Recherche

L'équipe de Sciences Po :

• Jan Rovny, CEE (Principal Investigator)
• Caterina Froio, CEE
• Jean-Philippe Cointet, médialab
• Romain Lachat, CEVIPOF
• Elena Cossu, CEE

Partenaires : 

Ce projet rassemble huit partenaires de toute l'Europe, coordonnés par le CEU Democracy Institute (Hongrie) : l’Université d'Oxford (Royaume-Uni), Sciences Po (France), l’université Charles (République tchèque), la Scuola Normale Superiore (Italie), l’université SWPS (Pologne), The Transatlantic Foundation (Belgique) et l’université de Vienne (Autriche).

Période : 2022-2025

Le projet vise à saisir les dynamiques des idéologies néo-autoritaires et illibérales dans toute l'Union européenne, mais se concentrera plus particulièrement sur l'Autriche, la France, la Hongrie, l'Italie, la Pologne, la République tchèque et le Royaume-Uni. 

L'équipe de Sciences Po produira une cartographie des idéologies démocratiques et néo-autoritaires dans l'Europe d'aujourd'hui, basée sur le plus important travail de collecte de données du projet : les documents de partis, les discours de personnalités et l'activité sur les médias sociaux de l’élite politique et de citoyens politisés. L’équipe combinera des enquêtes auprès d'experts et des analyses textuelles quantitatives. 

Les autres volets du projet permettront notamment : de replacer ces idéologies dans leur contexte historico-culturel, d’identifier les facteurs suscitant l’adhésion à ou le rejet de ces idéologies, de mieux comprendre les échanges transnationaux entre mouvements illibéraux et les politiques mises en œuvre lorsqu’ils arrivent au pouvoir. In fine, ces recherches fourniront aux décideurs politiques une boîte à outils complète pour contrer les diverses idéologies néo-autoritaires et améliorer le soutien à la démocratie.

En savoir plus.

L'équipe de Sciences Po :

• Séverine Dusollier, École de droit (PI)
• Rebecca Mignot-Mahdavi, École de droit (PI)
• Jean-Philippe Cointet, médialab (PI)
• Beatriz Botero Arcila, École de droit
• Raphaële Xenidis, École de droit
• Carola Kloeck, CERI
• Bernard Reber, CEVIPOF

Partenaires :

• L’Alliance de Sorbonne Université : Sorbonne Université, l’université de technologie de Compiègne (UTC), le Muséum national d’Histoire naturelle (MNHN), le Pôle supérieur d’enseignement artistique Paris Boulogne-Billancourt, France Education International, INSEAD, l’ENSCi-Les Ateliers, CNRS, INRIA, INSERM, IRD
• L'AP-HP
• L'université Paris-Panthéon-Assas
• Le Cnam
• La Cour de Cassation
• Le Conseil Économique, Social et Environnemental
• l’Ircam
• l’IEA
• L'Onera
• La Fondation sciences mathématiques de Paris  

PostGenAI@Paris a été développé pour répondre aux défis posés par les avancées les plus récentes de l’IA, dans un monde où les frontières entre la technologie et l’intelligence humaine deviennent progressivement indistinctes. Cette nouvelle ère de l’« IA post-générative » (IA qui va au-delà de la génération de contenu pour comprendre et interagir de manière plus autonome et contextuellement mieux adaptée à son environnement) est porteuse de progrès scientifiques et technologiques, mais suscite en retour des inquiétudes et des questionnements profonds liés à la confiance, à la réglementation ou encore à l’éthique de l’IA. Le cluster couvre trois domaines principaux : les technologies de rupture, la santé et les sociétés résilientes.

La mise en œuvre du programme scientifique de PostGenAI@Paris repose sur deux grands piliers. Le premier de ces piliers prend la forme de plusieurs programmes d’accélération collaboratifs (PACs), qui permettent d’articuler étroitement recherche, formation et innovation. Véritables clés de voûte du cluster, ces PACs sont directement impliqués dans le développement de la formation par la recherche, tout en promouvant des collaborations industrielles sur des sujets de pointe. Le second pilier du cluster PostGenAI@Paris s’articule autour d’un programme transverse, conçu pour encourager l’interconnexion entre les membres du consortium et garantir l’agilité et l’évolution du cluster. PostGenAI@Paris souhaite ainsi s’affirmer à l’horizon 2030 comme un pôle d’excellence international, fort d’une expertise transverse dans le domaine de l’IA et d’un réseau robuste de collaborations socio-économiques. Au travers de toutes ces actions, le cluster participe à la formation des prochaines générations en augmentant de manière significative le nombre d’étudiantes et d’étudiants formés chaque année à l’IA à tous les niveaux.

Sciences Po est impliqué dans deux programmes d'accélération collaboratifs : le premier, “AI Normative and Distributive Effects” et le deuxième “AI for augmented deliberation”.

L'équipe de Sciences Po :

• Jean-Philippe Cointet, médialab
• Martial Foucault, CEVIPOF

Partenaires :

• Make.org
• Sorbonne Université
• CNRS

Make.org, Sciences Po, Sorbonne-CNRS se sont associés pour monter le programme de recherche “Communs Démocratiques” qui a pour ambition d’utiliser le potentiel de l’IA générative afin de préserver et renforcer la démocratie. 

