Sign in

Thomas Renault

@thomasrenault.bsky.social
88 followers 66 following 18 posts

Associate Professor in Economics. University Paris-Saclay. #SocialMedia #AI #Misinformation www.thomas-renault.com

PostsRepliesMedia
Reposted by Thomas Renault
Mohsen Mosleh @mmosleh.bsky.social · 31/08/2026
New preprint: Community Notes are good at correcting posts. But do they change the people behind them? We tracked 19,854 corrected accounts on X — 11.9M posts, 4 weeks before/after each correction. The answer: it depends who you are. @oii.ox.ac.uk 📎 arxiv.org/abs/2608.27526
arxiv.org
Community corrections have divergent downstream effects across corrected accounts
Community-based fact-checking can reduce the spread of annotated misleading posts, but whether it produces lasting behavioral change among corrected authors remains unclear. Here, we conduct a large-s...
2158
Thomas Renault @thomasrenault.bsky.social · 14/08/2026
For those interested in the broader discussion, you can also read the original Science paper here: www.science.org/doi/10.1126/..., as well as the authors’ recent response to our paper here: arxiv.org/abs/2607.28968
010
Thomas Renault @thomasrenault.bsky.social · 14/08/2026
📄 Paper, open access: www.pnas.org/doi/10.1073/... 💻 Code and replication materials: github.com/trenault/llm... With @abergeaud.bsky.social and @clementbosquet.bsky.social
pnas.org
Scientific production in the era of large language models: Outcome-triggered treatment timing and spurious event-study dynamics | PNAS
Large language models (LLMs) are increasingly used in scientific writing, but their effect on individual productivity is difficult to identify beca...
120
Thomas Renault @thomasrenault.bsky.social · 14/08/2026
Many thanks to the editor, Josh Angrist, and to the two anonymous reviewers for their comments and suggestions.
100
Thomas Renault @thomasrenault.bsky.social · 14/08/2026
Importantly, our results do not mean that LLMs have no effect on productivity. Rather, they show that we do not yet know what the true effects are, and that more research and clearer empirical designs are needed to identify how AI affects researchers’ scientific productivity.
120
Thomas Renault @thomasrenault.bsky.social · 14/08/2026
We find that similar patterns emerge with random treatments, neutral keywords, and several placebo tests. We also provide a formal proof of the mechanism and reproduce the same pattern in simulations.
100
Thomas Renault @thomasrenault.bsky.social · 14/08/2026
We show that the large increase in productivity among researchers adopting Large Language Models reported by Kusumegi et al. (2025) in Science is substantially driven by a statistical artifact in how the LLM adoption date is defined.
120
Thomas Renault @thomasrenault.bsky.social · 14/08/2026
🚨Our new paper on generative AI and researcher productivity is now out in @pnas.org A 🧵 www.pnas.org/doi/10.1073/...
192
Reposted by Thomas Renault
David Rand @dgrand.bsky.social · 01/08/2026
🚨JOIN the Civic Discourse Mediator Competition🚨 In 1st week of August, create & test a prompt-based AI mediator that helps people have more thoughtful, productive conversations about important public issues. Win $400! Get started here: forms.gle/HkbhEk4hELQE...
034
Thomas Renault @thomasrenault.bsky.social · 19/05/2026
On the effect of LLM adoption on researcher productivity, you might find the comment we released today interesting: arxiv.org/abs/2605.17979
010
Thomas Renault @thomasrenault.bsky.social · 19/05/2026
We discuss the mechanism in detail in the paper. 📄 Paper: arxiv.org/abs/2605.17979 💻 Code and replication materials: github.com/trenault/llm... Comments and feedback are very welcome. With @abergeaud.bsky.social and Clément Bosquet
arxiv.org
Comment on Scientific production in the era of large language models
Kusumegi et al. (2025) study whether researchers' preprint output rises after adopting large language models (LLMs), dating adoption as the first month in which at least one submitted abstract exceeds...
030
Thomas Renault @thomasrenault.bsky.social · 19/05/2026
The core issue is that the treatment variable (being classified as an LLM adopter) is mechanically related to researchers’ publication volume: the more papers an author publishes, the more likely they are to be classified as treated.
171
Thomas Renault @thomasrenault.bsky.social · 19/05/2026
Similar “LLM effects” can be reproduced when replacing “LLM adoption” (detected from abstract content) with random keywords, random treatment assignment, or even publication periods predating the release of ChatGPT.
