Nathan Bartley

Two dose-response functions. When probabilities of exposure are not homogeneous across units, we can only partially identify the expected average outcomes from the average outcomes by exposure — even when the exposure map is correctly specified; see Corollary 11.2 — as shown here using data from Cai et al. (2015). AFEOs by exposure (left) only partially identify the EAO curve (right). Lines indicate bounds on the EAO, with red lines being the bounds when outcomes are assumed to be monotonic under exposure levels. Error bars and bands are 95% confidence intervals.

I'm giving the upcoming Online Causal Inference Seminar, this Tuesday 11:30am Eastern.

I'll be talking about different dose–response functions you might want to estimate when treatment effects may spill over from one unit to another.

Tune in & ask questions!
sites.google.com/view/ocis/home

🚨New paper

Are visual tokens going into an LLM interpretable 🤔

Existing methods (e.g. logit lens) and assumptions would lead you to think “not much”...

We propose LatentLens and show that most visual tokens are interpretable across *all* layers 💡

Details 🧵

Seeing people use this to experiment w/turning Claude time from automated lecture into active learning & self-testing has brought me joy!

I added a Measure This resource to the repo, because my first thought was: how can you all get more credit for this at work??

github.com/DrCatHicks/l...

“These findings provide clear evidence that data collected on MTurk simply cannot be trusted.”

just gonna drop some links to a few papers of mine about AI-CSAM and the law here, for no particular reason

cyber.fsi.stanford.edu/news/ai-csam... (May 2025)

partnershiponai.org/hai-research... (November 2024)

www.lawfaremedia.org/article/addr... (February 2024)

Handbook of Computational Social Science cover and authors

ICYMI -- Delighted that the Handbook of Computational Social Science is finally out. Amazing cast of coauthors, with special thanks to @tahayasseri.bsky.social for leading the effort. Happy Holidays!

www.elgaronline.com/edcollbook/b...

Consider submitting to the NLP+CSS workshop, and please share the call for papers with people who might be interested!

#PolCom #CompSocSci #ComputationalSocialScience #NLP #NLProc #langsky #polisky

Here we go! @lmesseri.bsky.social & I wrote about epistemic risks of synthetic participants ('AI Surrogates') in @cp-trendscognsci.bsky.social

(see 🧵 here)
bsky.app/profile/mjcr...

Mapping the online manipulation economy A market perspective on digital manipulation may help improve online trust and safety

New paper out in @science.org! We unveil the online manipulation market with the Cambridge Online Trust & Safety Index (COTSI). We show in real time the cost of purchasing fake accounts across every social platform around the world - so they can be held accountable

www.science.org/doi/10.1126/...

Angeline Era is OUT NOW!!!
🔍Every level hidden in plain sight…
🗺️… chart your own path through the world!
⚔️Tons of thrilling combat, mysterious places…
👼… and a soul-shattering story!
GET IT HERE!!
store.steampowered.com/app/2393920/...

You’re absolutely right — you are Pagliacci. It would certainly be difficult for you to attend your own performance! I should not have given such paradoxical advice, and I apologize deeply for the error. There is no excuse for my failure.

We first note that hand-wringing about the decline in US college enrollments has mistakenly linked such declines to the price of four-year colleges.

But the decline is entirely driven by two-year community colleges (and by for-profit colleges). The four-year sector is the dog that didn't bark.

AI Surrogates and illusions of generalizability in cognitive science Recent advances in artificial intelligence (AI) have generated enthusiasm for using AI simulations of human research participants to generate new know…

Can AI simulations of human research participants advance cognitive science? In @cp-trendscognsci.bsky.social, @lmesseri.bsky.social & I analyze this vision. We show how “AI Surrogates” entrench practices that limit the generalizability of cognitive science while aspiring to do the opposite. 1/

Technology, Labor Rights, and Political Power in Kenya and Across Africa | TechPolicy.Press Justin Hendrix speaks with Nerima Wako-Ojiwa, director of Siasa Place, and Odanga Madung, a tech and society researcher and journalist.

Content moderators and data workers across Africa are building political power, especially in Kenya. @techpolicypress.bsky.social @justinhendrix.bsky.social

www.techpolicy.press/technology-l...

