Moduleco
2. Markets are social systems with complex dynamics
and both individual and collective learning

page précédente sommaire page suivante


Main concepts in dynamic modeling of complex adaptive systems : attractors, hedge of chaos, critical system (phase transition), self-organized critical system.

Specific concepts from multi-agents literature : organization, emergence, constraints, interaction and communication structures. 

The aim of those models is to abstract general (analogue) formal properties from social and natural systems, independently of the specific context. The key feature is viewing the emerging order as a product of the system dynamics (system attractor), and more specifically of its element interactions. ( see Bonabeau [1994]).

One can associate some emergence proprieties with the notion of collective learning. But, in Economics, we must take into account the cognitive capacities of individual agents. With a special emphasis upon the revision rules and beliefs process, in a situation of incomplete information (individual learning).


Multi-agent systems are popular in computing science and within the domain of artificial intelligence. Generally, the Agent is defined as referring to a software component which is capable of acting in order to accomplish a specific task, such as finding information on the Internet.

In the theoretical research Area, Multi-agent systems were introduced in order to simulate "life" and investigate "historical" questions (like life emergence, or species co-evolution by experiments  "in silico", (Langton [1995.]) as well as to simulate with very simple rules the complex behavior of animals, like flocks of birds (Reynolds [1987]). The interest in Multi-agent systems in Economics is strongly correlated with the Artificial life Movement (Lane [1993]).

The most famous Multi-agent systems is SWARM, initiated by Langton [1995]. Applications to the economic field are reviewed by   Luna, Stefansson  [2000]. Some others (Parker's Ascape, Valente's LSD, among Others...) are listed at the end of bibliography

This paper is dedicated to the presentation of   MODULECO, a Multi-agent system, built in order to simulate a market phenomenon in a "virtual lab", and more generally models of social choices involving both individual learning and collective learning.


Denis.Phan@enst-bretagne.fr ;   Antoine Beugnard@enst-bretagne.fr