**Y**esterday, I gave my talk at the Seminar of philosophy of mathematics at Université Paris Diderot, in this new district of Paris where I always get lost because construction work continuously modifies the topology of the place. (This year, I ended up biking the mythical Rue Watt which has been beautifully renovated.) I managed nonetheless to get there in time and talked about Bayesian model choice and of the difficulties with Murray Aitkin’s proposal. The talk was presumably much too mathematical and not philosophical enough, but it was followed by a discussion launched by the two following speakers, Jan Sprenger and Bengt Autzen. Due to teaching duties, I could only attend the talk by Jan Sprenger, who covered the philosophical aspects of the difficulty in defining objective Bayesian inference, alas missing both Bengt’s and Steve Fienberg’s talks… He mostly focussed on MaxEnt priors, with an interesting counterexample by Teddy Seidenfeld, but also mentioned reference priors as suffering from the same difficulties. From my (non-philosophical) perspective, I consider that MaxEnt priors are beyond in terms of objectivity, because they first require the definition of a reference measure for the (entropy) divergence to be defined. During the talk, Jan also mentioned the book *In Defence of Objective Bayesianism* by Jon Williamson, which I will try to read (and comment) in the coming months. I just had a few words with who told me he had worked on Seber’s *Evidence and Evolution* as part of his PhD thesis, so I wished we had had more time to chat about that! (Steve has proposed to give his talk at the students’ seminar here in CREST so that we can discuss effects of causes versus causes of effects.)

### Like this:

Like Loading...