

About Indira Sen:
Indira Sen is a Postdoctoral researcher at the Political Science and Public Policy Department at the University of Konstanz. Her research is about understanding and characterizing the measurement quality of social science constructs like political attitudes and abusive content from digital traces, combining NLP and social science measurement theory. In the past, she was a doctoral student at RWTH-Aachen and GESIS-Leibniz Institute for Social Science, working in Computational Social Science. She has interned at Snap Inc. and Nokia Bell Labs. You can reach her at @indiiigosky on X/Twitter or https://indiiigo.github.io/


Learn more about the Complexity Science Hub: https://csh.ac.at/
  / complexity-science-hub  
  / cshvienna  
  / cshvienna  
  / cshvienna