One of the aims of simulating organizations is to understand the relationship between the recurrent patterns of group behaviour, called organizations, and recurrent patterns of individual behaviour, called roles. These recurrent behavioural patterns can be formally described either by fixed protocols for interaction between roles at the organizational level and with the external environment at the individual level, or by constraints on the possible behaviours by means of collective and individual norms. Individual norms are represented in various ways: as mental attitudes, as obligations, authorizations and conventions or as commitments, etc. One can think of collective norms as specifying a set of possible structures of interaction in a group. This last option is generally linked to a Belief-Desire-Intention (BDI) agent architecture.

We consider only the descriptions of structures of interaction as a set of protocols between roles. Therefore, the roles played by the individuals are completely determined by these structures in a kind of methodological holism. We further have to distinguish whether the recurrent patterns of behaviour refer to the actual or expected behaviours. The expected behaviours would be a specification of the kind of interactions we would like to emerge from the dynamics of the multi-agent system but which does not necessarily happen. The actual behaviour would be the description of the detailed interactions which has necessarily to occur when the agents interact with each other.

We use a formalism of the structures of interaction as a set of protocols between roles to describe the dynamics of the groups instantiating these structures with individuals creating, entering and leaving groups. Our aim is to provide both a formalism and a tool to investigate the dynamics of social networks in a simple way. More precisely, the aim is

a) to propose to use a formalism of the structures of interaction as a set of protocols between roles
b) to describe the dynamics of the groups instantiating these structures with individuals creating, entering and leaving groups
in order to investigate the dynamics of social networks in a simple way.

Thus, we partially address the challenge of allowing for open systems in social simulation, which have since a while been recognized as a difficult class of system to design. We recall that, in software engineering, open systems are defined as encapsulated, reactive and spatially and temporally extensible systems, whose components are entities that a designer has no knowledge of at the design time and which may dynamically join and leave the system at run-time.