|Juliette Rouchier, Research Fellow (CR1), CNRS|
2 rue de la Charité, 13002 Marseille, France
|juliette dot rouchier at univmed dot fr|
My work is part of a research area called cognitive economics. The tool which is at the centre of my work is multi-agent social simulation or agent-based simulation. I also personally need to perform regular field research to study actors' behaviours. At the moment, I mainly work on markets for professionals (like wholesale markets) organised through pair-wise interactions. I consider the forms of negotiations it implies, the trust (/ mistrust) and reputation that are the core of long-term relations and enable agents to anticipate in a fluctuating, hardly predictable, environment.
- phD at CIRAD, FRANCE, in François Bousquet's team, GREEN (1996-2000).
- post-Doc at CPM, Manchester Metropolitan University, UK, with Scott Moss and Bruce Edmonds (2000-2001).
Cognitive economics: The aim of this movement is to give an quite accurate, though minimal, representation of the rationality of actors while performing economic activities. Since a lot of diverse meanings can be put under the phrase "economic activities", quite a few branches and research topics are considered as being part of cognitive economics, with shallow frontiers that depend on who draws them. One can consider that it contains a lot of evolutionary approaches putting an emphasis on interactions (like evolutionary game-theory or network analysis in economics), works gathering data on actual behaviour as a central element (like experimental economics) and that new theoretical branches that tend to erase the notion of optimising agent to focus on path dependent situation are also included (like conventional economics or New Institutional Economics or systemics). Focus is put on context, interaction, negotiation, procedural agency, misunderstanding and limits in computabilities.
Multi-agent social simulation: One of the recent trend in the study of society in a new approach to structural properties, based on the creation of artificial worlds. These universes are conceived as softwares organising the interaction of diverse agents and objects that follow a common dynamics. An agent is usually defined by goals, an ability to act, an ability to perceive, stock and interpret information and rules that enables it to choose an action when facing a given situation. One can also say that an agent has a rationality. Agents interact: they directly send messages to each other through a communication framework, or they act on their environment and others perceive it. When performing simulations, one needs to define the multi-agent universe: agents and their rationality, the global institution (information gathering processes, possible actions and common rules, temporal environment), and an initial state from which the dynamics can be run. One also needs to have observable variables to describe the states of the universe. When defining the universe, one needs a lot of parameters (of any kind: number of agents, sensitivity of their perception, memory length, rules of attribution,...). The discovery of the dynamics properties is a discovery of the influncee of the change of one parameter over the global results. Simulations are mainly used for sensitivity analysis in the created complex system. After that one needs to interpret using data from the outside world, which cqn get quite complex.
my french page for publications