The work presented in this paper is part of the project SocLab, which proposes a formalization of the sociology of organized action (Crozier et Friedberg). This formalization is based on a meta-model of the structure of social organizations, which provides means to describe the structure of a particular organization, to develop an analytical study of its properties and mainly, to calculate by simulation the behaviors that the actors of the organization are likely to adopt one to each other. Under this approach, an organization is viewed as a system that, depending on the behavior of the actors to each other, gives every of them a certain capacity of action to achieve its objectives, without distinguishing those related to his role within the organization and those that are its own. These behaviors are relatively stable. This is an essential condition for the coordination of the actors so that they can coordinate in performing, at least partially, what constitutes the raison d'être of the organization. These behaviors appear also to be generally cooperative facilitating the achievement of personal objectives of each one as well as those of the collective as a whole. This thesis focuses on the modeling of the rationality which leads a social actor to adopt such behavior in the " social game " constituted by a context of organizational interactions. According to the sociology of organized action, this rationality is strategic, guided by the research of own interest, and it is exercised within the framework of a (very) limited rationality. The proposed model seeks to be plausible, from the social and the psycho-cognitive points of view, and it fits into the paradigm of reinforcement learning. Insofar as the structure of the organization allows it, the simulations converge towards configurations that can be described as Pareto optima. We also study variants of this algorithm corresponding to rationalities that drive an organization to regulate toward other configurations that are elitist, protective or egalitarian, or Nash equilibria.
