Learning for Social Systems Modeling
The macroscopic or overall behavior of social systems depends on the characteristics of individuals (their idiosyncratic preferences, their learning abilities and types of interactions) in a way that is not predictable from the individual data. In our theoretical work, we try to surpass the models that use symmetric and positive interactions (where individuals tend to adopt mimetic behavior or compliance with the other), in order to deal with cases such as public property where interactions are not symmetric. This allows us to predict cycles in the system evolution or hysteresis effects depending on the public policies. We also work on empirical data, such as the development of delinquency and its modeling.
We are also interested in the study and modeling of online social networks, and particularly the dissemination of information in these networks and the modeling of their dynamics. We have proposed a new model of information dissemination based on several factors: the content of the disseminated information, the network topology, the propensity of each user that broadcasts.