Thomas Renault
Current Organizations / Institutes
The organizations/institutes from the scholar's latest article
Previous Organizations / Institutes
The organizations / institutes from the scholar's other articles in the past ten years
Research Subjects
Thomas Renault has a total of 37 co-authors. Here are the top co-authors by number of shared publications:
- David Gertler Rand
Massachusetts Institute of Technology
4 shared publications - Mohsen Mosleh
Oxford Internet Institute, University of Oxford, Oxford, Oxfordshire, United Kingdom, and Sloan School of Management, Massachusetts Institute of Technology, Cambridge, MA, USA
4 shared publications Brent H. Meyer
Federal Reserve Banks - Federal Reserve Bank of Atlanta
2 shared publicationsDavid Altig
Federal Reserve Banks - Federal Reserve Bank of Atlanta; Federal Reserve Bank of Cleveland; University of Chicago - Booth School of Business
2 shared publications- David Restrepo Amariles
HEC Paris
2 shared publications Emil Mihaylov
Federal Reserve Banks - Federal Reserve Bank of Dallas
2 shared publications
Latest Publications
- Open Access
- 2 November 2026
- Published by Center for Open Science in PsyArXiv
Can AI Scale Human Fact-Checking? Evidence from Community Notes
Community Notes---a crowd-based fact-checking program through which social media users can flag a post with a note explaining why the post is misleading, and the note is displayed if a diverse set of users rate the note as helpful---has become central to online content moderation. The program, however, faces substantial scalability constraints, as the generation of notes takes considerable time and effort, and the number of user participants is limited. Here, we investigate whether allowing AI models to write Community Notes helps to alleviate these constraints by analyzing the full corpus of AI-generated Community Notes on X through May 2026. Encouragingly, AI-generated notes are submitted more quickly, are longer, and cite more and higher-quality sources than human-written notes, while being rated helpful at a similar rate and being associated with similar decreases in reposts following the publication of a rated-helpful note. Furthermore, a pre-registered survey experiment finds no statistically detectable difference between AI- and human-generated fact-checks in their effect on belief updating, regardless of whether the author was revealed to participants. On the other hand, AI-generated note writing is highly concentrated, as a handful of accounts produce the large majority of notes, and are significantly less likely to engage with posts involving controversial and polarized topics such as politics and elections. While those findings suggest that AI may be a powerful tool for scaling crowd-based fact-checking systems such as Community Notes, this evolution also entails important risks: shifting the primary bottleneck from note production to note evaluation, moving the system from a decentralized toward a more centralized model, and creating incentives to prioritize claims that are more likely to achieve consensus over those of greater societal importance.
- 2 November 2026
- Published by Center for Open Science in PsyArXiv
Open AccessCan AI Scale Human Fact-Checking? Evidence from Community Notes
Community Notes---a crowd-based fact-checking program through which social media users can flag a post with a note explaining why the post is misleading, and the note is displayed if a diverse set of users rate the note as helpful---has become central to online content moderation. The program, however, faces substantial scalability constraints, as the generation of notes takes considerable time and effort, and the number of user participants is limited. Here, we investigate whether allowing AI models to write Community Notes helps to alleviate these constraints by analyzing the full corpus of AI-generated Community Notes on X through May 2026. Encouragingly, AI-generated notes are submitted more quickly, are longer, and cite more and higher-quality sources than human-written notes, while being rated helpful at a similar rate and being associated with similar decreases in reposts following the publication of a rated-helpful note. Furthermore, a pre-registered survey experiment finds no statistically detectable difference between AI- and human-generated fact-checks in their effect on belief updating, regardless of whether the author was revealed to participants. On the other hand, AI-generated note writing is highly concentrated, as a handful of accounts produce the large majority of notes, and are significantly less likely to engage with posts involving controversial and polarized topics such as politics and elections. While those findings suggest that AI may be a powerful tool for scaling crowd-based fact-checking systems such as Community Notes, this evolution also entails important risks: shifting the primary bottleneck from note production to note evaluation, moving the system from a decentralized toward a more centralized model, and creating incentives to prioritize claims that are more likely to achieve consensus over those of greater societal importance.
