HAL-Paris1
Not a member yet
124252 research outputs found
Sort by
À qui la faute ?
No full textInternational audienceConseil d'État, 15 juillet 2025, no 494073, Société Nouvelle Laiterie de la Montagne, Lebon T. ; AJDA 2025. 140
Audio Deepfake Detection in the Age of Advanced Text-to-Speech models
No full textRecent advances in Text-to-Speech (TTS) systems have substantially increased the realism of synthetic speech, raising new challenges for audio deepfake detection. This work presents a comparative evaluation of three state-of-the-art TTS models-Dia2, Maya1, and MeloTTS-representing streaming, LLM-based, and non-autoregressive architectures. A corpus of 12,000 synthetic audio samples was generated using the Daily-Dialog dataset and evaluated against four detection frameworks, including semantic, structural, and signal-level approaches. The results reveal significant variability in detector performance across generative mechanisms: models effective against one TTS architecture may fail against others, particularly LLM-based synthesis. In contrast, a multi-view detection approach combining complementary analysis levels demonstrates robust performance across all evaluated models. These findings highlight the limitations of single-paradigm detectors and emphasize the necessity of integrated detection strategies to address the evolving landscape of audio deepfake threats
Optimal Land Use and Pollution Abatement under Discounting: A Spatiotemporal Analysis with State Constraints
No full textWe develop a rigorous spatiotemporal framework that incorporates ecological state constraints and reversible pollution dynamics with fertile land as the sole bounded production input. Using an extended Pontryagin Maximum Principle and generalized Kuhn-Tucker conditions, we characterize the structure of optimal solutions across discounting regimes. Thus, this paper investigates optimal strategies for land use and consumption under varying time discount rates in spatial settings. Under low discounting, the system supports sustainable dynamics, avoiding excessive consumption and enabling full restoration of fertile land. Fertility restoration is local in the heterogeneous case and global in the homogeneous case. In contrast, high discounting leads to boundary behavior, partial degradation, or irreversible loss, depending on critical thresholds. We derive explicit solutions when space is homogeneous. In heterogeneous settings, we construct hybrid solutions where the system transitions from a finite-horizon control problem to a structured long-run regime. Our results provide analytical benchmarks and highlight the pivotal role of time preferences in shaping long-term environmental and economic outcomes. This work contributes to the literature on spatial growth, optimal control with state constraints, and sustainable resource management under ecological limits
Campaigning across platforms: how party status, social profiles, and emotional tone shaped user engagement with French MPs during the 2022 presidential election
No full textInternational audienceThis article examines audience engagement with posts published by 387 French MPs on Facebook, Instagram, and Twitter during the 2022 presidential election (N = 35,348). Using a supervised machine learning method, we investigate how engagement varies according to MPs' political affiliation, social characteristics, post topics, and emotional tone. The findings reveal that party status shapes online visibility: posts from opposition MPs, particularly those from radical parties, consistently elicit higher engagement than those from the governing party, which tends to favor neutral and coordination-oriented messaging. Social underrepresentation also plays a role: younger, rural, and female MPs generate more engagement in specific contexts, yet these effects vary substantially across platforms. For instance, women and urban MPs receive more reactions on Twitter, whereas Facebook and Instagram favor rural MPs. These platform-dependent patterns suggest that audience engagement is shaped not only by asymmetries in political status and social background, but also by platform-specific affordances: Facebook and Instagram tend to reward locally embedded and coordination-oriented communication, particularly among younger and rural MPs, while Twitter amplifies adversarial rhetoric and politically polarizing content, especially among urban or radical actors
Incompatibilité d'une création d'entreprise avec les fonctions de l'agent : des rapports entre l'avis du référent déontologue et l'avis de la HATVP
No full textInternational audienceNote sous CE, 24 juillet 2025, nos 490199, 490249 (Lebon T.), M. B. c/ Communauté d'agglomération du pays voironnai
Les stratégies familiales de reproduction : comment se transmettent le capital économique et le capital culturel dans la France contemporaine ?
No full textInternational audienc
Local Communication in Repeated Games with Local Monitoring
No full textI consider repeated games with local monitoring: each player observes his neighbors' moves only. Hence, monitoring is private and imperfect. I assume local and public communication: communication is restricted to neighbors, and each player sends the same message to each of his neighbors at each stage. Both communication and monitoring structures are given by the network. The solution concept is perfect Bayesian equilibrium. In the four-player case, a folk theorem holds if and only if the network is 2-connected. Some examples are given for games with more than four players
When the dark core meets the digital: Toward a theory of dark digital leadership
No full textInternational audienceDigitalization is transforming the landscape of leadership, amplifying not only its constructive but also its destructive expressions. Despite growing interest in toxic leadership and digital leadership, existing research has largely treated these domains as separate. This paper theorizes Dark Digital Leadership (DDL) as a multilevel, socio-technical phenomenon that emerges from the interplay between leaders’ dark personality traits and the affordances of digital infrastructures. Drawing on the Dark Triad framework, Adaptive Structuration Theory, the Job Demands–Resources Model, and Upper Echelons Theory, we develop a Dynamic Socio-Technical Loop Model that explains how digital systems magnify, routinize, and legitimize destructive leadership behaviors. The framework delineates micro-level mechanisms, such as technostress and moral disengagement, meso‑level dynamics, including trust erosion and team fragmentation, and macro-level consequences, such as ethical drift and institutional instability, while identifying boundary conditions related to organizational culture, governance, and digital affordances. By embedding destructive leadership within the architecture of algorithmic organizations, this study bridges psychological and structural perspectives and contributes a socio-technical lens to understanding the dark side of leadership in the digital age
Effets d'une clause d'élection de for en cas de pluralité de défendeurs en droit commun français
No full textInternational audience(Civ. 1re, 8 oct. 2025, no 23-16.756, F-B, Unica Realty, D. 2025. 1745
