Agent Societies to Design and Implement Complex Systems

Software agents are usually exploited to model and implement complex and unpredictable systems, paying attention to the interactions between component agents, in particular in open and heterogeneous systems. Agents may face situations that was not completely predicted by their developers, which must rely on their adaptability derived from their features of proactivity, reactivity and sociality. But this could not be enough: social norms must be enforced to rule the agent interactions in open environments. Moreover, social norms enable to enact local policies as well, i.e. policies defined by the environment where agents live. To this purpose, the abstraction of agent societies has been proposed and applied to complex systems.In this paper we address the development of agent societies, in particular considering methodologies and infrastructures. We will present the existing methodologies and point out the degree of the available specific support for societies. Moreover, we will explain that appropriate support must be provided also by underlying infrastructures. Further, we will show that there is a gap between methodologies and infrastructures that leads to fragmented solutions; we will sketch some proposals to bridge such a gap by means of meta-models.