Le programme a pour objectif de développer et de diffuser en open source un cadre conceptuel détaillé pour l'application des principes démocratiques aux principales utilisations de l'IA, un modèle d'évaluation des biais des LLM par rapport à ce cadre, des LLM corrigés de ces biais ainsi que des plateformes de participation citoyenne conformes à ces principes. La diffusion de ces Communs Démocratiques permettra aux scientifiques du monde entier, ainsi qu’aux acteurs de la démocratie et aux institutions publiques elles-mêmes, de bénéficier d’outils et des connaissances acquises pour tirer le meilleur parti de l’IA pour des usages démocratiques.

Lire le communiqué de presse.

L'équipe de Sciences Po : 

• Guillaume Lachenal
• Daniel Benamouzig

Période : 2022-2026

Le projet Tractrust réunit des chercheurs en anthropologie, en sociologie politique de la santé, en sciences de la communication et en sciences sociales computationnelles. Il étudie l'acceptation, l'adoption et l'adhésion aux mesures de santé publique pour la prévention et le contrôle de la COVID-19, grâce à des méthodes combinant données numériques et entretiens. Au sein du médialab et en collaboration avec la Chaire Santé de Sciences po (Daniel Benamouzig), nous étudions dans le cadre du projet Tractrust les controverses sociotechniques sur le covid-19 à partir de l’espace public numérique, afin d’examiner le rôle des réseaux sociaux dans la reconfiguration de l’expertise et de l’évaluation scientifique.

En savoir plus.

L'équipe de Sciences Po : Nicolas Sauger

Partenaires : George N. Georgarakis (University of Vienna)

Période : 2023-2026

The rise of misinformation challenges a common conceptualization of truth or accuracy shared by citizens in modern politics, especially so in the age of internet and artificial intelligence. This disruption of a shared understanding of reality undermines the quality of decision-making, cooperation, and communication in the social, political, and economic life and increases affective polarization and political radicalization (Lazer et al. 2018). To mitigate the deleterious effects of misinformation, social scientists and policy stakeholders have proposed a plethora of interventions on online media environments. However, the comparative effectiveness of these interventions has not been tested in a systematic manner until now. 

The proposed project will examine the following question: How effective are the major behavioral interventions that aim to reduce the impact of misinformation on public opinion? To answer this question, the proposed study will examine the causal effect of exposure to misinformation and strategies to mitigate it on truth discernment, public trust, and political attitudes across different countries. 

En savoir plus.

L'équipe de Sciences Po : Frederico Varese

Période : 2021-2026

Nous savons tous que la criminalité organisée est nocive; toutefois, certains de ses aspects importants restent méconnus. Comment les groupes impliqués dans les activités criminelles (production, commerce ou gouvernance) diffèrent-ils les uns des autres? Chacun dispose-t-il d’une structure organisationnelle différente, de membres ayant des profils et des compétences «professionnelles» différents? Dans quelle mesure se recoupent-ils? Dans quelles conditions un groupe se spécialisant dans la production ou le commerce évoluerait-il vers un groupe de criminalité organisée de type de gouvernance? Le projet CRIMGOV, financé par l’UE, répondra à ces questions. L’attention sera portée sur les centres de production de cybercriminalité locale en Europe, le commerce international de drogue depuis la Colombie vers l’Europe, et l’émergence de la gouvernance criminelle à l’intérieur et à l’extérieur des prisons.

En savoir plus.

L'équipe de Sciences Po : Hélène Thiollet

Comment les médias et les hommes politiques encadrent-ils les flux migratoires dans le discours public ? Dans quelle mesure ces discours sont-ils liés aux flux migratoires et d'asile réels ?

Les médias et les hommes politiques réagissent différemment aux divers mouvements internationaux de personnes en fonction de l'origine des arrivants, de leur nationalité et des contextes socio-politiques et économiques dans lesquels les discours sont ancrés. Ils construisent également des discours sur les migrations et notamment sur les "crises migratoires", en l'absence d'afflux substantiels de migrants, mais en lien avec des questions politiques nationales ou internationales plus larges. Le projet DIMIG vise à mieux comprendre les interactions entre les flux migratoires, les médias et les discours politiques. Cette relation fait partie d'une énigme plus générale en sciences sociales : comment les discours façonnent-ils les (fausses) représentations des réalités sociales ?

Le projet rassemble des économistes, des politologues, des géographes et des spécialistes des données et s'appuie sur des méthodes mixtes. Il est organisé en deux work packages qui se dérouleront de 2023 à 2025. Les membres du projet remercient l'Institut Convergences Migrations et la Fondation de France pour leur soutien financier de 72 000 €, sans lequel le projet n'aurait pas pu voir le jour.

Le projet est divisé en deux groupes de travail et deux équipes :

• Une équipe basée au CERI-Sciences po, dirigée par Hélène Thiollet, travaille avec ses institutions partenaires sur les discours de crise migratoire, la polarisation et les imaginaires géographiques dans les médias à travers de multiples contextes (France, Iran, Italie, Pologne, Turquie, Royaume-Uni et Etats-Unis) : en savoir plus
• Une équipe basée au Département d'économie de l'Université Paris 1, dirigée par Léa Marchal et Claire Naiditch, travaille sur les discours politiques sur l'immigration en France : en savoir plus

En savoir plus.

