The question of how software agents can learn strategic behaviors in complex, continually changing, multi-agent environments is not only a challenging forefront of theoretical research, but potentially of immense practical importance as well. In such systems, it would be difficult at best to hand-code fixed strategies that would always perform well with high confidence, especially if the other agents in the environment change their behaviors over time using adaptive learning algorithms. Hence, the need for learning as a component of overall agent programming methodology is readily apparent. We expect this to be particularly true in the domain of “agent economies,” in which large populations of agents engage in various forms of economic activity with each other (and possibly with humans as well).