- Open Access
- 30 September 2026
- Published by Center for Open Science in PsyArXiv
Fact-Checking Using AI
Large language models (LLMs) can both correct inaccurate beliefs (1, 2, 3) and manipulate attitudes (4, 5, 6). As LLMs are increasingly deployed as fact-checkers on social media, their impact depends not only on the models’ technical capability, but also on the choices of their designers and the attitudes and behaviors of platform users. Here, we shed light on all three dimensions of this sociotechnical system. We analyze an exhaustive sample of 1,555,281 English-language fact-check requests sent to Grok and Perplexity bots on X from February through September 2025 and a random sample of 184,124 fact-checks from Grok from September 2025 to August 2026. We also benchmark 19 frontier LLMs against professional fact-checkers’ evaluations of hundreds of news headlines, and conduct a preregistered survey experiment (N=1,592) to evaluate correction efficacy. On the user side, Grok requesters skew heavily Republican; familiarity with and trust in specific LLMs vary substantially, such that trust in AI is no longer a unitary construct; and revealing that a fact-check comes from Grok reduces belief updating among Democrats while increasing belief updating among Republicans. Examining LLM outputs, we find that from a technical perspective, LLM fact-checks can match or exceed agreement among professional fact-checkers. Yet we also show that this capacity is readily undermined by designers’ choices: for example, restricting sources (e.g. Truth Social AI) or disabling search can degrade accuracy. Designers and users, more so than technical capability, are now the primary binding constraints.
- 30 September 2026
- Published by Center for Open Science in PsyArXiv
Open AccessFact-Checking Using AI
Large language models (LLMs) can both correct inaccurate beliefs (1, 2, 3) and manipulate attitudes (4, 5, 6). As LLMs are increasingly deployed as fact-checkers on social media, their impact depends not only on the models’ technical capability, but also on the choices of their designers and the attitudes and behaviors of platform users. Here, we shed light on all three dimensions of this sociotechnical system. We analyze an exhaustive sample of 1,555,281 English-language fact-check requests sent to Grok and Perplexity bots on X from February through September 2025 and a random sample of 184,124 fact-checks from Grok from September 2025 to August 2026. We also benchmark 19 frontier LLMs against professional fact-checkers’ evaluations of hundreds of news headlines, and conduct a preregistered survey experiment (N=1,592) to evaluate correction efficacy. On the user side, Grok requesters skew heavily Republican; familiarity with and trust in specific LLMs vary substantially, such that trust in AI is no longer a unitary construct; and revealing that a fact-check comes from Grok reduces belief updating among Democrats while increasing belief updating among Republicans. Examining LLM outputs, we find that from a technical perspective, LLM fact-checks can match or exceed agreement among professional fact-checkers. Yet we also show that this capacity is readily undermined by designers’ choices: for example, restricting sources (e.g. Truth Social AI) or disabling search can degrade accuracy. Designers and users, more so than technical capability, are now the primary binding constraints.
- Open Access
- 11 August 2026
- Published by National Academy of Sciences in Proceedings of the National Academy of Sciences
- Vol. 123 (33)
Scientific production in the era of large language models: Outcome-triggered treatment timing and spurious event-study dynamics
Large language models (LLMs) are increasingly used in scientific writing, but their effect on individual productivity is difficult to identify because adoption is rarely directly observed. [K. Kusumegi et al. , Science 390 , 1240–1243 (2025)] infer adoption from the first paper detected as LLM-assisted and report large productivity gains after adoption. We show that this treatment-timing rule mechanically generates positive event-study dynamics even in the absence of any causal effect. Because high-output months are more likely to produce a detected paper, treatment assignment becomes intrinsically linked to productivity. Using reconstructed arXiv data, we show that random treatment assignments, neutral-keyword triggers, inverted treatment, and pre-ChatGPT placebo periods all generate similar dynamics. Simulations with no treatment effect also reproduce the same posttreatment patterns reported in K. Kusumegi et al. , Science 390 , 1240–1243 (2025). Our results demonstrate that first-detection timing alone creates spurious evidence of productivity gains from LLM adoption.
