New Pre-print out today! We're releasing Political DEBATE -- a new set of language models for zero/few-shot classification of political text. The models are open source, small enough to run on your laptop, and as good as proprietary LLMs within domain. arxiv.org/pdf/2409.02078

Job ad is now live! 18-month postdoc. with me Chris Bail (chris_bail_duke 🧵) Petter Törnberg and Ancsa Hannak + great team of PhDs and other postdocs. my.corehr.com/pls/uoxrecruit… Come work with a great team. Deadline is Jan 31 📆

New Course: Prediction for (Individualized) Decision-making statmodeling.stat.columbia.edu/2024/12/06/new…

Very happy to see our review article finally in print!

New Media Bias Detector 🔎 Blog Post 📝 - 2024 Election Recap: What the Media Told Us Duncan Watts David Rothschild 🌻 mediabiasdetector.seas.upenn.edu/blog/2024-elec…

Studies of the effects of misinformation risk harming participants, according to a new study. Standard debriefing efforts to counteract falsehoods presented to participants during experiments often fail, leaving participants misinformed. In PNAS Nexus: ow.ly/YN8u50Untry

Though more women hold political office in India, it's still male family members that hold political sway. Soledad Artiz Prillaman's research collaborations explore how to achieve genuine empowerment and representation. Read the Stanford Impact Labs story! bit.ly/3ZQSoVT

"The concentration of AI power in a few corporations threatens democracy. Public accountability is the only antidote." Schneier Blog + Nathan Sanders weigh in on the debate around open AI models, arguing open source models are a public good. bostonreview.net/forum_response…

Here's a preprint for you to read on the first week of 2025! It's my largest paper yet, where we analysed ~200mil users over 7 events from X to define "What is a Social Media Bot", and the bot-human dichotomy arxiv.org/abs/2501.00855 (1/5)

7 steps to junk science that can achieve worldly success statmodeling.stat.columbia.edu/2025/01/17/7st…

When estimating a treatment effect with a cluster design, you need to include varying slopes, even if the fit gives warning messages. statmodeling.stat.columbia.edu/2025/01/23/slo…

Deadline approaching: The CEPR Workshop on Media, Technology, Politics, and Society is hosted by Bocconi University in June 2025! Submission deadline is January 31, 2025. We accept extended abstracts but priority will be given to full papers.

Tiktok boosted Republican messages 11.8% more than Democrat messages during the 2024 U.S. Presidential Race according to this new preprint that employed bots to view ~394,000 videos: arxiv.org/pdf/2501.17831

Graphical display of election forecast uncertainty statmodeling.stat.columbia.edu/2025/02/02/gra…

We have a new version of our 𝕡𝕣𝕠𝕞𝕡𝕥𝕤𝕥𝕒𝕓𝕚𝕝𝕚𝕥𝕪 paper now on arxiv! This is a significant update that testa *a lot* more data, suggests post-processing techniques, outlines how to compare across models, and tests with new models...

The advance copy of my book, Safe Havens for Hate, is here! It offers a deep dive into the moderation of harmful content on social media and explains why efforts to combat online extremism keep failing. Official release March 4! press.princeton.edu/books/hardcove…

The R-squared on this is kinda low, no? statmodeling.stat.columbia.edu/2025/02/28/the…

Benjamin Franklin wearing a coonskin cap while being received by the French court. "Proper" attire at the time would have been a powdered wig, but Franklin enjoyed playing up his "rustic" American identity. On these trips, he asked for military support for newly independent US.

“A Post Mortem on the Gino Case”: “Committing fraud is, right now, a viable career strategy that can propel you at the top of the academic world.” statmodeling.stat.columbia.edu/2025/03/08/a-p…

Thrilled to be hosting the 3rd MEEW (Media Effects Empirical Workshop) at Columbia again. Hope to see some of you there! RSVP link is available at: sites.google.com/view/columbia-…