HAL Paris Dauphine-PSL
Not a member yet
21932 research outputs found
Sort by
Inter‐Group Cooperation and Conflict. Experimental Evidence From Yemen
No full textInternational audienceThis study investigates pro‐sociality among individuals during an ongoing war. We focus on Yemen, where the recent large‐scale conflict has divided the country geographically, putting different regions in opposition with one another. We collect original survey data and develop a mobile phone‐based version of a public good game to allow participants from regions on opposite sides of the conflict to play in pairs across the entire country. Participants play two games: one with an in‐group partner (from the same region) and one with a randomly assigned partner from a different region—either from the antagonistic out‐group (opposite conflict side) or from the non‐antagonistic out‐group (same conflict side). We further merge our data with the Armed Conflict Location and Event Data Project (ACLED) conflict data to explore whether conflict intensity affects participants' behaviours in the games. Our results show a lower level of cooperation with the rival out‐group compared to the non‐rival out‐group, along with higher in‐group cooperation among participants heavily exposed to the war. Further results show that lower pro‐sociality towards the rival out‐group is mostly found among players from the North of Yemen, where sectarian identity is most clearly tied to the conflict division, and this effect is exacerbated by the intensity of conflict exposure. These results provide some of the clearest empirical support to date for concerns about the destruction of inter‐group cooperation during an ongoing war
L’émission d’une facture d’honoraires par une SELARL d’avocats est un acte de gestion (Cass. com. 26 nov. 2025)
No full textInternational audienceLa Cour de cassation décide, à propos de l’émission d’une facture d’honoraires par l’associé-gérant d’une SELARL d’avocats, que l’émission d’une facture par une société constitue un acte de gestion susceptible comme tel d’engager la responsabilité de son dirigeant. Tout en expliquant son fondement, le commentaire discute la solution retenue en montrant qu'elle n'est pas aussi évidente qu'il n'y paraît. Puis il analyse la portée de cette solution au-delà des SELARL d'avocats
Composite likelihood inference for the Poisson log-normal model
No full textInternational audienceThe Poisson log-normal model is a latent variable model that provides a generic framework for the analysis of multivariate count data. Inferring its parameters can be a daunting task since the conditional distribution of the latent variables given the observed ones is intractable. For this model, variational approaches are the golden standard solution as they prove to be computationally efficient but lack theoretical guarantees on the estimates. Sampling-based solutions are quite the opposite. We first define a Monte Carlo EM algorithm that can achieve maximum likelihood estimators, but that is computationally efficient only for low-dimensional latent spaces. We then propose a novel inference procedure combining the EM framework with composite likelihood and importance sampling estimates. The algorithm preserves the desirable asymptotic properties of maximum likelihood estimators while circumventing the high-dimensional integration bottleneck, thus maintaining computational feasibility for moderately large datasets. This approach enables grounded parameter estimation, confidence intervals, and hypothesis testing. Application to the Barents Sea fish dataset demonstrates the algorithm capacity to identify significant environmental effects and residual interspecies correlations
Learning to Play Two-Player Perfect-Information Games without Knowledge
No full textThis paper proposes several techniques for learning game state evaluation functions by reinforcement. The first generalizes tree bootstrapping to reinforcement learning with non-linear functions, ensuring no information loss. The second modifies Unbounded Best-First Minimax by extending the best action sequences to terminal states. The third replaces the classic +1/ -1 game outcome with reinforcement heuristics such as quick wins, slow defeats, scoring, or mobility. The fourth introduces a completion technique that leverages state resolution. The fifth defines a new distribution for action selection.Experiments show these techniques significantly improve play strength. We combine them in a single algorithm, Athénan, and compare it to ExIt, a leading self-play reinforcement learning algorithm without prior knowledge. Results demonstrate that Athénan outperforms ExIt.Athénan is then applied to the games Hex, Othello, and Arimaa. In all cases, it surpasses the state of the art without using domain knowledge. We also test Athénan on the single-player game Morpion Solitaire, where it again reaches state-of-the-art performance without prior knowledge.Overall, Athénan shows that reinforcement learning can achieve cutting-edge results across a diverse set of games without relying on handcrafted heuristics or expert input
Équations aux dérivées partielles et applications: [résumé des cours et travaux : 2021-2022]
No full textInternational audienc
Strong large deviation principles for pair empirical measures of random walks in the Mukherjee-Varadhan topology
No full textFinal version.International audienceIn this paper we introduce a topology under which the pair empirical measure of a large class of random walks satisfies a strong Large Deviation principle. The definition of the topology is inspired by the recent article by Mukherjee and Varadhan~\cite{MV2016}. This topology is natural for translation-invariant problems such as the downward deviations of the volume of a Wiener sausage or simple random walk, known as the Swiss cheese model~\cite{BBH2001}. We also adapt our result to some rescaled random walks and provide a contraction principle to the single empirical measure despite a lack of continuity from the projection map, using the notion of diagonal tightness
Fieldwork as inquiry in management research: dancing with the field
No full textInternational audienceMost approaches to qualitative research methods in management and organization studies (MOS) implicitly enact spatial perspectives, crystallized in the idea of “accessing the field”. Field sites are thus performed as objective spaces in which research takes place. In this paper, we set out to challenge this spatial perspective by developing a processual approach that conceptualizes fieldwork as ‘inquiry’ as defined by John Dewey. To this end, we draw on an inter-case analysis of four independent research projects to demonstrate the processual, fluid, continuous, and shared constitution of a field of inquiry. We call this alternative approach ‘dancing with the field’. It allows us to move beyond the highly spatialized notion of access to the field and to materialize orprioritize different forms of relationality. We understand ‘accessing the field’ and ‘dancing with the field’ as two intertwined and complementary modes of both attachment to and detachment from the field