- 11 August 2026
- Published by National Academy of Sciences in Proceedings of the National Academy of Sciences
- Vol. 123 (33)
Open AccessScientific production in the era of large language models: Outcome-triggered treatment timing and spurious event-study dynamics
Large language models (LLMs) are increasingly used in scientific writing, but their effect on individual productivity is difficult to identify because adoption is rarely directly observed. [K. Kusumegi et al. , Science 390 , 1240–1243 (2025)] infer adoption from the first paper detected as LLM-assisted and report large productivity gains after adoption. We show that this treatment-timing rule mechanically generates positive event-study dynamics even in the absence of any causal effect. Because high-output months are more likely to produce a detected paper, treatment assignment becomes intrinsically linked to productivity. Using reconstructed arXiv data, we show that random treatment assignments, neutral-keyword triggers, inverted treatment, and pre-ChatGPT placebo periods all generate similar dynamics. Simulations with no treatment effect also reproduce the same posttreatment patterns reported in K. Kusumegi et al. , Science 390 , 1240–1243 (2025). Our results demonstrate that first-detection timing alone creates spurious evidence of productivity gains from LLM adoption.
- Open Access
- 2 August 2026
- Published by Elsevier in SSRN Electronic Journal
A Multilingual LLM-Based News Indicator of Euro Area Inflationary Pressures
We construct a novel multilingual news-based indicator of inflationary pressures in the euro area. The indicator has a positive and significant impact on households' inflation expectations, even after controlling for market- and survey-based measures. Newspapers are thus shown to contain early information about euro-area households' inflation expectations. Using a large language model, we score more than one million newspaper articles in French, German, Italian, and Spanish on a six-class directional scale, simultaneously filtering irrelevant content and measuring price pressures. We then fine-tune a multilingual encoder to enable scalable annotation of a 17-million-article corpus (2006-2024). Validation against a human-annotated benchmark for France yields a correlation of 83%. To facilitate replication and future research, we make both the trained classifier and the complete annotation pipeline publicly available.
- 2 August 2026
- Published by Elsevier in SSRN Electronic Journal
Open AccessA Multilingual LLM-Based News Indicator of Euro Area Inflationary Pressures
We construct a novel multilingual news-based indicator of inflationary pressures in the euro area. The indicator has a positive and significant impact on households' inflation expectations, even after controlling for market- and survey-based measures. Newspapers are thus shown to contain early information about euro-area households' inflation expectations. Using a large language model, we score more than one million newspaper articles in French, German, Italian, and Spanish on a six-class directional scale, simultaneously filtering irrelevant content and measuring price pressures. We then fine-tune a multilingual encoder to enable scalable annotation of a 17-million-article corpus (2006-2024). Validation against a human-annotated benchmark for France yields a correlation of 83%. To facilitate replication and future research, we make both the trained classifier and the complete annotation pipeline publicly available.
- 3 April 2026
- Published by Agence Bibliographique de l'Enseignement Supérieur
Three essays on the informational efficiency of financial markets through the use of Big Data Analytics
Trois essais sur l'efficience informationnelle des marchés financiers : une approche big data: L’augmentation massive du volume de données générées chaque jour par les individus sur Internet offre aux chercheurs la possibilité d’aborder la question de la prédictibilité des marchés financiers sous un nouvel angle. Sans prétendre apporter une réponse définitive au débat entre les partisans de l’efficience des marchés et les chercheurs en finance comportementale, cette thèse vise à améliorer notre compréhension du processus de formation des prix sur les marchés financiers grâce à une approche Big Data. Plus précisément, cette thèse porte sur (1) la mesure du sentiment des investisseurs à fréquence intra-journalière, et le lien entre le sentiment des investisseurs et les rendements agrégés du marché,(2) la mesure de l’attention des investisseurs aux informations économiques et financières en temps réel, et la relation entre l’attention des investisseurs et la dynamique des prix des actions des sociétés à forte capitalisation, et enfin, (3) la détection des comportements suspicieux pouvant amoindrir le rôle informationnel des marchés financiers, et le lien entre le volume d’activité sur les réseaux sociaux et le prix des actions des entreprises de petite capitalisation. Le premier essai propose une méthodologie permettant de construire un nouvel indicateur du sentiment des investisseurs en analysant le contenu des messages publiés sur le réseau social Stock-Twits. En examinant les caractéristiques