Séminaire de Logiques Non Classiques : Denis Bonnay
Denis Bonnay (joint work with M. Cozic)
Université Paris Ouest Nanterre La Défense
An axiomatization of individual and social updates
In this talk, I will consider update rules, which an agent may follow in order to update her subjective probabilities and take into account new information she receives. I will consider two different situations in which this may happen:
*** individual updates: when an agent learns the probability for a particular event to have a certain value.
*** social updates: when an agent learns the probability an other agent's gives to a particular event.
Jeffrey's conditioning and weighted averaging are two famous update rules, in individual and social situations respectively. I will show that both can be axiomatized by means of one and the same invariance principle, related both to Carnap's use of invariance in his work on probabilities and to the principle of minimum information.