Modélisation et simulation en sciences humaines et sociales

- Frédéric Amblard, Informatique SMA, IRIT Toulouse,

Titre : “Influence sociale et morphogénèse des réseaux sociaux"

- Sylvie Huet (IRSTEA, Clermont-Ferrand)

Titre : “Various polarization types emerging from a population of ego-involved and non ego-involved agents”

Résumé : (Huet & Deffuant, 2010) showed polarization on minor issues can be reproduced by a dynamics inspired from experiments (Wood and al, 1996) considering a population of self-involved agents on a main issue. This paper studies the impact of a mixt population of self-involved and not self-involved agents. The dynamics of self-involved agents is based on (Huet & Deffuant, 2010) meaning agents are attracted on every issues if they are close onto the main issue but reject their peer’s opinion if they are only close onto secondary issue. The dynamics of non self-involved agents is a simple 2D bounded confidence model (Deffuant et al, 2001). From the analysis of the simulations from a large experimental design, we observe the emergence of various types of “extreme minor clusters”. They can lead the population to a polarization on minor issues, but also on fundamental ones (as well as bipolarization and multi-polarization). A very strong in some case, and a very long transitory state in others cases of in-group polarization on secondary issues has been also exhibited. We define when polarization and depolarization occur on each dimension. Moreover, only few self-involved agents in the population are sufficient to obtain a smaller number of clusters compared to a population of only non self-involved agents. Trajectories of agents are studied to understand how the various polarizations types emerge.

- Cyrille Imbert (Archives Poincaré, CNRS, UL)

Titre : “No need for a secret ballot? Reputational cascades in deliberative groups”

À partir d’un travail avec Christine Bourjot (LORIA, Université de Lorraine), Thomas Boyer-Kassem (University of Tilburg), Vincent Chevrier (LORIA, Université de Lorraine)

Résumé : People, when discussing, do not always speak their mind but may partly falsify their opinion based on what they perceive of what other people think. We study with a deliberately crude model how much this is a worry within small deliberative groups. We first show that, even with opinion falsification, agents can sometimes converge towards what they would have agreed upon without falsification. We further analysis factors that may block this convergence process, in particular the discretization of the space of opinions that agents may publicly express. We investigate how much the corresponding decisions publicly taken by the group agree with those that they would have taken if they had taken a secret ballot and assess how much these decisions can be biased, how much they look biased to the voters themselves and how much they look biased for external witnesses.

- Samuel Martin (CRAN, LORIA, Université de Lorraine)

Titre : “Continuous opinions and discrete actions in social networks: an analysis”

- Philippe Mathieu (équipe SMAC du CRIStAL, Université Lille 1)

Titre : “Simulation dynamique de trafic routier à partir de données réelles”

Résumé : La simulation microscopique de trafic routier est depuis longtemps un domaine d’application privilégié des Systèmes Multi-Agents. L’approche centrée individus permet en effet d’intégrer la variété comportementale nécessaire pour rendre compte de situations réelles. Plus récemment, des bases de données géographiques fournissent des informations environnementales précises dans des formats ouverts, offrant ainsi la possibilité de réaliser facilement des simulateurs de trafic à base d’agents, informés en direct de modifications de conditions de circulation. En utilisant ces données et grâce à l’adaptativité des SMA, il est ainsi possible de construire des systèmes d’aide à la décision capables d’intégrer des changements environne- mentaux et comportementaux de manière immédiate ; et de comparer ainsi divers scénarios construits sur des hypothèses différentes en matière d’acteurs, de comportements, d’environnement et de flux. Nous décrivons ici un processus permettant la réalisation d’un tel outil.

- Jean-Pierre Nadal, Laboratoire de Physique Statistique, ENS, Paris & Centre d'Analyse et de Mathématique Sociales, EHESS, Paris

Titre : “Riot contagion: data-driven epidemic modeling of the 2005 wave in France”

Résumé : About 10 years ago, after two youths died while trying to escape from a police control on October 27, 2005, riots spread in poor neighborhoods in French cities,  hitting more than 800 municipalities and lasting about 3 weeks. Thanks to an access to a detailed account of these events, we analyzed the dynamics of these riots.  In this talk, I will show that a parsimonious (less than 10 free parameters) data driven epidemic-like model, taking into account both local (within cities) and non-local (through geographic proximity or media) contagion, allows to reproduce the full spatio-temporal dynamics of the riots.  I will make explicit the specificity of the model as compared to the modeling of the spread of infectious diseases. Finally, I will discuss the methodology in the broader context of modeling social phenomena, showing in particular how modeling provides a kind of regularization of the data. In the case of the riots in the suburbs of Paris, this allows to visualize the wave propagating around Paris starting from the triggering event in Clichy-sous-Bois.

This talk is based on works done in collaboration with Laurent Bonnasse-Gahot, Henri Berestycki, Marie-Aude Depuiset, Mirta B Gordon, Sebastian Roche and Nancy Rodriguez.

- Denis Phan

Titre : TBA

- Denise Pumain (Géographie – Cités, UMR 8504)

"Modéliser les trajectoires urbaines