Leveraging AI to Inform American Voters

Investment / Stage 1: Seed Partnerships

Woman voting in a U.S. election with the assistance of AI
Illustration: Eric Nyquist

    Voter participation in the United States remains low compared to other democracies, in part because of long ballots and frequent elections that span many offices and measures. Voter guides can improve participation, yet individuals with limited political knowledge are least likely to use them. Recent surveys find that 10 to 15 percent of Americans now use chatbots like ChatGPT or Claude to learn about elections, suggesting voters may already be using AI tools to understand their ballots. However, these tools were not designed to provide election information and can hallucinate facts, reflect partisan biases, or fail to provide jurisdiction-specific or up-to-date information.

    To address these limitations, our team developed and evaluated an AI voter guide grounded in nonpartisan Ballotpedia data during the 2024 election cycle. In a preregistered experiment among 2,474 eligible California and Texas voters, those who used the AI voter guide rated it as accurate, trustworthy, and unbiased. In addition, study participants reported increased intention to vote, reduced partisan animosity, and greater alignment between their policy preferences and their chosen candidates relative to a group of prospective voters who used their preferred source of election information.

    To expand this research, our team is partnering with VoteMate, a nonprofit that developed a nonpartisan AI assistant designed to work alongside a voter’s actual ballot by answering questions using jurisdiction-specific ballot and election data from Democracy Works, along with information from Ballotpedia, Congress.gov, and other trusted, verified nonpartisan sources. We will test whether a nonpartisan AI assistant can give voters the knowledge they need to make informed decisions on their ballot, and how it compares to other tools voters use to learn about elections. In a parallel audit study, we plan to compare VoteMate to frontier LLMs on accuracy, neutrality, and the kinds of election questions each is willing to answer. AI tools are already shaping Americans' voting behavior. We hope to learn whether, and under what conditions, that shift might help voters overcome informational barriers to participation and make well-informed voting decisions.

    A photograph of a person sitting on a wooden bench in a park, wearing a white top. The person's face is obscured with a blurred rectangle. The background features trees and greenery, creating a natural and relaxed atmosphere.
    Zeynep Biyikli

    Research Coordinator, POLITICS AND SOCIAL CHANGE LAB & AI FOR PUBLIC BENEFIT LAB

    A close-up photograph of a person standing in a park, wearing a dark blue patterned shirt. The person's face is obscured with a blurred rectangle. The background features soft greenery and blurred trees.
    Izzy Gainsburg

    Associate Director, POLITICS AND SOCIAL CHANGE LAB & AI FOR PUBLIC BENEFIT LAB

    A close-up photograph of a person standing outdoors, wearing a light blue shirt. The person's face is obscured with a blurred rectangle. The background is softly focused, showing architectural features.
    Jonne Kamphorst

    Assistant Professor, Political Science & Quantitative Social Science Methods, Sciences Po

    A portrait of a person wearing a dark blazer over a blue button-up shirt. The person's face is obscured with a blurred rectangle. The background is a dark, textured surface, giving the photo a professional look.
    Matt Latourelle

    Director of Technology Operations, Ballotpedia

    A portrait of a person with long, blonde hair wearing a black sweater. The person's face is obscured with a blurred rectangle. The background features a smooth, gray surface, enhancing the simplicity of the image.
    Alex Iovine

    Chief Executive Officer, VoteMate

    Joe Mernyk
    Joseph S. Mernyk

    PhD Candidate, Sociology; 2023 PhD Fellow

    Chrystal Redekopp
    Chrystal Redekopp

    Research Director, Politics and Social Change Lab & AI for Public Benefit Lab

    Robb Willer
    Robb Willer

    Professor, Sociology; Faculty Director, Politics and Social Change Lab & AI for Public Benefit Lab