A Stratification of Possibilistic Partial Explanations

Abstract

Several problems are connected, in the literature, to causality: prediction, explanation, action, planning and natural language processing… In a recent paper, Halpern and Pearl introduced an elegant definition of causal (partial) explanation in the structural-model approach, which is based on their notions of weak and actual cause [5]. Our purpose in this paper is to partially modify this definition, rather than to use a probability (quantitative modelisation) we suggest to affect a degree of possibility (a more qualitative modelisation) which is nearer to the human way of reasoning, by using the possibilistic logic. A stratification of all possible partial explanations will be given to the agent for a given request, the explanations in the first strate are more possible than those belonging to the other strates. We compute the complexity of this strafication.