A gradient flow on control space with rough initial condition
No full textInternational audienceWe consider the (sub-Riemannian type) control problem of finding a path going from an initial point to a target point , by only moving in certain admissible directions. We assume that the corresponding vector fields satisfy the bracket-generating (Hörmander) condition, so that the classical Chow-Rashevskii theorem guarantees the existence of such a path. One natural way to try to solve this problem is via a gradient flow on control space. However, since the corresponding dynamics may have saddle points, any convergence result must rely on suitable (e.g. random) initialisation. We consider the case when this initialisation is irregular, which is conveniently formulated via Lyons' rough path theory. We show that one advantage of this initialisation is that the saddle points are moved to infinity, while minima remain at a finite distance from the starting point. In the step -nilpotent case, we further manage to prove that the gradient flow converges to a solution, if the initial condition is the path of a Brownian motion (or rougher). The proof is based on combining ideas from Malliavin calculus with Łojasiewicz inequalities. A possible motivation for our study comes from the training of deep Residual Neural Nets, in the regime when the number of trainable parameters per layer is smaller than the dimension of the data vector
Jumeaux numériques dans la gestion de la chaîne logistique : portée et problèmes méthodologiques
No full textInternational audienceThis paper investigates the implementation of Digital Twins (DTs) in Supply Chain Management (SCM), highlighting the gap between their conceptual promise and practical applications. DTs are recognised for their potential to correct real-time deviations and anticipate and prevent disruptions as they emerge; however, operational deployments in SCM remain rare. Numerous studies mislabel simulation models or Digital Shadows (DSs) as DTs, blurring essential distinctions. To address this issue, this paper adopts a praxeological approach that aims to situate observed implementations within their decision-making context. From this perspective, we propose a novel methodological framework that integrates the historical evolution of supply chain information and decision systems with a multidimensional analysis grid, outlining technological progress from simulators to DSs and DTs. This grid evaluates core DT functionalities (simulation, detection, anticipation and correction) across 51 empirical case studies, providing granular insights into maturity levels and AI enhanced patterns. The results show that most models support monitoring and decision-making, but only 16% achieve closed-loop capabilities typical of fully functional DTs, mainly in closed systems. In contrast, open systems still depend on human intervention, although AI can increasingly support such contexts. This praxeological approach provides critical and evidence-based snapshots of actual implementations. It offers researchers a clarified conceptual lens and practitioners empirically grounded guidance, outlining avenues for future inquiry and a reflective framework to guide the development and governance of DTs in supply chains.Cet article examine la mise en œuvre des jumeaux numériques (DT) dans la gestion de la chaîne logistique (SCM), et souligne l’écart entre leurs fondements conceptuels et leurs applications opérationnelles. Les DT sont reconnus pour leur potentiel à corriger les écarts en temps réel et à anticiper et prévenir les perturbations dès leur apparition ; cependant, leur déploiement opérationnel dans la SCM reste rare. De nombreuses études qualifient à tort les modèles de simulation ou les Digital Shadows (DS) de DT, brouillant ainsi les frontières. Pour remédier à ce problème, cet article adopte une approche praxéologique visant à situer les mises en œuvre observées dans leur contexte décisionnel. Dans cette perspective, nous proposons un cadre méthodologique fondé sur l'évolution historique des systèmes d'information et de décision de la chaîne logistique, présenté dans une grille d'analyse multidimensionnelle, s'appuyant sur les progrès technologiques allant des simulateurs jusqu'aux DS et aux DT. Cette grille évalue comment sont traitées les fonctionnalités essentielles des DT (simulation, détection, anticipation et correction) à partir de 51 études de cas empiriques, fournissant des informations détaillées sur les niveaux de maturité et les les améliorations potentielles apportées par l'IA. Les résultats montrent que la plupart des modèles prennent en charge la surveillance et la prise de décision, mais que seuls 16 % d'entre eux atteignent les capacités de rétroaction typiques des DT pleinement fonctionnels, principalement dans des systèmes fermés. En revanche, les systèmes ouverts dépendent toujours, au final, de l'intervention humaine, bien que l'IA puisse de plus en plus prendre en charge ces contextes. Cette approche praxéologique fournit des analyses critiques fondées sur des cas de mise en œuvre réels. Elle offre aux chercheurs une perspective conceptuelle clarifiée et aux praticiens des conseils fondés sur l'expérience, décrivant des pistes de recherche futures ainsi qu'un cadre de réflexion pour guider le développement et la gouvernance des DT au sein des chaînes logistiques
Climate shocks and banking sector stability: Evidence from El Niño southern oscillation
No full textInternational audienceThis study introduces a novel ex ante approach to assess the short-term impact of climate shocks on banking sector stability by examining the effect of El Niño Southern Oscillation (ENSO) on banking sector distance-to-default. Using dynamic panel data econometric modeling, we investigate the macroeconomic implications of ENSO-induced climate shocks, such as El Niño and La Niña events, on banking sector stability in 51 countries across three regions particularly exposed to the consequences of ENSO oscillations (East Asia and Pacific, Latin America and the Caribbean, and Sub-Saharan Africa) during the period 2000–2020. Our findings show that the adverse effects of these climate shocks on banking sector stability are unevenly distributed among countries, with more pronounced and robust adverse effects of El Niño events in the short-term, particularly in Latin America and the Caribbean and, to a lesser extent, Sub-Saharan Africa. We also document the short-term adverse effects of La Niña events for Latin American and the Caribbean countries. Further estimates suggest that the increase in non-performing loans is a key transmission channel linking El Niño events to banking sector stability. As global warming should intensify the frequency and magnitude of ENSO's cyclical pattern, these findings can help estimate the potential adverse effects of climate change-related natural disasters on banking sector stability and inform future mitigation and adaptation policies