130
Thomas Renault @thomasrenault.bsky.social · 19/05/2026
🚨 A few months ago, a paper published in Science reported that “LLM adoption is associated with a large increase in researchers’ scientific output.” In a comment released today, we show that the reported effects are driven by a methodological issue in the empirical design. arxiv.org/abs/2605.17979
110235
Reposted by Thomas Renault
Alexios Mantzarlis @mantzarlis.com · 12/05/2026
Hello -- I'm looking for active X Community Notes contributors who want to talk to me for a piece on @indicator.media. Happy to use their pseudonym in the story. Please share with your networks!
014
Reposted by Thomas Renault
Matthew Facciani @matthewfacciani.bsky.social · 08/05/2026
New study finds that Community Notes on Twitter/X reduced reposting of misleading posts by about 61% after the note became visible, meaning users were much less likely to reshare content once they saw a fact-check. www.nature.com/articles/s41...
nature.com
Community-based fact-checking reduces the spread of misleading posts on X (formerly Twitter) - Nature Communications
Community-based fact-checking is increasingly adopted by social media platforms, but its real-world impact remains unclear. Here, the authors show that community notes can reduce the spread of mislead...
23713
Reposted by Thomas Renault
Jay Van Bavel, PhD @jayvanbavel.bsky.social · 14/05/2026
Community-based fact-checking works Community notes reducethe subsequent spread of misleading posts by 61% and increase the odds that users delete their false posts by 94% But they appear too late to stop the viral stage of the diffusion making the effect modest. www.nature.com/articles/s41...
22914
Reposted by Thomas Renault
CEPII @cepii-paris.bsky.social · 23/04/2026
🆕La Lettre du CEPII "L’émergence de l’extrême droite façonne-t-elle le positionnement des partis traditionnels sur la question migratoire?" d'Anthony Edo, @thomasrenault.bsky.social et Jérôme Valette. www.cepii.fr/CEPII/fr/pub... #extremedroite #migration #immigration #econsky #elections
111
Reposted by Thomas Renault
Alexios Mantzarlis @mantzarlis.com · 11/03/2026
NEW on @indicator.media: We launched a tracker to monitor who gets the most "@grok-is-this-true" replies on a given day on X. indicator.media/p/who-gets-g...
283
Reposted by Thomas Renault
David Rand @dgrand.bsky.social · 04/02/2026
Grok fact-checks our paper on Grok fact-checking - and it approves!
1297
Reposted by Thomas Renault
David Rand @dgrand.bsky.social · 03/02/2026
🚨New WP "@Grok is this true?" We analyze 1.6M factcheck requests on X (grok & Perplexity) 📌Usage is polarized, Grok users more likely to be Reps 📌BUT Rep posts rated as false more often—even by Grok 📌Bot agreement with factchecks is OK but not great; APIs match fact-checkers osf.io/preprints/ps...
211747
Reposted by Thomas Renault
Alexios Mantzarlis @mantzarlis.com · 28/01/2026
In a new working paper, @thomasrenault.bsky.social, @mmosleh.bsky.social & @dgrand.bsky.social break down how Grok is being used to fact-check, and on whom. osf.io/preprints/ps...
172
Reposted by Thomas Renault
Alexios Mantzarlis @mantzarlis.com · 28/01/2026
New on @indicator.media: "@grok is this true" was the single most frequent reply tagging X's AI chatbot in the six months following its launch.
indicator.media
@Grok is this true: How X’s chatbot performs as a fact-checking tool
New research explores whether the chatbot might replace the crowdsourced fact-checking program – and what that might mean for getting to the truth on X
22910
Reposted by Thomas Renault
Martin Anota @manota.bsky.social · 21/12/2025
Comment l’irruption du FN a bouleversé le jeu politique (A la marge) blogs.alternatives-economiques.fr/anota/2025/1...
blogs.alternatives-economiques.fr
Comment l’irruption du FN a bouleversé le jeu politique
L’extrême-droite a eu une place extrêmement marginale dans le paysage politique des décennies qui ont immédiatement suivi la Seconde Guerre mondiale. Celui-ci était dominé par deux blocs, un bloc de g...
064
Reposted by Thomas Renault
andyguess @andyguess.com · 09/12/2025
The Center for Information Technology Policy at Princeton invites applications for a Postdoctoral Fellow to work with an interdisciplinary team (me, @bstewart.bsky.social, and @manoelhortaribeiro.bsky.social). Link: puwebp.princeton.edu/AcadHire/app... Please apply by THIS SUNDAY, Dec. 14!
We are seeking a Fellow to lead cutting-edge research on short-form video content and its societal implications.

