|InterJournal Complex Systems, 1654
|Manuscript Number: |
Submission Date: 2006
|Cooperation networks: endogeneity and complexity|
This paper employs insights from Complex Systems literature to develop a computational model of endogenous strategic network formation. Artificial Adaptive Agents, implemented as finite state automata, play a modified two-player Iterated Prisoner's Dilemma game with an option to further develop the interaction space as part of their strategy. Several insights result from this relatively minor modification: first, I find that network formation is a necessary condition for cooperation to be sustainable but that both the frequency of interaction and the degree to which edge formation impacts agent mixing are both necessary conditions for cooperative networks. Second, within the FSA-modified IPD frame-work, a rich ecology of agents and network topologies is observed, with consequent payoff symmetry and network `purity' seen to be further contributors to robust cooperative networks. Third, the dynamics of the strategic system under network formation are investigated and show that initially simple dynamics with small interaction length between agents gives way to complex, a-periodic dynamics when interaction lengths are increased by a single step. Subsequent analysis of the developing network topology through time shows scaling behaviour in both time and space, indicating the attainment of a self-organised critical state, and thus apparently driving the complex dynamics of the overall strategic system.
|Submit referee report/comment|