Multi-Agents for modeling Complex Systems  (MA4CS'05)

Satellite Workshop of
the European Conference on Complex Systems (ECCS'05)
Paris, France,
14-18 November 2005


Registration :
See the ECCSRegistration
   

   
Presentation
       
Submission
Publication

Important Dates
Scientific Committee



Accepted P


































































































   

Invited Talks



Micro-economic multi-agent systems that trade populations
Pr. Philippe de Wilde
Heriot Watt University, Edinburgh, Scotland.
Abstract
The decisions of economic agents, and their effect on market equilibrium, have been the basis of micro-economic theory for more than half a century. Goods are produced, priced and traded. In this talk we present a theory of the exchange of goods that are part of an evolving population. Software agents are an example of such goods. They can be evolved using genetic algorithms. Economic agents, the enterprises, will soon possess populations of software agents. The software agents will specify and implement business processes and services. Instead of trading single products, it now becomes possible to trade part of populations. In this talk, we show how to organize this. This system possesses both structural and functional complexity. The structural complexity is provided by the large evolving populations in multiple habitats. The functional complexity
of the populations is in the economic value they represent to the organizations that own them. We have designed the first mechanism that allows efficient trade to take place in this micro-economic multi-agent system of populations. We illustrate our mechanism is a simulation of a market of software providers. The exchange of populations naturally emerges as a successful, profitable behavior of the software providers. We demonstrate two results: a simple mechanism is sufficient to give rise to an exchange in a software market, and knowledge exchange increases the efficiency of the market.


Multiagent systems as artificial societies
Pr. Jacques Ferber
LIRMM, University of Montpellier II, France

Abstract

The multiagent approach (MAS) has seen a widespread audience in the last decade as a tool for analyzing complex systems, in the natural, human and social sciences, and most notably at the intersection of these domains (e.g. management of sustainable resources). However, multiagent systems tend to be mixed up with other modeling and problem solving methods (differential equations, cellular automata, CSP, taboo techniques, simulated annealing, etc.) as a tool among others, without understanding what makes the specificity and originality of the MAS approach.
From a multi-perspective approach, we will show how MAS, as virtual societies building tools, give an integrated understanding of modeling and problem solving issues, by allowing one to use both an abductive/deductive reasoning process to analyze a phenomenon and an analogical thinking mode to give a meaning to it. In this regard, MAS propose a profound understanding of complex systems, in a quite specific and innovative way.
We will see from several examples how MAS may be regarded not as modeling tools or as solving problems platforms, but as tools for building artificial societies in which we may explore ideas and observe their consequences and in which we may capture the complexity of natural systems along several dimensions.



Tribal Programming for a Better World
Dr. David Hales
Dept. of Computer Science, University of Bologna, Italy

Abstract
A number of recent agent-based simulation models, inspired by sociological concepts, claim to demonstrate a novel approach for constructing co-operative and robust self-organising groupings (or tribes). But these models tend to be highly abstract, focusing on simple games (like the Prisoner's Dilemma). Even application orientated work using similar approaches presents simulations related to specific application domains. If we are to take this "tribal approach" seriously as a way of constructing self-organising distributed information systems we need a larger and more general vision coupled with actual implementations of useful applications that work for real "in the wild". In this talk we outline some of the previous work based on "tags" and evolving networks and then sketch our more general "tribal vision". Finally we discuss on-going work towards implementations and argue that some existing and successful applications embody aspects of the vision - though not constructed with this in mind.



Multi-Agent Systems Featuring Emergence and Evaluation
Dr. Yves Demazeau
Leibniz, Institut IMAG, Grenoble, France

Abstract

Multi-Agent Systems are one of the most prominent and attractive paradigms in computer science at the beginning of the new millennium. The technologies, methods, and theories of agents and Multi-Agent Systems are currently contributing to many diverse domains such as user interfaces, electronic commerce, robotics, computer mediated collaboration, computer games, education and training, ubiquitous computing, and social simulation. They enable to co-ordinate distributed computing, to simulate open systems in a decentralized way. They not only are a very promising technology, but are also emerging as a new way of thinking, a conceptual paradigm for analyzing problems and for designing systems, for dealing with complexity, distribution, and interactivity, while providing a new perspective on computing and intelligence. The presentation will explore some state of the art in Multi-Agent Systems from the Emergence and Evaluation perspective and will discuss prospects for the domain along the VOWELS approach we have been developing for several years. In a first part, we will expose the characteristics of Multi-Agent Systems and Applications and we will position the domain within both landscapes of computer science and complex systems. In a second part, we will present Emergence and Evaluation dimensions as being two major cornerstones of future Multi-Agent Systems. We will present some preliminary results and we will discuss the methodological implications of these dimensions. In a last part, we will present and discuss some related applications, including a first one in modular robotics and a second one in artistic creation.