Bridging the gap between computer vision, causal inference, and computational social science, the Fellow will focus on the large-scale analysis of TikTok data (using an existing, massive dataset).

The position involves developing novel multimodal methods to understand how short-form algorithmic content shapes public opinion, political polarization, and online culture.

This is a unique opportunity to work at the forefront of the field, combining rigorous social science research designs with state-of-the-art computational techniques.
02218
Reposted by Thomas Renault
David Rand @dgrand.bsky.social · 15/08/2025
Apply for funding to work with me and Gord at Cornell!
03114
Thomas Renault @thomasrenault.bsky.social · 14/08/2025
Would love to read more. Is there a full paper ? Or full version of the poster ? Thx
000
Reposted by Thomas Renault
Alexios Mantzarlis @mantzarlis.com · 17/06/2025
3 months into Meta’s pivot to Community Notes, the company has yet to share any data about its crowdsourced moderation feature. With help from three @indicator.media readers in the pilot, I tried to piece together how it’s going: indicator.media/p/a-first-lo...
indicator.media
A first look at Meta’s Community Notes
Zuckerberg's pivot to crowdsourcing features three jacked dudes, a flirty farmer, and some useful notes
12515
Reposted by Thomas Renault
David Rand @dgrand.bsky.social · 17/06/2025
🚨In PNAS🚨 The right often accuses fact-checkers of political bias But we analyzed Community Notes on Musk's X and found posts flagged as "misleading" are 2.3x more likely to be written by Reps than Dems! The issue is Reps sharing misinformation, not fact-checker bias... www.pnas.org/doi/10.1073/...
7344129
Thomas Renault @thomasrenault.bsky.social · 17/06/2025
With @dgrand.bsky.social and @mmosleh.bsky.social 🔗 Read the Brief Report (Open Access) here: www.pnas.org/doi/10.1073/...
pnas.org
PNAS
Proceedings of the National Academy of Sciences (PNAS), a peer reviewed journal of the National Academy of Sciences (NAS) - an authoritative source of high-impact, original research that broadly spans...
021
Thomas Renault @thomasrenault.bsky.social · 17/06/2025
— We find a real, measurable partisan gap in misinformation sharing : even a “crowdsourced,” supposedly neutral system ends up flagging more Republican posts — This challenges arguments that fact-checkers should be dismissed on the grounds that they’re too biased.
112
Thomas Renault @thomasrenault.bsky.social · 17/06/2025
We analyzed every English Community Note on X over 1.5 years. In Community Notes: — Users from across the political spectrum propose and rate notes — Notes are only published when there is agreement between people who typically disagree (via X’s “bridging algorithm”)
100
Thomas Renault @thomasrenault.bsky.social · 17/06/2025
💡 Why is this important? Many past studies have found that Republicans share more misinformation. But critics argue those findings could be driven by : — Bias in how fact-checkers or academics define misinformation — Bias in which news stories researchers choose to study
100
Thomas Renault @thomasrenault.bsky.social · 17/06/2025
🚨 New in PNAS 🚨 Posts by Republicans are 2.3 times more likely to be flagged as misleading than those by Democrats on X's Community Notes. A 🧵 pnas.org/doi/epub/10.... (Open Access)
176
Reposted by Thomas Renault
Laura Bronner @laurabronner.bsky.social · 07/02/2025
Interested in how people talk to each other online? Care about causal inference and/or NLP? Want to design and implement field experiments? Come do a PhD with Dominik Hangartner, me, and a bunch of awesome people at IPL in Zurich: jobs.ethz.ch/job/view/JOP...
0109
Reposted by Thomas Renault
David Rand @dgrand.bsky.social · 31/01/2025
Many thanks to lead author @thomasrenault.bsky.social who does amazing work on Community Notes and other topics; and the always-wonderful coauthor @mmosleh.bsky.social For more of my group's work on misinformation, check out this doc: docs.google.com/document/d/1...
docs.google.com
Misinformation-related papers
Papers related to misinformation from David Rand and Gordon Pennycook’s research team Key papers The Psychology of Fake News TiCS 2021 [Tweet thread] [15 minute video summary] Durably reducing conspi...
194
Reposted by Thomas Renault
David Rand @dgrand.bsky.social · 31/01/2025
We examine all English notes 1/23-6/24. More notes are proposed on tweets written by Reps than Dems. The partisan diff is MUCH bigger when restricting to "helpful" (ie ~unbiased) notes: 63% on Reps, 37% on Dems. The "vox populi" on X has spoken and concluded that Reps share more misinfo than Dems!
141
Reposted by Thomas Renault
David Rand @dgrand.bsky.social · 31/01/2025
Accusations of bias against conservatives drove Musk to buy Twitter- then gut fact-checking and up Community Notes. This month, Zuckerberg did the same at Meta. But greater sanctioning of conservatives could just be the result of conservatives sharing more misinformation bsky.app/profile/dgra...
2315
Reposted by Thomas Renault
David Rand @dgrand.bsky.social · 31/01/2025
🚨New WP🚨 Remember Musk+Zuck+Trump+Jordan etc crying fact-checker bias b/c Reps were flagged more than Dems? We analyzed Community Notes on Musk's X and guess what: posts flagged as "misleading" are 67% more likely to be written by Reps! The issue is Reps, not fact-checkers... osf.io/preprints/ps...
6453121
Reposted by Thomas Renault
David Rand @dgrand.bsky.social · 28/01/2025
🚨OpEd+data: Meta is out of step with public opinion🚨 Zuck cut moderation b/c he said people no longer want it. But he's wrong! We polled 1k Americans and most people, including majority of Reps: i) want content moderation ii) don't want Community Notes w/o fact-checkers thehill.com/opinion/tech...
623380