L'équipe de Sciences Po : 

• Kevin Arceneaux
• Martial Foucault

Democracy is in retreat around the world. Over the past decade, the decline in democracy has been notable not only in countries that have just recently transitioned to democracy but also in advanced, stable democratic systems in Europe and North America (Diamond 2021). Political polarization and the spread of misinformation are often cited as culprits in the rise of contested democracy (Finkel et al. 2020; Kingzette et al. 2021; Osmundsen 2021; Vosoughi et al. 2018). In turn, the Online Public Sphere and more specifically the online media sphere have been implicated in both the growth of polarization and the spread of misinformation. The goal of our project is to understand the breadth and contours of the Online Public Sphere (both positive and negative) as well as investigate how it could be structured in ways that encourage more healthy behaviors and support good digital governance.

In this project we propose developing and administering an innovative research design for both understanding how people engage with the online public sphere as well as test interventions that are designed to cultivate healthier and more constructive online behaviors. We plan to do so in the context of the French presidential and legislative elections slated for spring 2022. Following the experience of the last national elections in France, we anticipate that the electoral campaign will be emotionally charged, polarizing, and filled with attempts to spread misinformation. Our research approach will cast a broad net by studying a wide range of attitudes and behavior, both on- and offline.

En savoir plus.

L'équipe de Sciences Po : Jen Schradie

Période : 2023-2027

Qu'est-ce qui détermine la façon dont les gens réagissent à la désinformation ? Fake news, tromperie en ligne et propagande informatique sont autant de termes pour décrire un phénomène qui a donné lieu à un large éventail d'études, en particulier au cours des cinq dernières années, mais une énigme demeure. Les personnes qui croient à la désinformation sont beaucoup plus nombreuses que celles qui déclarent l'avoir consommée. Comprendre les mécanismes de cette déconnexion est l'objectif principal du projet de recherche ACTIVEINFO. 
La plupart de ces recherches existantes se sont concentrées sur les facteurs descendants de la diffusion de la désinformation à l'ère numérique, notamment les entreprises de plateformes Big Tech, le rôle de la réglementation gouvernementale et les puissants acteurs politiques et financiers. Toutes ces approches suggèrent que le public de la désinformation est passif, plutôt que de participer activement à la génération et à la recirculation des nouvelles et des informations, ce qui va à l'encontre de décennies de recherche en communication (Katz et Lazarsfeld 1955). Ce projet veut faire progresser l'état de l'art en examinant un éventail très large de pratiques médiatiques ascendantes et quotidiennes. On entend découvrir ici ce que les gens partagent, mais aussi ce qu'ils ne partagent pas et pourquoi. Le projet mène une comparaison entre deux pays (la France et les États-Unis) avec un travail de terrain riche en qualité et l'apport de méthodes mixtes, souvent absentes des procédures de collecte de données en ligne. Responsable du projet : Jen Schradie

En savoir plus.

L'équipe de Sciences Po : Jen Schradie

What shapes whether people receive, believe, and share disinformation? The rise of ‘fake news’ has become an area of deep concern worldwide, raising questions about the role of information in democratic societies. Observers often point to the January 6 attack on the U.S. Capitol, based on falsehoods about election results, as a critical turning point in how disinformation has dire consequences. Despite a dramatic increase in disinformation research, a crucial and remaining puzzle is that a large number of people believe fake news claims while only a small number of people (often below 1%) consume fake news in their daily news diet (Allen et al. 2020; Fletcher and Nielsen 2018; Grinberg et al. 2019; Tsfati et al. 2020). How and why do people report that they trust unverified information if they are not actually consuming this news directly? To understand this empirical disconnect in the diffusion of disinformation in the digital era (3Ds), this mixed-methods research, DeCodingDisInfo advances the state of the art that selects on the dependent variable of digitally visible fake news and top-down levers of distribution. Current scholarship skews toward top-down powerful players: platforms (like Twitter), politicians (like Trump), or policies (like the GDPR). The audience for disinformation, however, is usually viewed as passive individuals without institutional affiliation. This extant research ignores the broader array of everyday bottom-up active media practices and mechanisms of sharing—or not sharing. Instead, DeCodingDisInfo, an interdisciplinary and mixed-methods project, will uncover how information – both fake news and otherwise – circulates in the digital media environment and in offline spaces. Taking a deeply contextualized and community-based approach, our team will harness the power of a two-country comparison and examine how ideology and institutions shape information flows. This will result in publicly available tools to better decode and deconstruct fake news.

En savoir plus.

L'équipe de Sciences Po : Philipp Brandt

Partenaire : Soror SAHRI, Université Paris Cité

Our objective is to complement the ML applications with insights from social science to understand discriminatory hiring decisions and ensure fair treatment for all. Specifically, we intend to introduce a framework for identifying and mitigating gender fairness issues arising in the composition of the underlying datasets (e.g. demographics dataset) used by ML applications. We apply an interpretable approach to demographics data to identify and observe the potential data features caused by drifts for demographic predictions that are unknown from social scientists before, and can raise concerns in gender fairness. This approach allows us to analyze the interpretability of prediction models, particularly black-box models that are known to be uninterpretable.

L'équipe de Sciences Po : 

• Sylvain Brunier
• Jean-Noël Jouzel

Période : 2023-2028

The Coevolution of Equipment, Digital Technologies and Agroecological models project, funded by the ANR's priority research program and equipment, is managed by Pierre Labarthe, Director of Research at INRAE. It brings together 16 research units and around 50 permanent researchers and academics in the social sciences, including Sylvain Brunier and Jean-Noël Jouzel for the CSO, alongside AGIR, DATASPHERE, INNOVATIONS, ITAP, LEREPS, TERRITOIRES and TETIS.

