Robb Willer
banner
robbwiller.bsky.social
Robb Willer
@robbwiller.bsky.social

Professor of Sociology, Psychology, & Org Behavior, Stanford University
Director, Polarization and Social Change Lab, @pascl-stanford.bsky.social
Co-Director, Stanford Center on Philanthropy and Civil Society

Robb Willer is an American sociologist and social psychologist whose research has examined American politics, political psychology, moral persuasion, and cooperation.

Source: Wikipedia
Political science 34%
Sociology 23%
Pinned
Saturday's No Kings protests were massive, inclusive, and nonviolent. It was an honor to team up with
@owasow.bsky.social to analyze how nonviolent resistance can wins hearts, minds, and elections.

Reposted by Robb Willer

Thank you for leading us, @kristianvsf.bsky.social!

Reposted by Robb Willer

49 years ago today

The Clash begin recording their debut album at CBS Studios in London with producer Micky Foote. The album was recorded over three weekend sessions.

📸 Caroline Coon

#punk #punks #punkrock #theclash #history #punkrockhistory #otd

Reposted by Robb Willer

Reposted by Robb Willer

Reposted by Robb Willer

Reposted by Robb Willer

Clickable link to the working paper here: osf.io/preprints/ps...

H/t to @ballotpedia.org for their terrific work producing an excellent source of political information. Also to @yamilrvelez.bsky.social, @semrasevi.bsky.social, & Don Green for contributing to this growing area of study on AI voter guides.

**Major** kudos to @jsmernyk.bsky.social &
@jonnekamphorst.bsky.social for leading this project, and to co-authors Keshav Sivakumar and @adambonica.bsky.social

And thanks as well to
@pascl-stanford.bsky.social and the AI for Public Benefit Lab for supporting this project!

In summary: an AI voter guide grounded in a high-quality, nonpartisan information source can provide trusted and trustworthy information, constituting a potentially useful tool for voters:

👉https://osf.io/preprints/psyarxiv/whsm8_v1

But are people willing to use AI voter guides?
In a national survey (N≈2,800), we found...
⚡50% of Americans said they would use a validated AI voter guide
⚡14% reported already using AI to learn about past elections

We also found evidence that the AI voter guide helped voters align preferences and decisions. Voters using the guide reported higher intentions to vote for candidates whose issue positions and political ideologies were closer to their own views (reported at baseline).

Participants who used the AI voter guide reported
⬆️turnout intentions
⬆️intentions to vote for Democrats (though no change in intentions to vote for Republicans). This may seem surprising for a nonpartisan tool, but aligns with other work on how voter info impacts candidate preferences

We audited the guide: 89% of politically relevant sentences came from Ballotpedia data, suggesting minimal hallucination. Many remaining sentences were common knowledge (e.g., # of votes required to win the electoral college).

Judge accuracy for yourself: jsmernyk.shinyapps.io/voter_bot_tr...

Democrat, Republican, and independent participants alike generally liked, trusted, and engaged with the guide:
⚡On avg, they sent 4 msgs and spent 7 min interacting with it (only 2 min required)
⚡They also rated the guide as trustworthy, accurate, not biased toward either party, and enjoyable to use

We tested the AI voter guide the week before the 2024 US general election with 2,474 eligible voters in CA and TX who had not yet voted. Participants completed a baseline survey, used the AI voter guide (or searched for election info from their usual sources), then completed an endline survey.

Our AI voter guide:
⚡Relied on a trusted, nonpartisan source (@ballotpedia.org)
⚡Used best practices to ensure neutrality and adhere to source material
⚡Suggested prompts based on issues and races participants said they cared about

Voting in US elections is complex! This creates informational challenges, likely ⬇️ turnout, and makes it hard to choose candidates who align w/ voters' preferences. Voters increasingly turn to LLMs. Therefore responsible design and rigorous evaluation of LLMs' provision of election info is essential

🚨New WP: Can an AI voter guide (grounded in information from a nonpartisan, fact-checked source) help voters’ decision making? 🚨

We built and evaluated an LLM-based chatbot that provided voting info in CA & TX (N=2,474) right before the 2024 election. 🧵👇

Reposted by Robb Willer

Reposted by Robb Willer

Reposted by Robb Willer

Reposted by Robb Willer

Reposted by Robb Willer

Reposted by Robb Willer