Social polarization in the metropolitan area of Marseille. Modelling uncertain knowledge with probabilistic and possibilistic networks

Giovanni Fusco (1) , Cristina Cao (1) , Didier Dubois (2, 3) , Henri Prade (2, 3) , Floriane Scarella (1) , Andrea G. B. Tettamanzi (4)

1 ESPACE - Études des Structures, des Processus d’Adaptation et des Changements de l’Espace
2 IRIT-ADRIA - Argumentation, Décision, Raisonnement, Incertitude et Apprentissage
3 CNRS - Centre National de la Recherche Scientifique
4 WIMMICS - Web-Instrumented Man-Machine Interactions, Communities and Semantics

A Bayesian Network and a Possibilistic Network are used to produce trend scenarios of social polarization in the metropolitan area of Marseille (France). Both scenarios are based on uncertain knowledge of relationships among variables and produce uncertain evaluations of future social polarization. We show that probabilistic models should not be used just to infer most probable outcomes, as these would give a fallacious impression of certain knowledge. The possibilistic model produces more uncertainty-laden results which are coherent with model uncertainties and respect elicited values of possibilities. Results of the two models converge when probability values are "degraded".

Mots clés: Social polarization; Uncertainty; Bayesian networks; Possibilistic networks; Marseille

Domaines: Géographie; Intelligence artificielle [cs.AI]

Giovanni Fusco, Cristina Cao, Didier Dubois, Henri Prade, Floriane Scarella, et al.. Social polarization in the metropolitan area of Marseille. Modelling uncertain knowledge with probabilistic and possibilistic networks. Plurimondi, 2015, 8 (17), pp.117-126. ⟨hal-01531095⟩