Possibilistic Explanation

The philosophy literature has been struggling with the problem of defining causality. There has been extensive discussion about it. Hume taught that talk of causation was metaphysics and should be consigned to the flames (D. Hume, 1748). We present a comparison between two approaches that model this notion: The first one is the normative approach which is based on an interventionist conception of the cause (D. Kayser, A. Mokhtari, 1998). The second one is the structuralmodel approach that gives a new definition of actual causes, using structural equations to model counterfactuals (J. Y.Halpern, J. Pearl, 2005). Both approaches propose a modelling of the explanation using the notion of causation. Halpern and Pearl define the notion of partial explanation, so that every explanation will not be considered equally good. In order to make difference between them, the authors add a probability to the set of possible contexts (J. Y. Halpern, J. Pearl, 2005). Our purpose in this paper is to modify this definition, rather than to use a probability (quantitative modelling). We suggest to affect a degree of possibility (a more qualitative modelling) which is nearer to the human way of reasoning (P. Prade, D. Dubois, 1994). We propose to give a stratification of all possible partial explanations for a given request of an agent. The explanations in the first stratum are more possible than those belonging to the other strata.