• vendredi 21 juin 2002 : MAP, UMR CNRS-MCC N° 694, École d'Architecture de Marseille-Luminy, 2° étage, salle de réunion
  Annonce commune MAP-LSIS
  Pour toutes informations complémentaires prendre contact avec :
  Mme Camilla SCHWIND au 04 91 82 71 90 ou Camilla.Schwind@gamsau.map.archi.fr
  
  
  • 14h30 : Belief Revision And Conditionnal Logic - Nicola OLIVETTI
    Professeur à l'Université de Turin
    Professeur associé à l'Université de la Méditerranée
    Actuellement accueilli au MAP et au LSIS
    
    Résumé :
    Belief revision have been widely investigated in the last twenty years since the seminal work by Alchourron, Gärdenfors and Makinson (AGM theory). On the other hand, conditional logics were introduced in the Sisties by Stalnaker and Lewis to provide a formal theory of hypothetical reasoning. Establishing a correspondence between the two areas is interesting in both directions. For belief revision, as the correspondence may provide an object-level formalization. For conditional reasoning, as the correspondence can be used do express an intutive (and operational) semantics of conditional sentences dating back to F.P.Ramsey.The relation between the two concepts was formalized by Gärdenfors by means of the so-called Ramesy Test. However, Gärdenfors proved a famous negative result, known as "Gärdenfors Triviality result", according to which, no conditional logic can represent any signicant belief revision system via the Ramsey Test.Since Gärdenfors's negative result the problem of the Ramsey Test (whence, more generally, of the relation between belief revision and conditional logic) has been studied for more than fifteen years, stimulating a wide literature. In this seminar we show how we can reformulate the AGM postulates for belief revision systems that contain conditional formulas, by weakening the postulates in a natural way. Our reformulation allows us to establish a mapping between belief revision systems and conditionals by means of the Ramsey Test, without incurring in Gärdenfors' negative result. Moreover, we show that we can derive a system of conditional logic from the revision postulates by means of the RamseyTest itself. This logic provides an object-level formalization of belief revision in the language of conditional logic, as belief revision systems and models of this logic turns out to be isomorphic.