Chapter meeting -
Colloquium talk
Colloquium talk
Silicon sampling—the use of large language models to simulate responses from human populations—has opened new methodological opportunities and challenges for public opinion research. Its promise rests on the possibility that patterns embedded in human language can support useful simulations of political attitudes, judgments, and behavior. Realizing that promise requires confronting fundamental questions about representation, validity, and inference. This presentation will begin with a brief overview of the varied schools of academic research that form the intellectual foundations of silicon sampling. This will be followed with a discussion of the many challenges and uncertainties facing silicon sampling in practice, particularly how to distinguish the production of plausible responses from the generation of valid evidence. The central focus of the presentation is where silicon sampling can contribute to valid scientific knowledge and what standards and practices should guide its use.
Dr. Lisa P. Argyle is an Associate Professor of Political Science at Purdue University and an Associate Co-PI of the Cooperative Election Study. Her research brings together public opinion, political behavior, and computational social science. She examines how people form and communicate political views, online and offline, and has become a leading contributor to the growing discussion of how researchers can use large language models responsibly in social-science research.
Dr. Argyle’s recent work on “silicon samples” explores whether large language models can help researchers study public opinion by generating responses that reflect patterns in human survey data. This work is of particular interest to statisticians because it raises fundamental questions about representativeness, bias, validation, and the reliability of inference. Her research offers a thoughtful perspective on the opportunities and necessary cautions of bringing AI-based tools into survey research.
This event is co-hosted with the Data Analytics Colloquium at UTD's School of Economic, Political and Policy Sciences (EPPS).
Date and Time: November 17, 2026 (Tuesday), 8:00–9:00 PM CST
Location: Online via Microsoft Teams
Free of charge and open to everyone, including non-ASA members
You may be asked to complete a brief registration on the Microsoft Teams webpage before joining.