In open multi-agent systems (MASs), heterogeneous agents representing different stakeholders can join, leave and use the MAS infrastructure at any time in order to achieve their goals. Due to the nature of open MASs, social approaches have been proposed in order to study the dynamics of these systems. These solutions are based on the idea of creating agent organizations inside the MASs, regulated by social norms, in order to mediate the behavior of the different agents inside them. Agent environments are first-class abstractions that: i) provide the surrounding conditions for agents to exist which means that the environment is an essential part of every multi-agent system and ii) provide an exploitable design abstraction for building multi-agent system applications. The talk basically consists of three parts. In the first part, I will argue about the necessity of first-class representation of MASs. In particular, I will explain which properties of agent environments are useful for agent organizations and what problems they do solve.
In the second part, I will present MANET (Multi-Agent Normative EnvironmenTs). MANET is the meta-model that I have suggested in order to specify the institutional (organizational) part of first-class agent environments. Finally, the last part discusses how first-class normative MASs can be applied to solve real and complex problems. I will analyze the case of an Electronic Energy Marketplace and I will present a first approach towards modeling the system using MANET meta-model.
- Acerca del IIIA