A Dartmouth study showing that artificial intelligence can rig public opinion polls has been recognized as one of the most significant pieces of scientific research published last year.
Associate Professor of Government Sean Westwood’s study on the threat of large language models to online survey research received the 2025 Cozzarelli Prize, awarded by the journal Proceedings of the National Academy of Sciences (PNAS), in the Behavioral and Social Sciences category.
The prize recognizes scientific excellence and originality. This year’s winners were chosen from more than 3,600 articles published in PNAS, one of the world’s most-cited multidisciplinary scientific journals.
“This recognition reflects the urgency of Sean’s work,” says Professor of Government Benjamin Valentino, associate dean for the social sciences. “For years, changes in communications technologies have been making accurate polling more difficult, but Sean shows that AI presents a new and potentially greater threat to the survey methods we rely on to understand public opinion. Dartmouth is proud to have faculty doing work this consequential.”
The research documents a striking vulnerability at the heart of modern social science. Autonomous AI agents, Westwood found, can complete online surveys with coherent responses that evade nearly all existing detection methods. The agents reason through complex questions, maintain a convincing demographic persona, as if they are human, and remember prior responses to build consistent answers across an entire survey. Even without malicious intent, they can infer a researcher's hypothesis and produce data that confirms it. With a single instruction, the AI agents can systematically bias polling results as part of coordinated information warfare.
The real-world stakes are significant. In the seven major national polls before the 2024 election, adding as few as 10 to 52 fake AI responses—at five cents each—would have flipped the predicted outcome. The bots work even when programmed in Russian, Mandarin, or Korean, yet produce flawless English answers.
“We can no longer trust that survey responses are coming from real people,” says Westwood. “With survey data tainted by bots, AI can poison the entire knowledge ecosystem.”
The implications reach well beyond election polling. Surveys underpin scientific research across disciplines—from psychology to public health to economics—and thousands of peer-reviewed studies published each year rely on survey data to shape research and policy. The threat is one Westwood had already been tracking: a 2024 report from his Polarization Research Lab—a nonpartisan collaboration with researchers at the University of Pennsylvania—found that half of Americans believed AI would make elections less civil, and 65% worried it would harm personal privacy.
Westwood tested every AI detection method currently in use and found that all failed to identify his tool. His study calls for transparency from companies that conduct surveys, requiring them to prove their participants are real people.
The study is a natural extension of Westwood’s career-long focus on political behavior, public opinion, and the forces that shape—and distort—democratic participation. His work examines how partisanship and information from political elites affect the behavior of citizens, with a focus on understanding where partisan biases originate, where they manifest, and their bounds. His research is frequently covered by the New York Times, Washington Post, Wall Street Journal, NPR, CNN, and Bloomberg, among other outlets.
“We need new approaches to measuring public opinion that are designed for an AI world,” he says. “The technology exists to verify real human participation; we just need the will to implement it. If we act now, we can preserve both the integrity of polling and the democratic accountability it provides.”
Westwood was recognized at an awards ceremony during the National Academy of Sciences Annual Meeting in April. Established in 2005, the Cozzarelli Prize honors late PNAS editor-in-chief Nicholas Cozzarelli and his uncompromising scientific standards and dedication to advancing knowledge.