Gossip as a model of inference to composite hypotheses

In this paper we seek an inferential and cognitive model explaining some characteristics of abduction to composite hypotheses. In the first section, we introduce the matter of composite hypotheses, stressing how it is coherent with the intuitive and philosophical contention that a single event can be caused not only by several causes acting together, but also by several kinds of causation. In the second section, we argue that gossip could serve as an interesting model to study the generation of composite hypotheses at a larger scale: several characteristics of gossip (for instance its being diluted over time and its collaborative dimension) make it extremely prone to produce composite hypotheses considering different levels of causation. In the third and final section, we try to illustrate some specificities of abduction to composite hypotheses for individual agents basing on the analysis of collective agents.