The project will begin from March 2023 to February 2028.

The CoEDiTAg project (Coevolution of Equipment, Digital Technologies and Agroecological models) aims to understand the co-evolution between 1) the trajectories of development of equipment and digital technologie, and 2) the transformations of the structures, arrangements and organisations that frame agroecological transitions. This project posits that the direction of this coevolution is not predetermined and depends, to a large extent, on the strategies of actors, their socio-economic interactions and public policies. Indeed, Equipment and Digital Technologies (EDiTs) can reinforce the industrialisation of agriculture or be a lever for a stronger ecologisation of agriculture. CoEDiTAg gathers research in different social sciences to understand the mechanisms of this co-evolution.

En savoir plus.

L'équipe de Sciences Po : Kévin Mellet

Partenaire : 

• Christine Balagué, IMT-BS et Titulaire de la Chaire Good in Tech
• Inna Lyubareva, IMT Atlantique
• Cecile Bothorel, IMT Atlantique

Période : 2023-2026

Even if virtual reality and augmented reality have been sources of research publications, the metaverse remains an emerging area of research. The objective of this project is to study the dynamics of metaverse communities’ users’ behaviors, i.e. real behaviors, to better understand the benefits and risks of such immersive technological universes and the determinants of their adoption. The expected results of the netnography (qualitative part of the project) will be a better understanding of the interactions on the metaverse and the governance of the communities. The quantitative part of the project (NLP, dynamic graphs, algorithmic community detection) will provide the insights on the communities’ structures, their dynamic evolution, and the role of intra and extra factors. Finally, on this basis we expect to provide the recommendations for the industrial actors, policy makers and users of metaverses communities. The coordination team includes complementary skills in sociology (Sciences Po), computer science, computational social sciences and management (IMT). Additionally, the research team has a strong expertise in analysis of online platforms, communities, and social networks.

The project will include the recommendations for the industrial actors, policy makers and users of metaverses communities as well as a specific report on the benefits and risks of metaverse communities based on real users’ experiences. These publications will be available on the GoodInTech website. In terms of impact and valorization of the results, 3 GoodInTech webinars, corresponding to the three parts of the project, will be organized during the course of the study. The project will end up with a final conference on the metaverse challenges.

En savoir plus.

L'équipe de Sciences Po : Eduardo Perez

Période : 2021-2026

Informational environments are largely endogenous. They can be, and often are, chosen or designed by individuals or organizations with specific objectives in mind. As recognized by a large literature in economics, information plays a crucial role in shaping the outcome of downstream decisions by strategic players (i.e. at the receiving end of informative signals). However, the structure of information also impacts decisions by strategic agents upstream of the generation of signals, as agents mould the underlying reality differently depending on how other players will eventually be informed about it. Finally, designed information production systems are susceptible to manipulations by third party agents pursuing their own interests.

I will seek to further our understanding of socially or privately optimal information designs, how they shape upstream and downstream decisions, how they can be manipulated by private interests, and how to best anticipate and counter such manipulations. I will rely on the analysis of a largely unexplored designer-agent-receiver class of games, in which the designer picks an information generation system, the agent takes an upstream decision affecting the states of the world, or manipulates the production of information, and receivers choose downstream actions based on realized signals.

The project is organized around the different technologies available to the agent. I will consider fake news production, which is the fabrication of signals that pass as informative but are in fact independent of the truth; state falsification, which consists in falsifying the state of the world, or feeding the information production process with falsified data; pure agency, which is the possibility for the agent to secretly deviate to a different but undistinguishable information generation technology; and state shifting, which is the upstream effort an agent can exert to actually transform the probability distribution of states of the world.

En savoir plus.

L'équipe de Sciences Po : Julia Cagé

Période : 2021-2026

The goal of PARTICIPATE is to develop a comprehensive approach to the study of campaign finance, information and influence, using new individual-level data and investigating recent changes in participation behaviors. What are the motives behind small campaign contributions? Does tax policy affect political giving? Why are so few politicians from the working class and can this change? What is the ability of the media to induce citizens to make electoral decisions they would not make if reporting were unbiased? While there is evidence at the macro level on the flow of money in elections, news consumption, or the extent of charitable giving, relatively little is known at the micro level, e.g. on individual-level behaviours such as the motivations of small donors, the tax-price elasticity of political donations, or the exposure to competing information flows. PARTICIPATE will help fill this gap. By providing groundbreaking evidence on new forms of participation, it will lead to the reassessment of influential theories of special interest groups and policy formation. One distinguishing feature of the proposal is to combine comprehensive individual-level datasets and the use of computer sciences tools (such as natural language processing techniques and machine learning).

PARTICIPATE will advance the existing research in three steps. First, I will propose a unified analysis of political contributions. This will include a groundbreaking assessment of the importance of small campaign donations, and a combined study of charitable giving and political contributions investigating the impact of tax policy on donations. Fundraising success can lead to the emergence of new candidates. I will then consider citizens’ decision to run for elections, and investigate the role played by network in political selection. Finally, given the importance of media organizations in shaping participation, I will study the changing patterns of information propagation and political influence.

En savoir plus.

