This problem is similar to the bead jar guess problem. Essentially the problem is where priors come from and it doesn't have a general solution within the context of Bayesianism. Bayes can tell you how to update your priors, but not what your initial priors should be.