In this paper, we show how self-* mechanisms give rise to complex but predictable and therefore steerable global system behavior in a co-operative computing environment. We simulate the operation of and interactions between a set of networked autonomic devices. These are used as access points to a number of services, have the ability to accept or delegate execution of the associated tasks, and can adjust their internal state in response to the demand. We study the emergence of co-operation, and find that it spontaneously occurs when specific conditions are met that allow individual devices to focus on performing a single task, sacrificing their ability to efficiently perform others.