Team : SMA
Agent-based approaches to the study of human behaviors
Broadly speaking, my work aims to contribute to the study of the human systems' behavior. By human system, I mean any system composed of one (individual) or several (collective) human beings, in interaction with each other or with some environment.
I will show how computational approaches -- mostly multi-agent systems -- could contribute to a better understanding of some human behaviors. Human systems are by nature complex, delicate to study for a human (studying another human; the axiological neutrality at stake), free not to follow scientific rules and needs to account for explanation and meanings of their actions. To address all these peculiarities, I propos a general methodology, based on my psychomimetism theory, to (1) derive the computational model (and especially of the individual behaviors) from theories, facts, and data from Human and Social Sciences (HSS; psychology, sociology, economics, etc.) that are robust and empirically grounded;
(2) use a data-driven approach, and validate the model on real data;
and (3) at the implementation level, ensure that the program stays compatible as much as possible with the conceptual model and do not violate its core principles (e.g. like bounded rationality).
Several projects from various fields will illustrate my approach, all of them agent-based: Coban (marketing, psychology) is a model dynamics of opinions, diffusion of innovation (e.g. the launch of a new product) and Polias (social psychology) a model of attitude dynamics (e.g. for a population in the middle of the war conflict). HappyWork proposes an agent-based model of job satisfaction for an employee, including a model of employee’s activities at work within a firm. WorkSim is a one of the most comprehensive agent-based model of the French labor market, and will be used to asses various labor policies. For each project, I will present the model, the validation procedure (all models are calibrated on real data) and simulation simulations, in order the assess the benefits of these contributions.
: 12/09/2015 - 14hJury members
Philippe Mathieu, Université de Lille 1 - LIFL (Rapporteur)
Christophe Sibertin-Blanc Université Toulouse 1 - IRIT (Rapporteur)
Jaime Sichman Escola Politécnica da Universidade de São Paulo, Brasil (Rapporteur)
Amal El Fallah Seghrouchni, UPMC - LIP6
Patrice Perny, UPMC - LIP6
Matteo Richiardi,University of Torino & University of Oxford
1 PhD student (Supervision / Co-supervision)
- CHAPUIS Kevin : Happywork : modélistaion multi-agents de la satisfaction au travail.
- BROUSMICHE Kei-Léo : Modélisation et simulation de la formation et de la dynamique des attitudes basées sur les croyances.
- GOUDET Olivier : La modélisation multi-agent du marché du travail français.
- LEWKOVICZ Zach : Modélisation et simulation du marché du travail à l'aide de systèmes multi-agents.
- THIRIOT Samuel : Vers une modélisation plus réaliste de la diffusion d'innovations à l'aide de la simulation multi-agents.