AI Applications for Survey Research: A MAPOR (Online) Summer Camp
Friday, July 17th, 12pm-1pm CT: Validation and Inference for Silicon Sampling in Survey Research
Friday, July 24th, 12pm-1pm CT: Using Large Language Models to Identify Issues with Question Construction
Friday, July 31st, 12pm-1pm CT: Why Survey Researchers Are Well Prepared for Generative AI: A Hot (Deck) Take
The Midwest Association for Public Opinion Research (MAPOR) is pleased to announce the Summer Webinar Series “AI Applications for Survey Research: A MAPOR (Online) Summer Camp”. As generative AI tools become increasingly accessible, survey researchers are exploring how they can be incorporated into existing research workflows. From generating synthetic survey responses to evaluating questionnaire quality, AI offers new opportunities to improve efficiency and support decision-making. This three-part webinar series examines timely applications of AI in survey research, providing an evidence-based assessment of where these tools show promise, where caution is warranted, and how practitioners can integrate them responsibly into their work.
This webinar is free for MAPOR members and all students.
Validation and Inference for Silicon Sampling in Survey Research
Friday, July 17th, 12pm-1pm CT
In both research and commercial applications, AI is increasingly being used to create synthetic representations of humans. This raises at least two important questions for scientists as they consider the integration of this new tool: How can a researcher know if the silicon survey output is valid? And how can a researcher make new inferences on the basis of synthetic output? Drawing on a synthesis of research on silicon sampling, I discuss the intellectual foundations of this approach, the ongoing practical and theoretical challenges, and cutting-edge methodological tools. I conclude by proposing some best practices for conducting transparent open science using large language models.
 | Dr. Lisa Argyle is an Associate Professor of Political Science at Purdue University. She uses computational social science tools, including AI, surveys, experiments, and text analysis, to better understand how people talk with each other about politics, and the impact that public deliberation has on polarization and political division. Her recent work emphasizes the use of AI as a research tool, with applications in public opinion surveys and political communication research workflows. |
Using Large Language Models to Identify Issues with Question Construction
Friday, July 24th, 12pm-1pm CT
Evaluating survey questions using established questionnaire design best practices is essential for minimizing burden and error and maximizing data quality, but manually identifying construction issues can require substantial time and deep expertise. Large language models (LLMs) such as ChatGPT and Claude offer the potential to automate parts of this evaluation process, yet questions remain about their accuracy and how prompting strategies influence performance. In this webinar, we’ll show results from a study that evaluates ChatGPT and Claude’s ability to identify common question construction issues across a diverse set of survey questions. You’ll learn where LLMs can improve efficiency, where human expertise remains essential, and how these tools can be incorporated into the questionnaire design process.
 | Ashley Griggs is a Survey Methodologist at RTI International with extensive expertise in crafting survey questions and designing questionnaires for mail, web, and in-person interviews. She leads instrument development on surveys of individuals, households, and establishments, focusing on aspects such as question wording, pretesting (including cognitive and usability testing), and contact strategies. She specializes in reducing measurement error, minimizing mode effects, and boosting participation in longitudinal studies. |
 | Jerry Timbrook is a Survey Methodologist and Applied Social Sciences Manager at RTI International with a specialty in designing questionnaires for mail, web, and computer-assisted telephone interview (CATI) surveys. He employs behavior coding, analysis of paradata (e.g., keystroke logs, response timing data), cognitive interviews, and usability tests to inform questionnaire development. As a computationally minded social scientist, Dr. Timbrook also uses computer-based methodological techniques like machine learning and generative/agentic AI to improve the quality of survey data. |
Why Survey Researchers Are Well Prepared for Generative AI: A Hot (Deck) Take
Friday, July 31st, 12pm-1pm CT
One of the first studies assessing how generative artificial intelligence would alter the workplace pegged survey research as among the most impacted fields. Yet it is becoming increasingly clear that LLMs cannot replace survey methodologists or the need for real data collection. But this does not mean that generative artificial intelligence does not have a role in survey research. In this talk I argue that survey researchers already have a strong lens for understanding when and how GenAI is likely to improve survey interpretation and administration—that lens is data imputation.
Imputation has long been central to how survey researchers handle uncertainty: sampling error, item nonresponse, imperfect measurement, and the conceptual distance between observed answers and the constructs we seek to capture. Thinking of LLMs as imputation tools—rather than replacements for human expertise—helps clarify what they can and cannot do. Their strengths lie in narrow, pattern-based tasks: generating plausible text, categorizing open-ended responses, or filling in auxiliary detail. In these contexts, they operate much like complex, data-driven hot decks—useful, but bounded.
This analogy also illuminates their limitations. Synthetic-data studies show that LLMs often fail to reproduce the true variability of human thinking. These are hallmarks not of broad intelligence, but of constrained pattern completion. And this way of understanding the technology fits much more closely with what LLMs are actually doing than the AI-as-person-replacement model. It also highlights the importance of key questions about the replicability and black box nature of the results they generate.
Framing GenAI this way can help to both identify productive and responsible integrations. By treating generative AI as another imputation method, survey methodologists can make informed decisions about where it adds value, where it introduces risk, and how to ensure that interpretation remains grounded.
 | Josh Pasek is Professor of Communication & Media and Political Science at the University of Michigan. His research examines how information, identities, and communication environments shape political beliefs, public opinion, and political behavior. He also studies how surveys, digital trace data, and other forms of social data can be used to measure public opinion and better understand society. |