Professor of Sociology, Psychology, & OB @Stanford Direct @PASCL_Stanford and the AI for Public Benefit Lab (@ai4pb_stanford)
Palo Alto, CA
Joined August 2014
- Tweets8.4K
- Following2.8K
- Followers7.6K
- Likes22.3K
Pinned Tweet
Saturday's No Kings protests were massive, inclusive, and nonviolent. It was an honor to team up with @owasow to analyze how nonviolent resistance can wins hearts, minds, and elections.
Robb Willer retweeted
Here's the under/over performance in some key Senate races, using 2024 presidential results + 2026 national environment as the baseline expectation
Robb Willer retweeted
This contentious interchange between the NY Times and @GinaHinojosaTX was truly amazing. Hinojosa offered one of the best explanations of populism I've ever seen. Her comments were a total rejection of Corporate Dems' neoliberal formula. Her Texas victory is a game changer.
Robb Willer retweeted
We find that Democratic counters have little effect and Republican counters have modest effects. But the Joe Rogan/Tucker Carlson counter almost halves following and significantly lowers issue-specific and overall Trump approval.
Robb Willer retweeted
Using the Many Authors data, Claude counts 34-40 unretracted papers in which Gino was responsible for a study and no co-author reports seeing the data, and Claude also says these papers have been cited over 1000 times since 2024.
Robb Willer retweeted
Also, here's the arXiv link: arxiv.org/abs/2610.38883
Robb Willer retweeted
Stevens made a list of demands of Abdul - denouncing not just one dumb ad from an outside group but also apologizing for other statements she didn’t like and pulling another ad that included Tucker Carlson - all to keep her endorsement while ALSO saying her endorsement was never in doubt.
Did Stevens make any demands of Rogers to earn her support?
Robb Willer retweeted
I think frontier labs have an ethical responsibility to make efforts to communicate to the public about the risks they’re taking proportionate to the scale and severity of those risks, even if they believe (and are correct about) their work improving the odds that things go well.
Robb Willer retweeted
There's a vogue for what I call the Techno-Optimist Fallacy, wrongly inferring "all instances of technological progress are good" from the true premise "technological progress has been very good on average."
There are many counter-examples.
slowboring.com/p/the-techno-…
Robb Willer retweeted
New blogpost: my lab's AI policies and norms.
I'm trying to figure out how researchers should make use of AI to accelerate discovery without letting it replace human understanding or falling off the "slop cliff." Here are some ideas.
babieslearninglanguage.blogs…
Robb Willer retweeted
The Filedrawer.org platform includes a “light pass” review layer that checks every reported estimate against the result tables and every analysis against the pre-analysis plan. I’ve also been adding integrations for automated reviewers: Coarse, @RefineDotInk, OpenAIReview, and my colleague @JunyanJiang’s new system Edyth.
If you file a study on Filedrawer, you can use Edyth for free on that report (thanks @JunyanJiang!) A neat thing about pointing automated review systems at these reports is that they can help you motivate follow-up studies.
A couple of weeks ago, @RobbWiller mentioned creating a public resource that generates automated research reports as a way of addressing publication bias.
After many rounds of tinkering, I’ve created a repository for my own site, but I'm opening it up to anyone who wants to contribute. Filedrawer.org is an agent-friendly platform that takes an existing project, creates a report, runs a light pass that checks every reported estimate against the results and every analysis against the pre-analysis plan, and suggests new research designs with downloadable QSFs so you or other scholars can extend the work. Reports can even obtain a DOI!
You host all data and scripts on GitHub, with Filedrawer.org only hosting metadata from your public repo. It checks locally for PII before you post, lets you import outside reviews, and you can use either local open-weight models or proprietary models via @OpenRouter.
I’m committing to host it, as I already have several projects there, and thanks to GitHub, hosting the metadata is very cheap. No individual-level data reach Filedrawer.org or the models: they see only the survey schema, the plan, and aggregate results.