L'équipe de Sciences Po : Julia Cagé

Période : 2021-2026

Over the past few years, the issue of disinformation has become particularly salient around the world due to the Brexit referendum and Donald Trump’s presidency (see among other Allcott and Gentzkow 2017). Citizens are not equal in front of misinformation. Less-educated people as well as younger ones seem to be more vulnerable to fake news. More generally, “information inequality” is an important concern: people in certain socio-economic segments are much more likely to be informed about recent political events – and to be able to confidently distinguish real news from fake news – than individual in other ones (see in particular Angelucci and Prat 2021). Free access to high-quality news is often discussed as one tool that could be of use to reduce information inequality. In this spirit, a number of countries such as France and Canada have recently introduced policies such as tax credit for a subscription to a newspaper. In 2019, as part of the Protecting Journalism in the Age of Digital Platforms report, Cagé, Prat and others argued in favor of “media vouchers”: give each adult a voucher to donate to her favored media outlets.

But to which extent giving free access to high-quality news (and ensuring high-quality news production) guarantees that citizens who do not consume news will do so when provided with free information, in particular in a digital world where news compete online with entertainment (Gavazza, Nardotto, and Valletti 2019) ? Past research has shown that people may decide to select themselves out of news (see e.g. Prior 2007), and the issue of whether people are used to consuming news is thus a central one. This seems of particular importance for young generations confronted from their early age with an infinite amount of choices. People indeed tend to develop a taste for high-quality news reporting when they are young, so with habit formation they might continue to consume high-quality journalism after they graduate.

While a small number of recent papers in the literature have measured the effects of exposing citizens to politically biased news outlets (Levy 2021; Broockman and Kalla 2021), this research project will be the very first to investigate whether exposing young people to high-quality digital news impacts their political knowledge and overall ability to fight misinformation.

En savoir plus.

L'équipe de Sciences Po : 

• Julia Cagé
• Sergei Guriev
• Emeric Henry

In this project, PI propose to develop and test new solutions and methods to slow down or block the spread of false news and alternative facts. We propose solutions at two levels: upstream, to improve the regulation of platform companies and improve fact-checking procedures, and downstream, to enhance users’ ability to recognize false news. Our project will therefore both enhance digital governance, by proposing and evaluating changes to the design of social media, enhance the impact of fact checking, by evaluating best practices, and improve digital literacy, by designing and testing a training app that will be used in practice.

In the first pilar of the project, we will focus on the downstream actions of improving users’digital literacy and sensitivity to misinformation. Specifically, we propose to design a smartphone application based on two key principles. First, the app will train users to recognize false news and alternative facts. Second, the app will provide tools for users to realize the impact they have on others when they share. We will then carry out a randomized evaluation of the impact of this application on sharing true and false news on social media.

In the second part of pilar 1, we will explore solutions to improve digital governance. In particular, we will explore behavioral interventions that could discourage users from sharing false news. The intervention is to ask social media users to confirm their intention to share content. We will evaluate whether this simple intervention has a larger effect on false compared to true news and propose a system to adapt this cost to the potential accuracy, using a system of user ratings.

We will also measure the benefits of exposure to checked vs. unchecked news for the receivers of information; this will allow us to carry out a welfare analysis of the intervention and propose optimal policies.

The key idea of Pilar 2 is to build a partnership with AFP Factuel, the largest fact-checking group in France part of the third largest international news agency in the world (the Agence France Presse), to first study the impact of fact-checking and second to improve the design of factchecking and propose innovative tools to tackle disinformation. The key objectives of this pilar are to evaluate the impact of fact-checking on the circulation of fake news. and to determine the optimal format of a fact check (length, level of detail), and the best timing to maximize its impact.

Our unique collaboration with the AFP Factuel allows us to study the universe of French language fake news topics, ranging from recent covid misinformation in Congo, antisemitic conspiracy theories in Canada or fake news during the upcoming French presidential election 2022. Our approach is thus suitable to establish best practices for fact-checking and social media platforms around the globe.

En savoir plus.

L'équipe de Sciences Po : Alfred Galichon

Période : 2020-2026

This project seeks to build an innovative economic toolbox (ranging from modelling, computation, inference, and empirical applications) for the study of equilibrium models with gross substitutes, with applications to models of matching with or without transfers, trade flows on networks, multinomial choice models, as well as hedonic and dynamic pricing models. While under-emphasized in general equilibrium theory, equilibrium models with gross substitutes are very relevant to these problems as each of these problems can be recast as such.

Thus far, almost any tractable empirical model of these problems typically required making the strong assumption of quasilinear utilities, leading to a predominance of models with transferable utility in applied work. The current project seeks to develop a new paradigm to move beyond the transferable utility framework to the imperfectly transferable utility one, where the agent’s utilities are no longer quasi-linear.

The mathematical structure of gross substitutes will replace the structure of convexity underlying in models with transferable utility.

To investigate this class of models, one builds a general framework embedding all the models described above, the “equilibrium flow problem.” The gross substitute property is properly generalized and properties (existence of an equilibrium, uniqueness, lattice structure) are derived. Computational algorithms that rely on gross substitutability are designed and implemented. The econometrics of the problem is addressed (estimation, inference, model selection). Applications to various fields such as labor economics, family economics, international trade, urban economics, industrial organization, etc. are investigated.

The project touches upon other disciplines. It will propose new ideas in applied mathematics, offer new algorithms of interest in computer science and machine learning, and provide new methods in other social sciences (like sociology, demography and geography).

En savoir plus.

