Our new paper is out today in @pnasnexus.org with colleagues at Yale (@matthewshu.com, Danny Karell, @keitarookura.bsky.social)
We wanted to understand how using AI-generated summaries to learn about history influenced attitudes compared to existing resources like Wikipedia. 1/4
π£CFP: 7th edition International Workshop on Cyber Social Threats (CySoc)
We welcome papers that examine a diverse range of issues related to online harmful communications.
π
Submission: March 22nd, 2026
π
Notification: April 8th, 2026
π Details: cy-soc.github.io/2026/
I've seen a few history of T&S things but don't remember any with that focus. If you come across the right people to write it please send them our way! tsjournal.org/index.php/jo...
π¨π¨π¨
Can #AI #chatbots reliably tell you whether a political claim is true or false? If not, what would it take to make them trustworthy fact-checkers?
A new study led by Matt DeVerna tackles these questions by evaluating 15 #LLMs on more than 6K claims fact-checked by PolitiFact over an 18-year period.
Here's the recording! www.youtube.com/watch?v=MsH_...
Amongst other things, this again highlights the core point about due process. Itβs bad enough to be in a βshow me your papersβ situation, but your precious citizen/non-citizen, βlegalβ/βillegalβ distinctions are meaningless when you canβt prove anything to authorities because they donβt give a shit.
This is huge news. I have spent the past 6 months wondering wtf was up with Amazon: they filed 380,000 AI-related CyberTipline reports to NCMEC in the first half of 2025.
Turns out ALL of it was known CSAM they found by screening their AI training data. It's NOT AI-generated or AI-morphed CSAM.
"Machine learning research is not serious research & therefore hallucinated references are not necessarily a big deal, agrees prestigious group of ML researchers"
I β€οΈ when the titles write themselves
But seriously, I don't think there's much to debate statmodeling.stat.columbia.edu/2026/01/26/m...
This is courage. Looking a masked, government fascist dead in the eye, face to face, knowing full well that your life is on the line.
If elected officials had 1/10th of Alex Pretti's resolve, we would have far fewer troubles.
I have an op-ed in the NYT today about the Grok scandal, sharing my research from last year finding that legal risk hinders AI companies from making their models safer against CSAM - an echo of the years where white-hat hackers were chilled from good-faith research. www.nytimes.com/2026/01/12/o...
A picture of the website where you can register for the course.
π©βπ« I spent 6 months building something Iβm super proud of: a course for journalists on using AI for investigations. It officially launched today! These approaches have already improved my own investigations. Get started: careercatalyst.asu.edu/programs/ai-...
New research on Gifted dogs is out, in @science.org!
Huge congrats to SHANY DROR for her effort and this incredible achievement.
π www.science.org/doi/10.1126/...
π¨ New paper from an awesome group led by Noam Kolt and
@nickacaputo
.
We hear a lot about what important concepts and methods from AI research that lawyers need to understand. But it's really a two-way street...
π§΅π§΅π§΅
Reminds me of a passage from Kurt Vonnegut's Timequake. Can't fit the whole thing here, but the key part goes: "... the situation is social rather than scientific. Any work of art is half of a conversation between two human beings, and it helps a lot to know who is talking at you."
Johns Hopkins Data Science and AI Institute is hiring Postdoctoral Fellows (Deadline Jan 23rd, 2026)! π«
Reach out and apply if you're interested in working with me! I'm especially excited to work with postdocs on AI for social sciences/human behavior, social NLP, and LLMs.
Should scientists apply to OpenAI's fund for research on AI & mental health? Should policymakers consider it a credible safety effort?
Avriel Epps & I see it as "grantwashing," and it's an insult to anyone whose loved one's death involved chatbots. We explain:
www.techpolicy.press/beware-of-op...
We're hiring interns in the Computational Social Science group at Microsoft Research NYC!
If you're interested in designing AIβbased systems and understanding their impact at both individual and societal scales, apply here by Jan 9, 2026: apply.careers.microsoft.com/careers/job/...
The Center for Information Technology Policy at Princeton invites applications for a Postdoctoral Fellow to work with Andy Guess (Politics/SPIA), Brandon Stewart (Sociology), and me (CS).
puwebp.princeton.edu/AcadHire/app...
Please apply before Sunday, the 13th of December!
US CAISI (the equivalent of the US "AI Safety Institute") just put out their approach to AI measurement & there's such a significant portion on construct validity (nist.gov/blogs/caisi-...).
Great to see this after ongoing advocating about this issue (arxiv.org/abs/2511.04703)!
Happy to share our most recent paper, led by @allisonwan.bsky.social, auditing what domains Google shows when you search for a Member of Congress.
journalqd.org/article/view...
@journalqd.bsky.social
"Captain Gains" on Capitol Hill Shang-Jin Wei & Yifan Zhou WORKING PAPER 34524 DOI 10.3386/w34524 ISSUE DATE November 2025 Using transaction-level data on US congressional stock trades, we find that lawmakers who later ascend to leadership positions perform similarly to matched peers beforehand but outperform them by 47 percentage points annually after ascension. Leaders' superior performance arises through two mechanisms. The political influence channel is reflected in higher returns when their party controls the chamber, sales of stocks preceding regulatory actions, and purchase of stocks whose firms receiving more government contracts and favorable party support on bills. The corporate access channel is reflected in stock trades that predict subsequent corporate news and greater returns on donor-owned or home-state firms.
什 1 1 -9 -8 -7 -6 -5 -4 -3 -2 -1 1 2 3 4 5 7 8 9 Year Figure 2: Estimated dynamic quasi-difference-in-differences coefficient, di, of equation(3), with vertical dashed lines representing 90 percent confidence intervals. The point estimate of the year in which the lawmaker became a congressional leader (Year 0) is normalized to zero. BHAR over the 250 days following each trade is the dependent variable and calculated using the Fama-French five-factor plus momentum as the benchmark model.
After becoming a congressional leader, a politicianβs stock portfolio beats out those of peers by 47 (!!!) percentage points a year through trades timed around bills and firms that later get government contracts
www.nber.org/papers/w34524
via @florianederer.bsky.social
Google AI overviews now reach over 2B users worldwide. But how reliable are they on high stakes topics - for instance, pregnancy and baby care?
We have a new paper - led by Desheng Hu, now accepted at @icwsm.bsky.social - exploring that and finding many issues
Preprint: arxiv.org/abs/2511.12920
π§΅π
The underlying study: "Experimental evidence of the effects of large language models versus web search on depth of learning" academic.oup.com/pnasnexus/ar...
"participants reported developing shallower knowledge from LLM summaries even when the results were augmented by...web links" vs. search
to keep functioning, a society must discourage baldfaced lying, especially by authorities
bsky.app/profile/anna...
