Can AI tell us what people really think?

In this first AI in Society Speaker Series webinar, Francis Guesquiere, IPUR Senior Research Affiliate, speaks with Mark Polyak, Chief Operating Officer of MINT AI, about how AI is transforming public opinion research. The discussion explores why polling failures in 2016 reflected long-standing issues such as sampling bias, timing, and undercounted voter groups, and how the industry responded by combining traditional surveys with social media analysis, transactional data, and mobility data.

The webinar also examines how public behaviour has changed, with growing polarisation, declining trust in polls, lower response rates, and survey fatigue making opinion research more difficult. Mark then discusses the rise of ChatGPT and frontier AI models as a new source of opinion data, arguing that chat logs may become a kind of “new census” because of the scale and intimacy of information people now share with AI systems.

Other key topics include synthetic panels, digital twins, behavioural world models, AI guardrails, political persuasion, and the risks of highly personalised influence. The session closes with a preview of the next webinar featuring Carl Benedict Frey on automation, productivity, and public backlash against AI.