L'équipe de Sciences Po : Clément de Chaisemartin

Applied economists routinely evaluate the effect of economic policies. For instance, what is the effect of raising the minimum wage on employment? Angrist & Pischke (2010) argue that applied economics has recently experienced a ""credibility revolution'': by switching to transparent methods to evaluate policies, applied economists have increased the credibility of their findings, and the impact of their work. In the first part of this proposal, de Chaisemartin shows that the credibility revolution is not complete. Two-way fixed effects regression, a policy-evaluation method used in as many as 26% of the most-highly cited papers recently published in the American Economic Review, relies on the assumption that the policy's effect is constant, across units and over time. In most applications, this assumption is not credible. Therefore, the project proposes a series of new differences-in-differences estimators, that do not rely on this constant effect assumption, and that can be used in virtually all the applications where two-way fixed effects regressions have been used. In the second part of this proposal, de Chaisemartin argues that the credibility revolution's focus on unbiased estimators may be hard to defend, as variance also matters. The project explores two ways of trading-off bias and variance. The first amounts to combining an unbiased and a biased but potentially more efficient estimator of the same parameter. The second amounts to deriving the minimax estimator of the policy's effect, under the assumption that this effect cannot be larger than a (potentially large) constant B. Finally, the third part investigates the potential impact of the ``credibility revolution''. Specifically, a randomized controlled trial will be run to measure the effect of following an online course presenting policy evaluations, in two very different populations: policy makers and members of the general public, focusing on individuals with a low trust in institutions in that second group.

En savoir plus.

L'équipe de Sciences Po : Jean-Marc Robin

Research proposal MARNET aims at improving our empirical knowledge of markets structured as bipartite networks (all connections involve two different categories of agents) by providing better statistical models of network formation and of the effects of the network structure on outcomes.

The use of random graphs, in which an edge between two nodes is the realization of a random draw, to model networked social, technological and biological systems has been subject to a novel and vigorous effort over the last twenty years. Robin will contribute to the knowledge corpus on networked economic systems with two separate research lines: first, in studying marriage formation and intrahousehold resource allocation; second, in studying wage dynamics and employment mobility using linked employer-employee data.

The first research line aims to better understand why different marriage markets (regions, countries, time periods) look different. The aim is to disentangle and quantify the sources of differentiation that relate to standard socio-economic variables, such as education and wages, from those that relate to culture and social norms. Various extensions will be considered to improve the design and realism of the model (aging and fertility, lots of heterogeneity, unobserved heterogeneity, housing and asset accumulation).

In the second research line, recent advances in dynamic random graphs will be used to model individual wage dynamics and the mobility of workers across firms. The main empirical objective is to quantify the degree of assortative matching of workers and firms in the labour market, and the relative contributions of worker and firm heterogeneity in wages. Several extensions will be considered (time-varying types, subjecting the network structure to pre-determined firm sizes, opening firm nodes by modelling wage dynamics and occupational mobility within firms).

En savoir plus.

L'équipe de Sciences Po : 

• Beatriz Botero Arcila
• Séverine Dusollier
• Jeremy Perelman
• Guillaume Tusseau

How can the values and rights needed to sustain democracies and the common good be upheld and ensured in our digital world? What is the role of the law in doing so? Sciences Po’s Law School Towards a New Digital Rule of Law project will assess the legal governance/regulation of the Internet and of some of its technologies in identified case studies in order to develop a new line of research investigating and experimenting the necessary role of law in current internet evolution. The Law School intends to develop this line of research for the years to come around different specific issues challenging the rule of law. In the next three years, the Project will focus on examining the value and of constitutional law concepts for a digital rule of law, how to govern AI to make it compatible with democracies, and how to use infrastructures to govern and how to govern infrastructures to address data inequality and shape a less concentrated web. All projects will be developed emphasizing three main pillars (1) hosting and supporting rigorous academic research focused on relevant governance challenges;(2) channeling the research into early action and early impact through the law school’s clinical program; and (3) implementing public outreach, by hosting conferences and workshops where different stakeholders can discuss and influence policy on time-contingent issues.

En savoir plus.

L'équipe de Sciences Po : 

• Beatriz Botero Arcila
• Rachel Griffin

The 2022 Digital Services Act (DSA) establishes a new comprehensive regulatory framework for online platforms in the EU. It reserves the most extensive obligations for ‘very large online platforms’ (VLOPs), defined as those with over 45 million users. These platforms are considered particularly important for the broader media system and for political and democratic discourse. VLOPs are obliged to assess and evaluate ‘systemic risks’ in specified areas (e.g. fundamental rights, polarisation, public health and security); reasonably and proportionately mitigate these risks; and report to the Commission on their
mitigation measures.

Importantly, the DSA also introduces mechanisms for vetted researchers to access platforms’ internal data “for the sole purpose of conducting research that contributes to the detection, identification and understanding of systemic risks in the Union”. Policymakers envisage that independent scrutiny from academic researchers and civil society organisations will play an essential role in identifying and defining systemic risks associated with social media, as well as holding VLOPs accountable for effectively mitigating them.

This project will bring together a team of researchers from Sciences Po’s Law School and Médialab to be a central part of that effort. The project aims to monitor and critically evaluate how the systemic risk framework is being implemented, and to position itself as a key resource for regulators implementing and enforcing the DSA. It will provide actionable guidance and resources for regulators, civil society and researchers on how to utilise this framework to strengthen the governance of online media platforms.

We will focus on two main lines of research: How are systemic risks and appropriate risk mitigation measures understood and defined by VLOPs and other stakeholders within the DSA framework? What data should researchers have access to in order to evaluate and monitor systemic risks within the DSA framework?

En savoir plus.

