I am an associate professor, First Class with Habilitation, at Sorbonne University Sciences, Paris, France.
I work at the Computer Science Laboratory (LIP6), to do research on artificial intelligence and cognitive sciences.
My research focuses on the study of human complex systems, that is trying to model, simulate and understand human behaviors, economies and societies.
How can we fight unemployment more effectively? Will automation via robots and algorithms take our jobs? How can we find new ways to fight global warming and preserve biodiversity? These questions have in common their complexity in the sense of complex systems: by definition they are difficult to study and impossible to predict.
In my research , I designed a method to address these complex, yet crucial issues for our societies, using a particular form of artificial intelligence called multi-agent simulation. The idea is to provide an innovative method for designing and evaluating policies (of all kinds) in a virtual world, before their implementation in the real world.
My approach is strongly multidisciplinary and relies heavily on modelling of individual and social behaviors, from theories derived from economics, sociology, cognitive and social psychology, biology, environmental sciences, etc.
Agent-based models enable to design decision-aid tools that are fully explainable Artificial Intelligence systems, to help the decision maker to build better products or design better policies.
Look at the research section to find about my projects on labor markets, digital automation, environment and climat change or diffusion of innovations.
You may also watch my TEDx Talk (in French) that introduces my research on AI and complex systems.
An innovative approach to reduce the human impact on Climate and Earth
How can we effectively combat global warming and preserve biodiversity? The TerraNeon project aims to design a decision-making tool to help companies and public bodies reduce their environmental footprint efficiently.
With TerraNeon, you can test different ideas and compare their effects, costs and social acceptability, thanks to a clear interface based on scientifically founded indicators.
Studying the impact of digital automation on labor with a multi-agent simulation
This project is a collaboration between the Multi-Agent Systems team of LIP6, and Pôle Emploi, the French leading public employment service and the Panthéon-Assas University (CRED). NumJobs aims to model and simulate the impact of digital technology, and in particular of Artificial Intelligence (AI), on employment, using a multi-agent system.
To do so, we have proposed a complete multi-agent model of entire French economy, including :
An agent-based model of the French labor market
Started in 2006, WorkSim is the most comprehensive labor market simulator available, including individual and corporate behavior, institutions and labor laws. One of the strong points is the endogenization of risk when choosing between a fixed-term or permanent job. We have calibrated it on real data and used it to evaluate public policies, such as the generation contract. In 2016, we were the only ones to propose a quantitative evaluation of the "El Khomri" labor law, proposed by the French government in 2016. This work has been cited and recognized by the IMF and the European Commission.
Major results
A novel agent-based approach to study the diffusion of innovations
Partnership with Orange Labs (Orange group, World Telecommunication leader)(2005-2009)
We designed COBAN, one of the most comprehensive and complete simulation model for the diffusion of any kind of innovation. Coban encodes people's beliefs through associative networks, in order to compute agents attitude and make adoption decision. We also proposed the first method to generate automatically a social network generator from statistical field data.
We also proposed the first method to generate automatically a social network generator from statistical field data.
Major results
How to study opinion and attitude dynamics with an agent-based model
Partnership with Airbus Defence and Space (2012-2015)
How would a population react when they are faced with information whether from a government or a company (ads) about e.g. an action, a product, an innovation ? This could be tackled through a core concept in social psychology: attitude dynamics. In this project, we propose a multi-agent based simulation model that will help us to better comprehend attitudes dynamics. Our goal is to articulate the cognitive and emotional dimensions of attitudes within a population.
Major results