Interventions in Possibilistic Logic
Résumé
An intervention is a tool that enables us to distinguish between causality and simple correlation. The use of interventions has been only implemented in Bayesian net structures (or in their possibilistic counterpart) until now. The paper proposes an approach to the representation and the handling of intervention-like pieces of knowledge, in the setting of possibilistic logic. It is compatible with a modeling of the way agents perceive causal relations in reported sequences of events, on the basis of their own beliefs about how the world normally evolves. These beliefs can also be represented in a possibilistic logic framework.