Asynchronous Time Evolution in an Artificial Society Model

"Artificial society" refers to an agent-based simulation model used to discover global social structures and collective behavior produced by simple local rules and interaction mechanisms. In most artificial society discrete-event simulation models, synchronous time evolution is used to drive the actions and interactions of the landscape and agents. Although for some applications synchronous time evolution is the correct modeling approach, other applications are better suited for asynchronous time evolution. In this paper we demonstrate that very different behavior can be observed in a typical artificial society model if agent events occur asynchronously. Using an adaptation of the artificial society model defined by Epstein and Axtell, we describe the implementation of asynchronous time evolution in a discrete-event simulation model. With output from this model, we show that the use of asynchronous time evolution can eliminate potential simulation artifacts produced using synchronous time evolution. Since the use of discrete-event simulation can produce an associated loss in computational performance, we also discuss means of improving the performance of the artificial society simulation model. We provide results demonstrating that acceptable computational performance for asynchronous time evolution can be achieved using an appropriate event list implementation.