Epistemic communities: description and hierarchic categorization

(v2: some typos corrected in sec. 3.2)Social scientists have shown an increasing interest in understanding the structure of knowledge communities, and particularly the organization of "epistemic communities", that is groups of agents sharing common knowledge concerns. However, most existing approaches are based only on either social relationships or semantic similarity, while there has been roughly no attempt to link social and semantic aspects. In this paper, we introduce a formal framework addressing this issue and propose a method based on Galois lattices (or concept lattices) for categorizing epistemic communities in an automated and hierarchically structured fashion. Suggesting that our process allows us to rebuild a whole community structure and taxonomy, and notably fields and subfields gathering a certain proportion of agents, we eventually apply it to empirical data to exhibit these alleged structural properties, and successfully compare our results with categories spontaneously given by domain experts