From Acceptation Relations to Causality Ascription in a Belief Function Framework

From a generic set of uncertain pieces of information about the normal course of things and a temporal sequence of reported facts, an intelligent artifact should be able to identify causally related events and distinguish between factors that facilitate or justify the occurrence of events from other facts. In this paper, we propose a model for causality and facilitation ascriptions when the background knowledge on n-ary variables is represented under the belief function framework. In order to ascribe causality, an agent has to judge if an event is accepted or rejected. We propose different definitions of acceptance and rejection to specify the strength of the causal link in the context of n-ary variables. Even accepted, an event can be confirmed or attenuated which leads to different forms of causality ascriptions. This paper shows that our model can treat these cases in the framework of belief functions.