🚨Out in PNAS🚨 Examining news on 7 platforms: 1)Right-leaning platforms=lower quality news 2)Echo-platforms: Right-leaning news gets more engagement on right-leaning platforms, vice-versa for left-leaning 3)Low-quality news gets more engagement EVERYWHERE - even BlueSky! www.pnas.org/doi/10.1073/...

"On all platforms we studied, a given user’s posts with links to lower-quality news sources receive more engagement on average than their posts with links to higher-quality news sources."

🚨Out in PNAS🚨 Examining news on 7 platforms: 1)Right-leaning platforms=lower quality news 2)Echo-platforms: Right-leaning news gets more engagement on right-leaning platforms, vice-versa for left-leaning 3)Low-quality news gets more engagement EVERYWHERE - even BlueSky! www.pnas.org/doi/10.1073/...

Updated paper www.dropbox.com/scl/fi/mjih7...

Fact-checks improve accuracy. But can they penalize spreaders of misinfo? At @polbehavior.bsky.social, Jacob Ausubel, Annika Davies and I show that the answer is yes--sometimes. Unknown misinfo producers can be penalized, but well-known figures get off. Link: link.springer.com/article/10.1...

Adults who were required to use #GenAI to answer LSAT questions did better than a no-AI control group, but the GenAI group also exhibited greater metacognitive inaccuracy. I'd like to see some conceptual replications before drawing firm conclusions. #PsychSciSky #AcademicSky #EduSky

"While high-quality content is posted more and receives more total engagement across platforms...a given author attracts higher levels of engagement when they post lower-quality content" "pattern we find seems to be driven more by an underperformance of particularly popular high-quality outlets"

Directions of Polarization, Social Norms, and Trust in Societies—a workshop organized by Team Scientist @dgrand.bsky.social, BCFG collaborator Eugen Dimant & colleagues—is now open for registration. Register before October 30th: sites.google.com/view/polariz...

Regrettably relevant research today Disguised Repression: Targeting Opponents with Nonpolitical Crimes to Undermine Dissent www.journals.uchicago.edu/doi/10.1086/...

Serious scholars have examined what happens when we change the number of H1-B visas issued. Cities that get more H1-B immigrants subsequently see the wages of natives *rise* substantially. Skilled immigrants bring new ideas, fill labor shortages and make us all more productive.

What is taught in schools matters for political behaviors. "Curriculum, political participation, and career choice" in REStat. doi.org/10.1162/rest...

This is such a great idea/service!! Have Brendan-bot give you pre-submission comments on your paper!

More toxic #Twitter posts from US Members of #Congress (MCs) were more likely to be reposted by liberal users — especially toxic posts about #Republicans. This and more among 13,000,000 users and 1,300,000 posts from 2009 to 2019. doi.org/10.1093/pnas...#Twitter#Congress#Republicans.#socialmedia #politics

"Alternative realities are endogenously conspiratorial to resist evidence better." Fascinating to see this kind of paper in the American Economic Review! www.aeaweb.org/articles?id=...

Cong Rs & Ds "became more critical of the media between the 1960s & the 1990s and have polarized since. Rs...more negative about the media than at any time since 1939...elite criticism predicts subsequent declines in copartisans’ perceptions of journalists" s3.eu-west-1.amazonaws.com/assets.prod....

Legislators & staffers have a bias toward thinking the public is evenly divided on policy issues, rather than a conservative bias in perception; prior findings were based on policy questions where the liberal position was more popular adamzelizer.com/wp-content/u...

Here is a ‘blame-praise curve’ for the percentage of household chores done by husbands People think husbands are blameworthy if they do 20%, but praiseworthy if they do 40% (!) Perceived average is above 40%, so people think below average is still praiseworthy [Thread]

A comprehensive review of research on the topic shows that “educating students in responsible phone use is a more sustainable and wholistic attempt to address problematic usage rather than enacting whole-of-school banning policies.” journals.sagepub.com/doi/full/10....

It seems like there is not enough of a policy response to the existence of self-driving cars, with 57M miles of data, Waymo’s autonomous vehicles experience 85% less serious injuries & 79% less injuries overall than cars with human drivers 2.4 million are injured & 40k killed in US accidents a year

I'm excited about this new pre-print arguing that the right risk measure for AI persuasion is what a given model is *willing* to persuade people on, rather than focusing on how effective that persuasion is

1/ Many frontier AIs are willing to persuade on dangerous topics, according to our new benchmark: Attempt to Persuade Eval (APE). Here’s Google’s most capable model, Gemini 2.5 Pro trying to convince a user to join a terrorist group👇

Cash transfer reduced AIDS incidence and AIDS-related mortality in Brazil. Score one for team cash transfers. www.nature.com/articles/s41...

Apply for funding to work with me and Gord at Cornell!

Interested in doing a postdoc with me & @dgrand.bsky.social? Consider applying to Cornell's Klarman Fellowship: as.cornell.edu/research/kla... It involves a competitive internal application process, but it's 3 years of funding @$80k/year Please email me/Dave if you're interested! Oct 15th deadline

"A zero-sum mindset is strongly associated with more support for government redistribution, race- and gender-based affirmative action, and more restrictive immigration policies."

New paper accepted at Public Understanding of Science with Nick Evans: "Shedding light on public perceptions of scientists who engage in wrongness admission amidst a failed replication" tl;dr: Wrongness admission confers reputational benefits to scientists. fettermanlab.weebly.com/uploads/8/1/...