L'équipe de Sciences Po : Dina Waked

Partenaire : Stanford Center for Legal Informatics (CodeX)

Contemporary digital markets are dominated by business models based on proprietary software, centralising decision-making power in a few hands, and allocating revenues to shareholders and management. Contrastingly, the novel technologies referred to as “Web3” and artificial intelligence (“AI”) come with the promise of fostering new governance structures that enable the transition to open standards and protocols, collaborative decision-making, and giving a voice to a broader range of stakeholders. Desynchronised with the new technological frontiers, organisational laws worldwide revolve around the standard corporate form with asset ownership, centralised management, and a narrow orientation towards shareholder value. Yet this corporate form was crafted in response to the technological affordances of the previous century, when transaction and agency costs prohibited the emergence of open-source, collaborative and multi-stakeholder governance mechanisms. Given the technological affordances of Web3 and AI, there is a misfit between technological development and the governance models recognised by the law. Building on this misfit, this project questions how legal systems can adapt to and foster innovative governance mechanisms. This overarching question is split in three sub-questions: 1) how can novel technologies be leveraged to create associational models that provide a meaningful alternative to the corporate form, 2) how the transition to open-source models and decomposable tokens will affect property rights, and 3) what legal forms will Web3 and AI applications evolve to take.

These questions are topical for several reasons. First, instead of enabling a fairer distribution of economic opportunities and bolstering participatory mechanisms, technological developments have been instrumentalized for anti-democratic purposes, concentrating outsize economic power in the hands of a few. Second, we witness the shortcomings of traditional governance mechanisms in the case of companies active at the frontier of innovation, the OpenAI debacle being only an episode in a series of scandals. These trends render this research exceptionally timely. 

En savoir plus.

L'équipe de Sciences Po : Jean-Philippe Cointet

Partenaires : 

• Max Planck Institute
• University of Amsterdam
• Free University of Brussels
• University of Venice Cà Foscari
• Karlsruher Institute of Technology
• University of Leipzig
• University of Bristol

Current diagnoses that democracy is in crisis at the beginning of the 21st century share a common argumentative reference point: the (implicit) reference to the dysfunctional constitution of the political public sphere which is currently undergoing structural change. The rise of social media platforms is considered as one of its main constituents. While social media make the public arena more open and thus more responsive, these platforms also lead to new mechanisms of fragmentation and exclusion, an erosion of norms in public debate and a loss of trust in traditional institutions.

The project will reconsider the diagnoses of this crisis by:

(1) providing better empirical evidence for the impact of social media on society with respect to political debates,

(2) understanding the main causal mechanisms of this impact and

(3) developing tools that improve the capacity of social media to contribute to the functioning of the public arena in a liberal democracy, i.e., deliberation, legitimation and the self-perception of the democratic subject.

En savoir plus.

L'équipe de Sciences Po : Donato Ricci, Pauline Gourlet, Dominique Cardon, Valentin Goujon, Maxime Crépel,  Axel Meunier 

Partenaires : 

• Concordia / INRS (Canada)
• ZeMKI / HIIG (Germany)
• CIM, Warwick (UK)

Période : 2020-2024

Shaping AI is a multi-national and multi-disciplinary social research project that examines the global trajectories of public discourse on artificial intelligence (AI) in four countries (Germany, UK, Canada and France) over a ten-year period 2012-2021. 

Our comparative analysis will enable us to understand the differences and similarities in the construction of AI as a new scientific paradigm, societal force and contested political reality in these four countries by contextualizing their geo-political, historical and situational peculiarities. Funded by the European Open Research Area initiative, Shaping AI brings together leading research teams from each of the four countries under scrutiny.

Our project aims to be critical, comparative, qualitative, and to this end, we have adopted a four-dimensional research design, with the four teams collaborating to map and analyse AI controversies in and across four different spheres in which public discourse about AI has taken shape in the last ten years: media, policy, research and participation.

The research on AI controversies in each of the four spheres is led by a different research team, with the study of "spill-over" controversies that have arisen between the spheres reserved for the second half of the project.

Publications récentes :

• PRISES , MÉ.PRISES, EM.PRISES , DÉ.PRISES , RE.PRISES… Les formes de participation au développement de l'IA en France
• Gourlet, P., Ricci, D., & Crépel, M. (2024). Reclaiming artificial intelligence accounts: A plea for a participatory turn in artificial intelligence inquiries. Big Data & Society, 11(2). 

En savoir plus.

L'équipe de Sciences Po : Sylvain Parasie

Partenaires : 

MEDIALEX est un projet de recherche collaboratif qui rassemble, autour de deux laboratoires de sciences sociales et deux laboratoires d’informatique, des chercheur·se·s qui sont habitué·e·s à travailler ensemble : 

1) Le médialab, spécialisé dans le développement des méthodes numériques pour les sciences sociales, 

2) Le LATTICE, laboratoire de l’École Normale Supérieure/PSL, apporte des compétences importantes dans le domaine du traitement automatique de la langue. 

3) Le service de la Recherche de l’INA, apporte ses compétences en TAL et permet d’accéder à des corpus de données provenant des médias audiovisuels français (chaînes TV, radio) dont l’INA assure la conservation et qu’ils enrichissent par des méthodes de TAL (speech to text, détection d’événement, etc.) 

4) Le CREST, laboratoire d’économie, de statistique et de sociologie quantitative, apportera son expertise de sociologie politique dans l’étude à la fois ethnographique et numérique du monde parlementaire français. 