Right now, it handles survey experiments, conjoints and methods comparisons, since that’s my wheelhouse. It’s open source, so if your work is observational, qualitative or field-based, I’d love help extending it.
Robb Willer retweeted
A couple of weeks ago, @RobbWiller mentioned creating a public resource that generates automated research reports as a way of addressing publication bias.
After many rounds of tinkering, I’ve created a repository for my own site, but I'm opening it up to anyone who wants to contribute. Filedrawer.org is an agent-friendly platform that takes an existing project, creates a report, runs a light pass that checks every reported estimate against the results and every analysis against the pre-analysis plan, and suggests new research designs with downloadable QSFs so you or other scholars can extend the work. Reports can even obtain a DOI!
You host all data and scripts on GitHub, with Filedrawer.org only hosting metadata from your public repo. It checks locally for PII before you post, lets you import outside reviews, and you can use either local open-weight models or proprietary models via @OpenRouter.
I’m committing to host it, as I already have several projects there, and thanks to GitHub, hosting the metadata is very cheap. No individual-level data reach Filedrawer.org or the models: they see only the survey schema, the plan, and aggregate results.
Right now, it handles survey experiments, conjoints and methods comparisons, since that’s my wheelhouse. It’s open source, so if your work is observational, qualitative or field-based, I’d love help extending it.
What's even more impressive than publishing your own research file drawer? Helping others do the same!
@YamilRVelez consistently finds creative ways to apply new methods to longstanding scientific problems. This is so profoundly cool.
Check it out. 👇👇👇👇
A couple of weeks ago, @RobbWiller mentioned creating a public resource that generates automated research reports as a way of addressing publication bias.
After many rounds of tinkering, I’ve created a repository for my own site, but I'm opening it up to anyone who wants to contribute. Filedrawer.org is an agent-friendly platform that takes an existing project, creates a report, runs a light pass that checks every reported estimate against the results and every analysis against the pre-analysis plan, and suggests new research designs with downloadable QSFs so you or other scholars can extend the work. Reports can even obtain a DOI!
You host all data and scripts on GitHub, with Filedrawer.org only hosting metadata from your public repo. It checks locally for PII before you post, lets you import outside reviews, and you can use either local open-weight models or proprietary models via @OpenRouter.
I’m committing to host it, as I already have several projects there, and thanks to GitHub, hosting the metadata is very cheap. No individual-level data reach Filedrawer.org or the models: they see only the survey schema, the plan, and aggregate results.
Right now, it handles survey experiments, conjoints and methods comparisons, since that’s my wheelhouse. It’s open source, so if your work is observational, qualitative or field-based, I’d love help extending it.
Robb Willer retweeted
New working paper: Can AI forecast treatment effects in economics from pre-registration documents?
Answer: Yes. 1/
Robb Willer retweeted
"Reviewers without training found 27% of the major errors, and training barely helped. We gave the same papers to 15 LLMs. They found 53-96%. Even the smallest models found about twice as many major errors as human reviewers"
Why we need human AND LLM review
Lots of debate this week on whether journals should ban AI in peer review.
We actually tested this in our new preprint, and the results were pretty clear to me.
For context, I've reviewed for most of the big journals and evaluated 1,000 human reviews as an editor. I love to read and evaluate science.
TLDR: From now on I won't do any peer review without AI. For clinical trials, I'd even call human-only review dangerous 🧵
Robb Willer retweeted
Lots of debate this week on whether journals should ban AI in peer review.
We actually tested this in our new preprint, and the results were pretty clear to me.
For context, I've reviewed for most of the big journals and evaluated 1,000 human reviews as an editor. I love to read and evaluate science.
TLDR: From now on I won't do any peer review without AI. For clinical trials, I'd even call human-only review dangerous 🧵
Robb Willer retweeted
Out today in Nature Human Behaviour: our review of large language models in computational social science, with @xu_fengli, @yongli and colleagues at @Tsinghua_Uni, @hkust & @MSFTResearch Asia. A thread on what we found, including where the field is fooling itself.
Read @ doi.org/10.1038/s41562-026-0…