propres à chaque utilisateur (niveau d’expérience, approche d’investissement, période de détention), cet essai fournit des preuves empiriques montrant que le comportement des investisseurs naïfs, sujets à des périodes d’excès d’optimisme ou de pessimisme, a un impact sur la valorisation du marché action, et ce en accord avec les théories de la finance comportementale. Le deuxième essai propose une méthodologie permettant de mesurer l’attention des investisseurs aux informations en temps réel, en combinant les données des médias traditionnels avec le contenu des messages envoyés par une liste d’experts sur la plateforme Twitter. Cet essai démontre que lorsqu’une information attire l’attention des investisseurs, les mouvements de marchés sont caractérisés par une forte hausse des volumes échangés, une hausse de la volatilité et des sauts de prix. Cet essai démontre également qu’il n’y a pas de fuite d’information significative lorsque les sources d’informations sont combinées pour corriger un potentiel problème d’horodatage. Le troisième essai étudie le risque de manipulation informationnelle en examinant un nouveau jeu de données de messages publiés sur Twitter à propos des entreprises de petite capitalisation. Cet essai propose une nouvelle méthodologie permettant d’identifier les comportements anormaux de manière automatisée en analysant les interactions entre les utilisateurs. Étant donné le grand nombre de recommandations suspicieuses d’achat envoyées par certains groupes d’utilisateurs, l’analyse empirique et les conclusions de cet essai soulignent la nécessité d’un plus grand contrôle par les régulateurs de l’information publiée sur les réseaux sociaux ainsi que l’utilité d’une meilleure éducation des investisseurs individuels.
- 3 April 2026
- Published by Agence Bibliographique de l'Enseignement Supérieur
Three essays on the informational efficiency of financial markets through the use of Big Data Analytics
Trois essais sur l'efficience informationnelle des marchés financiers : une approche big data: L’augmentation massive du volume de données générées chaque jour par les individus sur Internet offre aux chercheurs la possibilité d’aborder la question de la prédictibilité des marchés financiers sous un nouvel angle. Sans prétendre apporter une réponse définitive au débat entre les partisans de l’efficience des marchés et les chercheurs en finance comportementale, cette thèse vise à améliorer notre compréhension du processus de formation des prix sur les marchés financiers grâce à une approche Big Data. Plus précisément, cette thèse porte sur (1) la mesure du sentiment des investisseurs à fréquence intra-journalière, et le lien entre le sentiment des investisseurs et les rendements agrégés du marché,(2) la mesure de l’attention des investisseurs aux informations économiques et financières en temps réel, et la relation entre l’attention des investisseurs et la dynamique des prix des actions des sociétés à forte capitalisation, et enfin, (3) la détection des comportements suspicieux pouvant amoindrir le rôle informationnel des marchés financiers, et le lien entre le volume d’activité sur les réseaux sociaux et le prix des actions des entreprises de petite capitalisation. Le premier essai propose une méthodologie permettant de construire un nouvel indicateur du sentiment des investisseurs en analysant le contenu des messages publiés sur le réseau social Stock-Twits. En examinant les caractéristiques propres à chaque utilisateur (niveau d’expérience, approche d’investissement, période de détention), cet essai fournit des preuves empiriques montrant que le comportement des investisseurs naïfs, sujets à des périodes d’excès d’optimisme ou de pessimisme, a un impact sur la valorisation du marché action, et ce en accord avec les théories de la finance comportementale. Le deuxième essai propose une méthodologie permettant de mesurer l’attention des investisseurs aux informations en temps réel, en combinant les données des médias traditionnels avec le contenu des messages envoyés par une liste d’experts sur la plateforme Twitter. Cet essai démontre que lorsqu’une information attire l’attention des investisseurs, les mouvements de marchés sont caractérisés par une forte hausse des volumes échangés, une hausse de la volatilité et des sauts de prix. Cet essai démontre également qu’il n’y a pas de fuite d’information significative lorsque les sources d’informations sont combinées pour corriger un potentiel problème d’horodatage. Le troisième essai étudie le risque de manipulation informationnelle en examinant un nouveau jeu de données de messages publiés sur Twitter à propos des entreprises de petite capitalisation. Cet essai propose une nouvelle méthodologie permettant d’identifier les comportements anormaux de manière automatisée en analysant les interactions entre les utilisateurs. Étant donné le grand nombre de recommandations suspicieuses d’achat envoyées par certains groupes d’utilisateurs, l’analyse empirique et les conclusions de cet essai soulignent la nécessité d’un plus grand contrôle par les régulateurs de l’information publiée sur les réseaux sociaux ainsi que l’utilité d’une meilleure éducation des investisseurs individuels.
