ASCRIBING STRONG CAUSES FROM OBSERVATIONAL AND INTERVENTIONAL DATA UNDER THE BELIEF FUNCTION FRAMEWORK

摘要

Ascribing causality amounts to detect causally related events. In this paper, we consider that these events are uncertain and expressed with the belief function formalism. They are either results of observations or interventions which are external actions forcing the variable to take a specific value. We show how the proposed concepts of strong accept/reject can be used, instead of changes in uncertainty, to discriminate between potential causes.