Our main issue is to design complex systems whose global behaviour emerges from the behaviours and interactions of the agents composing them. Each agent has a local objective and a behaviour built upon a cooperative social attitude. Since 1995, we study an approach for designing complex adaptive systems, which is based on adaptive multi-agent systems and emergence. A theory, called AMAS theory (for Adaptve Multi-Agent Systems) was established for this [Camps, 96], [Georgé, 04]. It gives local criteria for designing agents which enable an organisation to emerge within the system and therefore, the emergence of its global function. System adaptation, performed by self-organisation of its agents, enables changing the function a system executes. Cooperation is the engine of this self-organisation, because it locally guides the agent for taking decisions. 

Compared to other approaches for designing multi-agent systems [Weiss 99], [Wooldridge, 04], our originality is to use emergence for overcoming difficulties related to applications complexity. As designers, we define agents, the environment (if required) and their interaction means; the organisation emerge.

The global function performed by a MAS results from the organisation between its agents. Changing this organisation by the AMAS self-organisation mechanism is equivalent to changing ths function. We showed that the collective function is adequate when every agent acts in a cooperative way. Cooperation is the foundation of the internal reorganisation of an AMAS. Self-organisation based on cooperation means that the system and its environment try to mutually adjust for being in cooperative interaction and also means that all the agents within the system tend towards ccoperative interaction. The function performed at the collective level is emergent because agents do not have the knowledge of the global finality. The additional condition for speaking of emergence is that the means for obtaining the global function performed by the system is not coded into the knowledge of the agents.