Crowd-Anticrowd Theory of Collective Dynamics in Competitive, Multi-Agent Populations and Networks

Wednesday, November 5, 2003 - 8:45am - 9:20am
Keller 3-180
Neil Johnson (University of Oxford)
I present a crowd-based theory describing the collective behavior within a generic multi-agent population with limited resources. These multi-agent systems -- whose binary versions we refer to as B-A-R (Binary Agent Resource) collectives -- have a dynamical evolution which is determined by the aggregate action of the heterogeneous, adaptive agent population. Accounting for the strong correlations between agents' strategies, yields an accurate description of the system's dynamics in terms of the 'Crowd-Anticrowd' theory. This theory can incorporate the effects of an underlying network within the population, and is not just limited to the El Farol Problem and the Minority Game. By considering a variety of examples, I will show that the Crowd-Anticrowd theory offers a powerful approach to understanding the dynamical behavior of a wide class of agent-based Complex Systems [1].

[1] For applications in the financial domain, see 'Financial Market Complexity' (Oxford University Press, June 2003).