À ces partenaires s’ajoute une collaboration privilégiée avec le CARISM, de l’Université Panthéon-Assas.

Période : 2021-2025

L’essor des technologies numériques a profondément déstabilisé l’espace public, en élargissant l’accès à la parole publique. Les transformations médiatiques, notamment liées à l’essor des réseaux sociaux, ont permis à un plus grand nombre d’acteurs d’alimenter en sujets le débat public. En témoigne le mouvement des “gilets jaunes”, au départ très éloigné des sphères médiatiques et politiques, qui est parvenu à imposer de nouveaux thèmes dans le débat public notamment par le biais des réseaux sociaux, ce qui s’est traduit par l’adoption d’une loi de “mesures d’urgence économique et sociale” fin 2018. En témoigne également la démarche de nombreux parlementaires et élus, qui s’expriment sur les réseaux sociaux en courtcircuitant les médias traditionnels, intermédiaires incontournables de l’accès au public jusque-là.

Dans ce contexte, le projet MEDIALEX vise à répondre à la question suivante : dans quelle mesure la numérisation de l’espace public a-t-elle bouleversé la capacité des élus, des médias traditionnels et des publics à imposer les sujets prioritaires du débat public ? Autrement dit, comment les transformations de l’espace public et médiatique modifient-elles les dynamiques d’émergence, d’imposition et de circulation des thèmes dans le débat public ?

Pour répondre à cette question, le projet MEDIALEX propose de développer des méthodes numériques originales, en combinant des compétences de sciences sociales, à la fois quantitatives et qualitatives, avec des compétences en modélisation et en traitement automatique de la langue. Il s’agit ainsi de traiter de grands corpus de données et de traces numériques, de façon à pouvoir identifier les dynamiques d’influence entre les mondes parlementaire et médiatique et la sphère publique. Du point de vue du traitement automatique du langage, il s’agit de détecter des thèmes, événements, petites phrases et citations qui peuplent les différents sous-espaces. Ces méthodes permettront de modéliser à différents grains les logiques d’influence entre ces trois arènes discursives.

En savoir plus.

L'équipe de Sciences Po : 

• Jean-Philippe Cointet
• Donato Ricci
• Maxime Crépel
• Cosma Cazé

Partenaires :

• CNRS
• MNHN
• IFREMER
• Réseau des universités marines : Université Littoral Côte d’Opale, Université de Nantes, Université de Bordeaux, Université de Montpellier et Aix-Marseille Université
• Office français de la biodiversité

Période : 2021- 2028

Le projet de recherche FUTURE-Obs, lauréat de l’appel à projets « Un océan de solutions » du « Programme prioritaire de recherche » (PPR), a pour objectif de proposer des stratégies d’observations multi-échelles et multidisciplinaires des socio-écosystèmes qui associent aux approches traditionnelles d’observations de l’océan, de nouvelles techniques comme la génomique environnementale et l’imagerie in situ pour la biodiversité, ou encore l’utilisation des données provenant des réseaux sociaux pour les usages. Cette combinaison de méthodes ouvre de nouvelles voies d’étude et conduit à la mise en place d’observatoires dits augmentés. La multiplicité des données acquises à différentes échelles et sur l’ensemble des composantes des socio-écosystèmes favorisera l’élaboration de nouveaux indicateurs faisant appel aux outils issus du champ de l’intelligence artificielle comme le machine learning. Il s’agira ainsi, d’une part, de suivre et de mieux appréhender le fonctionnement et les trajectoires des socio-écosystèmes côtiers, et d’autre part, de faciliter la mise en œuvre d’initiatives conçues pour répondre aux principaux enjeux environnementaux actuels.

En savoir plus.

L'équipe de Sciences Po : 

• Sylvain Parasie 
• Pedro Ramaciotti Morales

Période : 2022-2024

The last two decades have seen the emergence of different hypotheses about disorders of digital public spaces, such as fragmentation (or “bubbles”), polarization, extremism, and the role that algorithms mediating this space might have on them by pushing the visibility and virality of particular contents. Conclusive results, however, have proved elusive, as a growing body of research seems to present a contradicting picture: disorders might be pervasive, capturing the attention of policy makers and the general public, but with no widely-accepted definitions or metrics emerging to quantify them or the role algorithm might have on them, let alone actionable means to design better algorithms that would minimize identified negative outcomes. Meanwhile, growing evidence suggests that Recommender Systems might be leveraging political opinions of users and other features of public debate associated with societal cleavages. This project takes inspiration in ideological social network embedding methods and political surveys for the analysis of party systems in policy spaces to propose a double network and opinion space modeling of digital public spaces. Using this double network and spatial opinion analysis, this project then proposes to test whether algorithms learn and leverage political opinions of users through algorithmic explainability, how they affect information dynamics in public debate, and to open a path towards actionable tools capable of guiding algorithm design, governance, and policy.

En savoir plus.

L'équipe de Sciences Po : Achim EDELMANN

Digital transformation has significantly changed how information is shared. Platforms like WhatsApp, Facebook, and Twitter have become key for spreading information, but they also allow for the spread of false or harmful content. This project will show the impact of scientific and political endorsements on its spread. It will help identify effective ways to reduce misinformation and its polarizing effects. Additionally, it will improve scientific outreach by providing clear strategies for science to act as a trusted information source. It will also impact education by informing future research and helping students navigate social media safely and effectively.

En savoir plus.