Complex Systems Paradigms for Integrating Intelligent Systems: A Game Theoretic Approach

Complex systems have provided not only an analytic view that computational intelligence could be attained at a critical point (edge of chaos) where a phase transition takes place, but also a synthetic view that computational intelligence could be embedded in the field where an open and evolutionary environment for sel.sh agents will lead to collective phenomena. In the synthetic view, using complex systems themselves for intelligent systems, such as DNA computing (we focus on immunity-based computing in another Chapter of this volume), grid computing, and parasitic computing, is another important paradigm.

This Chapter investigates the first step towards embedding computational intelligence in the Internet field by selfish agents, namely, whether selfish agents can ever cooperate and converge on some tasks. Selfish routing and task allocation have been studied extensively in the computational game community, but can intelligent tasks be done or can agents ever take care of themselves in the first place? We first pose the problem of self maintenance in an agent population, and then use a game theoretic approach to test whether cooperation would occur or under what conditions cooperation will occur.