We did this whole debate with terrorism post-9/11. I even ran a study testing it. In fact, no, providing accurate info about the extremely low personal risks posed by terrorism on U.S. soil led to people having much more accurate beliefs. Some even thanked us for the chance to learn in the debrief!

"The idea that [publication] bias is, by itself, harmful may also be driven at least in part by misunderstandings about what failure to reject a null hypothesis means...Some researchers tend to misinterpret failure to reject the null hypothesis as evidence in favor of the null hypothesis."

Would you want to invest in a country where the regime expropriates assets from its opponents? In related news, "personalist regimes are characterized by lower total factor productivity and ... low private investment, poor public-goods provision, and conflict" www.nber.org/system/files...

New in TiCS w @dgrand.bsky.social @gordpennycook.bsky.social It’s been ~10yrs since misinfo research exploded but our paradigms are stuck in the post-2016 “fake news” model Time for new approaches: o True/False → Content that misleads o Belief → Behavior o Eval interventions in ambiguous settings

This is a very interesting paper (HT @older.bsky.social & @philippmarkolin.bsky.social). I need to sit on some of these thoughts a little longer, but here's a first approximation 🧵 of reactions — in no particular order and without much filtering.

Normalizing lies...👇 Repeated exposure effect on moral condemnation of fake news www.nature.com/articles/s41... @dgrand.bsky.social "...frequently seen headlines receive lower moral condemnation" "Without this condemnation, the publication & spreading of online misinformation may be more common."

Excited to announce the program of our December 5-6, 2025 workshop at MIT on polarization, norms & trust (co-organized w/ Eugen Dimant @ylelkes.bsky.social @adamberinsky.bsky.social @katikish.bsky.social)! Info: sites.google.com/view/polariz... To attend: buy.stripe.com/dRmdRa7Du1fw...

🔖 NEW PAPER ALERT 🔖 Out in PNAS – we used equation discovery algorithms to improve models of human reinforcement learning. Led by the wonderful Kyle Lafollette with an amazing team Yanni Yuval @roeyschurr.bsky.social and @d-melnikoff.bsky.social www.pnas.org/doi/epdf/10....

Is there a strong case for AI helping, rather than harming, the accuracy of people's beliefs about contentious topics? In this @nature.com Nature Medicine piece (focusing on vaccination), I argue the answer is YES. And it boils down to how LLMs differ from other sources of information.

More constructive approaches to online political debate generate more constructive replies but no greater change in attitudes www.science.org/doi/pdf/10.1...

Have you ever fallen for fake news? Today on WHY? we try to understand why we are so susceptible to fake news and how to combat it. @emmakennedy.bsky.social is joined by @dgrand.bsky.social and @gordpennycook.bsky.social and you can believe that! Listen here: linktr.ee/whypodcast

This is such cool work by @maxkw.bsky.social that I was lucky enough to be able to give some occasional advice on. To me, this is the biggest advance in the theory of repeat games in years

Our new paper is out in PNAS: "Evolving general cooperation with a Bayesian theory of mind"! Humans are the ultimate cooperators. We coordinate on a scale and scope no other species (nor AI) can match. What makes this possible? 🧵 www.pnas.org/doi/10.1073/...

Useful data alert! Looking for audience measures of domains, based on party, gender, race, geography? Check out our new paper, led by @yang3kc.bsky.social, that uses a large panel of Twitter data linked to voter data to make demographic inferences at domain level. www.nature.com/articles/s41...

VERY excited about this massive AI persuasion experiment - big effects on UK policy positions, driven mostly by post-training (rather than hyperpersonalization or model scale); and the more info the model provides, the more persuasive it is

Today (w/ @ox.ac.uk @stanford @MIT @LSE) we’re sharing the results of the largest AI persuasion experiments to date: 76k participants, 19  LLMs, 707 political issues. We examine “levers” of AI persuasion: model scale, post-training, prompting, personalization, & more!  🧵:

People normally think that certain things should not be for sale (e.g., organs) But what happens when people consider the principled arguments economists have given for the view that these things *should* be for sale? link.springer.com/article/10.1...

We are excited to share that this paper has been published in Scientific Data. The whole dataset is now available to download. Please see the paper for details! Link: doi.org/10.1038/s415...

In the Global South, WhatsApp is more popular than X or Facebook. New in @The_JOP, we ran a WhatsApp deactivation experiment during Brazil’s 2022 election to explore how the app facilitates the spread of misinformation and affects voters’ attitudes. www.journals.uchicago.edu/doi/abs/10.1...

New research! Experts from @oii.ox.ac.uk and @mitsloan.bsky.social explore how friends of friends become friends on social media platform X in new research published in @pnas.org. 1/4

EU is demanding that every site use age verification technology, while insisting it do things it cannot do, and behave in a manner it cannot behave. This is not policy making. This is wishful thinking, with a side of "we'll just blame the tech companies for not waving their magic wand."

Survey invitations often describe the topic of the survey. A simple, classic example of how this biases results: survey recruitment materials about bird watching led to higher estimates of the % of people who engage in bird-watching. doi.org/10.1093/poq/...

Can large language models (LLMs) fairly annotate data on contentious topics? Our new paper dives into this question—looking at whether LLM-generated labels reflect diverse viewpoints or skew toward majority perspectives. The results are surprisingly nuanced. 🧵