That's definitely the hard part of switching to Bayesian methods. But the problem doesn't go away if you use p values, it just gets ignored. Typically people d

That's definitely the hard part of switching to Bayesian methods. But the problem doesn't go away if you use p values, it just gets ignored. Typically people doing research have prior information about what they are researching. E.g. previous studies have found effect sizes to be in some interval. You can fallback on an uninformative prior or other tricks if you really want to model the idea that you know literally nothing about what you're researching. But that should be very rare.