On: One Man's Modus Ponens

[gwern]

It's hard to say how Bayesian reasoning really works with classical logic (which is true of all statistical approaches), but you can set up plausible Bayesian models where you do get 'backfire' or 'polarization' effects in which people update in different directions based on the same observation because of the different models/priors they hold: https://www.gwern.net/docs/statistics/bayes/2007-bullock.pdf http://papers.nips.cc/paper/3725-bayesian-belief-polarizatio...