- 1 September 2025
- Published by Wiley in Annals of the New York Academy of Sciences
- Vol. 1552 (1) , 140-147
Social media and suicide: Empirical evidence from the quasi‐exogenous geographical adoption of Twitter
Social media usage is often cited as a potential driver behind the rising suicide rates. However, distinguishing the causal effect—whether social media increases the risk of suicide—from reverse causality, where individuals already at higher risk of suicide are more likely to use social media, remains a significant challenge. In this paper, we use an instrumental variable approach to study the quasi‐exogenous geographical adoption of Twitter and its causal relationship with suicide rates. Our analysis first demonstrates that Twitter's geographical adoption was driven by the presence of certain users at the 2007 South by Southwest (SXSW) festival, which led to long‐term disparities in adoption rates across counties in the United States. Then, using a two‐stage least squares regression and controlling for a wide range of geographic, socioeconomic, and demographic factors, we find no significant relationship between Twitter adoption and suicide rates.
- 1 September 2025
- Published by Wiley in Annals of the New York Academy of Sciences
- Vol. 1552 (1) , 140-147
Social media and suicide: Empirical evidence from the quasi‐exogenous geographical adoption of Twitter
Social media usage is often cited as a potential driver behind the rising suicide rates. However, distinguishing the causal effect—whether social media increases the risk of suicide—from reverse causality, where individuals already at higher risk of suicide are more likely to use social media, remains a significant challenge. In this paper, we use an instrumental variable approach to study the quasi‐exogenous geographical adoption of Twitter and its causal relationship with suicide rates. Our analysis first demonstrates that Twitter's geographical adoption was driven by the presence of certain users at the 2007 South by Southwest (SXSW) festival, which led to long‐term disparities in adoption rates across counties in the United States. Then, using a two‐stage least squares regression and controlling for a wide range of geographic, socioeconomic, and demographic factors, we find no significant relationship between Twitter adoption and suicide rates.
- Top Cited PapersOpen Access
- 16 June 2025
- Published by National Academy of Sciences in Proceedings of the National Academy of Sciences
- Vol. 122 (25)
Republicans are flagged more often than Democrats for sharing misinformation on X’s Community Notes
We use crowd-sourced assessments from X’s Community Notes program to examine whether there are partisan differences in the sharing of misleading information. Unlike previous studies, misleadingness here is determined by agreement across a diverse community of platform users, rather than by fact-checkers. We find that 2.3 times more posts by Republicans are flagged as misleading compared to posts by Democrats. These results are not base rate artifacts, as we find no meaningful overrepresentation of Republicans among X users. Our findings provide strong evidence of a partisan asymmetry in misinformation sharing which cannot be attributed to political bias on the part of raters, and indicate that Republicans will be sanctioned more than Democrats even if platforms transition from professional fact-checking to Community Notes.
- 16 June 2025
- Published by National Academy of Sciences in Proceedings of the National Academy of Sciences
- Vol. 122 (25)
Top Cited PapersOpen AccessRepublicans are flagged more often than Democrats for sharing misinformation on X’s Community Notes
We use crowd-sourced assessments from X’s Community Notes program to examine whether there are partisan differences in the sharing of misleading information. Unlike previous studies, misleadingness here is determined by agreement across a diverse community of platform users, rather than by fact-checkers. We find that 2.3 times more posts by Republicans are flagged as misleading compared to posts by Democrats. These results are not base rate artifacts, as we find no meaningful overrepresentation of Republicans among X users. Our findings provide strong evidence of a partisan asymmetry in misinformation sharing which cannot be attributed to political bias on the part of raters, and indicate that Republicans will be sanctioned more than Democrats even if platforms transition from professional fact-checking to Community Notes.
- Open Access
- 4 May 2025
- Published by Center for Open Science in PsyArXiv
Republicans are flagged more often than Democrats for sharing misinformation on X’s Community Notes
We use crowd-sourced assessments from X’s Community Notes program to examine whether there are partisan differences in the sharing of misleading information. Unlike previous studies, misleadingness here is determined by agreement across a diverse community of platform users, rather than by fact-checkers. We find that 131% more posts by Republicans are flagged as misleading compared to posts by Democrats. These results are not base rate artifacts, as we find no meaningful over-representation of Republicans among X users. Our findings provide strong evidence of a partisan asymmetry in misinformation sharing which cannot be attributed to political bias on the part of raters, and indicate that Republicans will be sanctioned more than Democrats even if platforms transition from professional fact-checking to Community Notes.
