Modeling the role of (ab)normality in the ascription of causality judgments by agents
Résumé
Abstract How agents ascribe causal links in face of a sequence of reported events, on the basis of their knowledge about how the world behaves, is a problem of interest, which should be distinguished from classical abduction in diagnosis tasks. The paper suggests an approach to this problem, assuming that agents knowledge is expressed in terms of non-monotonic consequence relations (which discriminate between what is normal and what is exceptional). This has several benefits. It enables to emphasize the role played by elements perceived as abnormal among the reported events when ascribing causal relationships, and distinguish between factors that facilitated the occurrence of events, from other events that are considered as being more instrumental. Besides, it also points out the bipolar nature of imprecisely known causal links where effects that are guaranteed to be possible are distinguished from effects that are just non-impossible