ALT An image of a card that reads "Auditing Large Language Models as Intermediaries of Democratic Discourse in the 2026 US Midterms" with pictures of MIT political scientist Adam Berinsky and MIT Sloan professor Chara Podimata. It lists four research themes: Bias, persuasion, Politicization, and Democracy.

ALT An image of a card that reads "LLMs have become a major gateway to political information, positioning them as de facto intermediaries of democratic knowledge. Yet little systematic evidence exists on what they say about elections, how their responses change over time, or how they vary across demographic or ideological framings. "This project addresses this gap through a longitudinal study of LLMs during the 2026 U.S. midterm elections. We pursue three aims: measuring bias and implicit predictions in portrayals of elections, examining persuasion and personalization, and mapping politicization across models. Using large-scale computational auditing grounded in political communication theory, we analyze how LLMs shape electoral understanding and democratic legitimacy."