Mardi 16 juin 2026 à 14h en salle 26-00/124
Comment se retrouver si l’on est dispersé dans une ville ? Se coordonner, en échangeant nos positions grâce aux moyens de communications modernes, semble facile. Mais comment parvenir au même résultat si les communications sont impossibles et qu’aucun point de rendez-vous n’est identifiable, comme c’est par exemple le cas lors d’un mouvement social réprimé ? Au cours de mon doctorat, mon travail fut d’identifier les ingrédients clés pour qu’un protocole de déplacement garantisse que des marcheurs se retrouveront au même endroit, quelles que soient leurs positions de départ.
]]>Mardi 14 avril 2026 à 14h en salle 26-00/534
Language is central to teaching, yet we lack principled ways to characterise which aspects of instructional text make explanations effective, particularly in cases where people are simultaneously learning from experience while being taught. We combined a two-stage human behavioural paradigm with a large language model–based analysis pipeline. “Teacher” participants first learned probabilistic two-armed bandit tasks through trial and error, then wrote free-text lessons for future “Pupil” participants, who received one such lesson (or none) before performing the same tasks. Lessons judged as high quality by external experts improved pupils’ reinforcement learning performance relative to low-quality lessons and no-instruction controls. To understand why, we introduced LLM-DISC, an inferential use of multi-step large language model processing to uncover the latent semantic dimensions in instructional text. Four dimensions (Memorization, Pattern Recognition, Option Ranking, and Randomness) predicted both expert judges’ rankings of lessons and pupils’ behavioural outcomes. Finally, by manipulating these dimensions in LLM-generated “Good” and “Bad” lessons, we causally altered pupils’ learning in a manner that replicated the original “human teacher” effects. These results show that a low-dimensional semantic structure of teaching language measurably shapes experiential reinforcement learning, and that it can be systematically discovered, interpreted, and used to probe the cognitive mechanisms underlying teaching and learning.
]]>In the 25th International Conference on Autonomous Agents and Multi-Agent Systems (AAMAS2026)
During social movements, protesters need to gather with limited communication means and limited knowledge other than what they observe in their direct surroundings. We propose BeWater, a fully distributed walking protocol that achieves gathering thanks to city information like street length, number of restaurants, number of lanes, or street names. Even though using only one of these observables performs poorly, we show that combining them in more advanced tactics rapidly leads to groups of significant sizes. To do so, our work leverages OpenStreetMap data to perform experiments on several real-world cities.
arxiv url to come.
]]>Communications of the ACM Blog
Instead of researching topics that serve the Machine, Matthieu Latapy advocates for more research on ecological and social movement needs.
]]>In France International Conference on Complex Systems (FrCCS), 2025
We propose a method to detect communities in multi-relational networks, based on a graph neural network pipeline. The method allows to target areas where communities are consensual over the different modes of the network, which are processed as different networks in the pipeline. This is done by combining the outcomes of multiple simple Graph Neural Networks, applied on each of the graphs representing different forms of interactions between users of the social platform. The method is validated on a synthetic benchmark, as a first step for further improvements. In particular, the flexible architecture of the pipeline allows to swap its subparts and create variants of community detection.
]]>In IEEE International Conference on Data Mining (ICDM), 2025
Detecting anomalies in link streams that represent various kinds of interactions is an important research topic with crucial applications. Because of the lack of ground truth data, proposed methods are mostly evaluated through their ability to detect randomly injected links. In contrast with most proposed methods, that rely on complex approaches raising computational and/or interpretability issues, we show here that trivial graph features and classical learning techniques are sufficient to detect such anomalies extremely well. This basic approach has very low computational costs and it leads to easily interpretable results. It also has many other desirable properties that we study through an extensive set of experiments. We conclude that detection methods should now target more complex kinds of anomalies.
arxiv url to come
]]>In International Conference on Complex Networks. Cham: Springer Nature Switzerland, 2025 (CompleNet2025)
We improve a flocking model on street networks introduced in a previous paper. We expand the field of vision of walkers, making the model more realistic. Under such conditions, we obtain groups of walkers whose gathering times and robustness to break ups are better than previous results. We explain such improvements because the alignment rule with vision guaranties walkers do not split into divergent directions at intersections anymore, and because the attraction rule with vision gathers distant groups. This paves the way to a better understanding of events where walkers have collective decentralized goals, like protests.
]]>PLoS One, e0338486, 20 (12), 2025
This study explores a model for the co-evolution of opinions and groups, related to opinion polarization and group identity in opinion dynamics. While traditional models focus on pairwise interactions, we incorporate the notion of groups thereby allowing agents to identify other agents as in-group or out-group. By modifying key parameters we examine how understanding interactions as in-group or out-group affects the convergence or divergence of opinions. Our findings reveal that attitude towards out-group plays a leading role in the polarization of such systems, while in-group interactions can temper extreme opinion shifts or even fragment groups. This model offers an understanding of the complex interplay between group identification and polarization, providing a contribution to ongoing debates about segregation and sectarianism in public and private spheres.
]]>Jeudi 22 janvier 2026 à 14h en salle 26-00/534
Guillaume Chelius, directeur du Programme Inria Quadrant (PIQ), viendra nous parler de ce programme très spécifique de soutien aux démarches de recherche à risque et à impact dans le numérique. Le programme, ouvert depuis fin 2024, accompagne et finance des scientifiques désireux d’engager des projets de recherche à risque et à impact dans le domaine des sciences et technologies du numérique, de ses fondements à ses usages. PIQ s’adresse à tous les scientifiques issus de l’ensemble les établissements publics de l’Enseignement Supérieur et de la recherche, chercheurs, enseignants chercheurs, ingénieurs de recherche. PIQ est ouvert à toute la communauté scientifique académique. PIQ accompagne des projets orientés « porteurs », de durée variable, de six mois à plus de trois ans, et sur des budgets pouvant aller de quelques dizaines milliers d’euros à plus d’un million d’euros. Aucun format n’est imposé. Durant cette session, Guillaume nous présentera le positionnement particulier du programme, la manière dont il s’empare de ces notions de risque et d’impact, ainsi que ses modalités spécifiques d’opération, son processus de candidature et l’ingénierie projet qu’il déploie. Les actualités du programme seront abordées, notamment à travers des projets accompagnés.
]]>Jeudi 18 décembre 2025 à 14h en salle 26-00/534
Depuis une vingtaine d’années, la pratique du blocage prend une place croissante dans les mouvements sociaux à l’échelle internationale, des mobilisations altermondialistes au mouvement pro-démocratie à Hong Kong, en passant par les Gilets Jaunes et l’appel « Bloquons tout ! » en France. Si ce mode d’action tranche par sa nouveauté, il s’inscrit également dans une longue histoire de subversion par la perturbation des flux, dont on trouve déjà des manifestations au XVIIe siècle. Cet exposé étudiera la manière dont ont progressivement émergé des tactiques fondées sur l’idée que le pouvoir devait être attaqué dans ses réseaux techniques et ses infrastructures, du XIXe siècle à nos jours.
]]>