UMR 5672

logo de l'ENS de Lyon
logo du CNRS logo UCBL
You are here: Home / Teams / SIgnals, SYstems & PHysics / Research Topics / Analyses of social and human activities

Analyses of social and human activities

The avalanche of digital data tracing social activities opens the way to combine data analysis and modeling with social science studies. Along this line, we study data representing human interaction or behaviour, or date pertaining to transportation (e.g., BSS or traffic on roads); in all these cases, the dynamics of the data (for instance of the networks uwed to represent them) is crucial and tailored methods are developed leveraging on methods in physics and/or signal processing.

Study of Vélo’v data and transportation.

(LIRIS (INSA), LET (Lyon 2), CMW et EVS (ENS Lyon)). From 2008 on, we had access to data about uses of Lyon’s Vélo’v system of shared bicycles, the first large scale bike sharing system (BSS) of Europe, and studied the mobility with BSS. We first studied the rhythms of use and statistical model for that. This was combined with spatial analysis of the trips to draw pictures of their use in the city. The data show that bicycles compete with the car in terms of speed in downtown Lyon. The work goes on in a ANR project VEL’INNOV. Other works model part of the use of the system, or discuss the need for mapping tools to display the data. Relying on network theory, we exhibited the different rhythms of the stations. Finally, in a paper on spatial networks (such as transportation networks), it was shown that a spatial hierarchy can emerge in network as a large-scale consequence of local cost-benefit considerations.

Human face-to-face interaction network.

Using active RFID tag and a dedicated experimental apparatus thanks to the collaboration to the Sociopatterns project (http://www.sociopatterns.org), data of time-resolved person-to-person interaction networks were collected at conferences, schools, hospital wards to analyze their dynamics. Behavioral characteristics can hence be studied, for instance in a school to quantify interaction between children.  Having only one realization of these networks, we developed a bootstrapping method using constrained graphs to probe with statistical confidence the behavior of a group in such a network.

Social systems and human behaviour.

Developing new sociological concepts. (Médialab, Sciences Po, Paris) A naive approach of social systems by physicists would be to start with interacting “social atoms” to probe collective phenomena, “emerging” from the microscopic level. However, for social systems, isolated (“atomic”) individuals do not exist. Therefore we argue that it is more interesting to use “collecting” entities instead of individual and collective levels, and develop this idea through the use of heterogeneous networks.

Scientometrics. This approach is tested on scientometric data (scientific articles). Mapping of scientific institutions was developed [599], for instance for ENSL and CNRS. The interdisciplinary practices of 600 laboratories were studied through their publications. Scientometrics is also useful to study scientific fields, for instance the “complex systems” domain, showing that it does not arise from a single universal theory, but from shared computational methods and concepts on self-organization. A study carried out on 7000 CNRS scientists regarding their public engagement activities, was propagated with a Special Issue in Public Understanding of Science.

Model for conference submission behavior. An empirical study of several datasets has revealed some ‘universal’ features in the temporal process of electronic submissions to conferences, leading to the proposition of a simple predictive model.

Collective Free Improvisation offers a situation of human interactions without any a priori reference frame. For musical production process, a model and studio experiments were done to gain a better understanding of emerging collective structures.

Contacts

Patrice Abry, Pierre Borgnat, Patrick Flandrin, Nicolas Garnier, Pablo Jensen

Filed under: