Global behavior in a population of adaptive competitive agents

We present computer simulations and analysis for the global behavior arisin g from a population of heterogeneous social agents acting with bounded rati onality. The particular model studied, termed the "bar-attendance" model, o ffers a simple paradigm for such complex adaptive systems involving competi tive agents. The model considers p adaptive agents, each possessing n predi ction rules chosen randomly from a pool of N, who attempt to attend a bar w here the seating capacity is s. The global attendance time-series x(t) has a mean near, but not equal to, s Surprisingly, the standard deviation or "v olatility" of x(t) can show a minimum with increasing adaptability of the i ndividual agents. Various arguments based on random walk models are discuss ed. It is shown that effects of crowding have to be included in order to un derstand the volatility in this system.