- 4 May 2025
- Published by Center for Open Science in PsyArXiv
Open AccessRepublicans are flagged more often than Democrats for sharing misinformation on X’s Community Notes
We use crowd-sourced assessments from X’s Community Notes program to examine whether there are partisan differences in the sharing of misleading information. Unlike previous studies, misleadingness here is determined by agreement across a diverse community of platform users, rather than by fact-checkers. We find that 131% more posts by Republicans are flagged as misleading compared to posts by Democrats. These results are not base rate artifacts, as we find no meaningful over-representation of Republicans among X users. Our findings provide strong evidence of a partisan asymmetry in misinformation sharing which cannot be attributed to political bias on the part of raters, and indicate that Republicans will be sanctioned more than Democrats even if platforms transition from professional fact-checking to Community Notes.
- Open Access
Collaboratively adding context to social media posts reduces the sharing of false news
We build a novel database of around 285,000 notes from the Twitter Community Notes program to analyze the causal influence of appending contextual information to potentially misleading posts on their dissemination. Employing a difference in difference design, our findings reveal that adding context below a tweet reduces the number of retweets by almost half. A significant, albeit smaller, effect is observed when focusing on the number of replies or quotes. Community Notes also increase by 80% the probability that a tweet is deleted by its creator. The post-treatment impact is substantial, but the overall effect on tweet virality is contingent upon the timing of the contextual information's publication. Our research concludes that, although crowdsourced fact-checking is effective, its current speed may not be adequate to substantially reduce the dissemination of misleading information on social media.
Open AccessCollaboratively adding context to social media posts reduces the sharing of false news
We build a novel database of around 285,000 notes from the Twitter Community Notes program to analyze the causal influence of appending contextual information to potentially misleading posts on their dissemination. Employing a difference in difference design, our findings reveal that adding context below a tweet reduces the number of retweets by almost half. A significant, albeit smaller, effect is observed when focusing on the number of replies or quotes. Community Notes also increase by 80% the probability that a tweet is deleted by its creator. The post-treatment impact is substantial, but the overall effect on tweet virality is contingent upon the timing of the contextual information's publication. Our research concludes that, although crowdsourced fact-checking is effective, its current speed may not be adequate to substantially reduce the dissemination of misleading information on social media.
- Open Access
- 1 January 2024
- Published by Elsevier in SSRN Electronic Journal
Collaboratively adding context to social media posts reduces the sharing of false news.
We build a novel database of around 285,000 notes from the Twitter Community Notes program to analyze the causal influence of appending contextual information to potentially misleading posts on their dissemination. Employing a difference in difference design, our findings reveal that adding context below a tweet reduces the number of retweets by almost half. A significant, albeit smaller, effect is observed when focusing on the number of replies or quotes. Community Notes also increase by 80% the probability that a tweet is deleted by its creator. The post-treatment impact is substantial, but the overall effect on tweet virality is contingent upon the timing of the contextual information’s publication. Our research concludes that, although crowdsourced fact-checking is effective, its current speed may not be adequate to substantially reduce the dissemination of misleading information on social media.
- 1 January 2024
- Published by Elsevier in SSRN Electronic Journal
Open AccessCollaboratively adding context to social media posts reduces the sharing of false news.
We build a novel database of around 285,000 notes from the Twitter Community Notes program to analyze the causal influence of appending contextual information to potentially misleading posts on their dissemination. Employing a difference in difference design, our findings reveal that adding context below a tweet reduces the number of retweets by almost half. A significant, albeit smaller, effect is observed when focusing on the number of replies or quotes. Community Notes also increase by 80% the probability that a tweet is deleted by its creator. The post-treatment impact is substantial, but the overall effect on tweet virality is contingent upon the timing of the contextual information’s publication. Our research concludes that, although crowdsourced fact-checking is effective, its current speed may not be adequate to substantially reduce the dissemination of misleading information on social media.