Interaction Routines and Selfish Behaviours in an Artificial Market

This paper describes an application of multi-agent systems to economics. The issue we address in our work is the actual functionning of a distributed market in terms of information acquisition for agents. We show here possible links between individual information acquisition and global results in the fluidity of exchanges. A multi-agent model has been built on the basis of field observations and interiews led at the Marseille Fruits and Vegetables wholesale market in 2002-2003. Two types of artificial agents interact, wholesale sellers and retailers, surrounded by an environment of exogenous supply and prices. Retailers can choose wholesale sellers according to two logic: either following some relational loyalty or searching for best prices (which takes time). We study the influence of the number of agents with each rationality so that to see the importance of both categories in the good supply of consumers and the limitation of garbage.