We have presented analysis of a simulation of mobile agents working together to map a network of computers. While agents that cooperate can collectively solve problems more quickly than agents acting alone, optimizing such a system for maximal performance is subtle. Strategies that appear profitable from the perspective of each individual agent are not necessarily efficient for the system as a whole when pursued by multiple, interacting agents. A major challenge for designers of multi-agent systems is to encourage diversity of behavior in a population of cooperating agents.
