6 juin 2017, 11h00–12h30
Toulouse
Salle MS 001
Economic Theory Seminar
Résumé
When the state space is binary or when only posterior means matter, it is possible to represent any experiment (joint distribution of truth and data) as a convex function. I present three applications of this representation: (i) manipulating posterior means, (ii) determining joint informational content of multiple datasets, and (iii) rationalizing reactions to unknowns sources of information.