This paper proposes initial steps towards a generic framework for modeling the scientific process. It is generic according to two main axes. First, it can be instantiated to cover various types of inferences usually considered relevant in science, second, and more important here, it allows for the modeling of the social dimension of scientific activity. After motivating this drive for genericity by looking at some results from philosophy of science, the paper presents the bases of the framework and its central reliance on the notion of consistency, both at the individual and group levels. It then instantiates the social dimension of the framework by proposing an actor-critic model of scientific interaction. The ideas proposed are illustrated with examples of hypothesis formation in medicine.