Bayesian inference only works for "the cat is on the mat" examples. In reality, hypothesis and theory discovery and model selection are highly complex and neither a theory/hypothesis/model's accuracy nor the subjective probability that it's true are the decisive criterion for choosing the best one.