Thanks for the interest in our recent paper arxiv.org/pdf/2410.24190, just published in #EMNLP2024! LLMs' political leaning and their potential influence on our democracy is a complex and important topic. What's the path forward? future-of-democracy-with-llm.org Recent work including

Like bubbles in the market, science can experience unsustainable hype cycles. With the amazing Donghyun Kang, Robert Danziger, and Jalees Rehman , our new Nature Human Behaviour paper (rdcu.be/d10OE) just out reveals how limited diffusion of scientific ideas—when research stays

🧵 Thrilled to share our new Nature Communications paper on how different approaches to fact-checking affect echo chambers on Twitter/X! With the amazing @junsolk, Zhao Wang, HaohanShi, and Hsin-Keng Ling 凌心耕. When someone gets fact-checked by an individual (via direct replies), they retreat

Excited to share that our paper on Community Notes has been published in Nature Communications! Nature Communications 📄 nature.com/articles/s4146… A huge THANK YOU to our incredible team! James Evans, Zhao Wang, HaohanShi, and Hsin-Keng Ling 凌心耕

[Behind the Paper] Individual Misinformation Tagging Reinforces Echo Chambers; Collective Tagging Does Not communities.springernature.com/posts/individu…

Research shows that limited diffusion of #knowledge predicts abrupt collapses in the use of #scientific #ideas, like bursting #market #bubbles. -By Donghyun Kang and James Evans The University of Chicago , Google Article: buff.ly/4aGnIKE #innovationinscience #knowledgediffusion

I'm excited to be travelling to Oxford tomorrow for the Artificially Intelligent Social Science Conference I'm organizing w Ray Duch. Any grad. students interested in attending, send me an email :)

Very excited about this work. Junsol Kim did an amazing job leading it!

Really excited about this new work with the amazing Junsol Kim and Aaron Schein that allows analysts to surface, evaluate, and precision-steer personas and perspectives deep within modern large language models. Check it out!

In the latest Science, Henry Farrell, Cosma Shailizi, James Evans and I make the case for LLM's as powerful, transformative cultural and social technologies, ways for people to learn from other people, rather than intelligent agents. science.org/doi/10.1126/sc…

New in Science Magazine: "Large AI models are cultural and social technologies" Working with brilliant colleagues Henry Farrell, Alison Gopnik, and Cosma Shalizi, we challenge the prevailing narrative about AI models as autonomous agents. science.org/stoken/author-…

Interested in how we can monitor and steer the subjective perspectives of AI? I’ll be presenting our work on how language models represent political perspectives this Saturday at #ICLR2025 — would love to connect! Linear Representations of Political Perspective Emerge in Large

Are biased AI systems always harmful — something to be eliminated at all costs? Our 7,500-trial study says: it’s more complicated, and often, NO. 🧵👇

Excited to share our latest preprint, led by amazing Shiyang Lai, Nadav Kunievsky, Yujin Potter, and James Evans. We find that culturally biased AI unintentionally improves human decision-making by reducing trust and fostering cognitive engagement. arxiv.org/abs/2508.09297