Project Overview Pew Research Center is seeking two Research Assistants to join its Religion research project. The Research Assistants will support the project’s demography of religion and U.S. survey research teams. The demography of religion team analyzes data on the religious identity and characteristics of populations. It produces reports describing the religious composition of countries as well as other topics at the intersection of religion and social trends (e.g. religion and educational attainment, gender differences in religious commitment levels). The U.S. survey research team produces reports that explore issues at the intersection of religion and public affairs. Through public opinion surveys and other quantitative tools, it studies religious groups in the U.S. and examines the influence of religion on politics and society. The people in these positions will work closely with other researchers on staff and will report to an Associate Director of Religion research. Primary Responsibilities Compiling and analyzing data from surveys and censuses Descriptive statistical analysis Creating tables and charts in Stata, Excel, Word and PowerPoint Contacting statistical agencies to request information and custom tables Data cleaning, maintaining and reviewing internal files and databases Number checking, fact checking, proofing Performing background research, including literature reviews, contacting subject matter experts and online data gathering Assisting with questionnaire development Coordinate the translation and checking of questionnaires

Education/Training/Experience B.A. degree is required. A concentration in sociology, political science, economics or another social science is desirable Strong quantitative skills, background in data management and analysis preferred Experience undertaking research using Stata, R, SPSS or other statistical software Mastery of descriptive statistics Knowledge and Skill Requirements Quantitative skills, including an ability to use Stata, SPSS or other statistical software Interest in public and/or international affairs, particularly in issues concerning religion and the intersection of religion and public life Strong verbal and written communications skills Facility using syntax commands to carefully document analysis using log files is desired Ability to balance numerous projects simultaneously Attention to detail, including exacting standards to maintain accuracy and impartiality in all work products Proven ability to communicate research results clearly and concisely Ability to work collaboratively and collegially with staff from the religion team, as well as with staff from other Pew Research Center projects and outside organizations Ability to balance multiple projects and meet tight deadlines while ensuring accuracy in data management, fact checking and research Ability to communicate in more than one language is advantageous

PLEASE SHARE:
We just posted an ad for TWO research assistant positions on @pewresearch.org's religion team.

We expect a lot of applications so it's best to apply this week for full consideration.

Come work with our great team!

pewtrusts.wd5.myworkdayjobs.com/en-US/Center...

Happy to contribute

Microsoft's Computational Social Science group may have the opportunity to hire one researcher

Senior: 0-3 yrs post PhD
jobs.careers.microsoft.com/global/en/jo...

Principal: 3+ yrs post PhD
jobs.careers.microsoft.com/global/en/sh...

Please note: our ability to hire this season is not certain

The Need to Make Content Moderation Transparent | TechPolicy.Press Sandra González-Bailón and David Lazer, contemplating results of a recent study on election misinformation, say Big Tech's power must be checked.

Content moderation is a power platforms exercise with consequence at moments of collective vulnerability, such as during elections. Assuming that platforms act for the betterment of all is, at this point, one assumption too many, write Sandra González-Bailón and David Lazer.

Discussing Bluesky migration in terms of liberal flight misses the simple fact that most people use social media for fun not influence and this platform, unlike X and cruddified Facebook, doesn’t feel like you’re wading through shit to reach your friends. It’s not that deep.

Screenshot of text: The months since 7 October have seen shocking attacks on freedom of expression and assembly on campus. Even before the stunning display of police brutality in recent weeks, campuses have been home to canceled speakers and events, arbitrary disciplinary hearings, and outright censorship. [Highlighting begins] The University of Southern California has canceled its entire commencement ceremony rather than let its valedictorian, Asna Tabassum, deliver a message of Palestine solidarity in her speech. [Highlighting ends] Universities have suspended chapters of Students for Justice in Palestine (SJP) and Jewish Voice for Peace, in decisions that PEN America has said are “united by a degree of opacity, in that university leaders have not been fully forthcoming in delineating how these student groups broke campus rules, or how the decision to suspend them was reached”.

Even people trying to describe how bad things are keep making the USC thing sound less bad than it was. The valedictorian hadn't written a speech at all! She had not announced any plans to say anything about Palestinians! A mob of bigots decided her religion made her offensive and USC caved to it.