> How can we learn more if we don't change our mind?

I think it's also possible to learn without having strongly held beliefs about the subject beforehand. This alternative reminds me of Descartes (don't assimilate knowledge until you are sure of it's veracity) and Bayes (keep track of degree of belief about traditionally non-probabilistic things). Maybe such an approach would help getting trapped in local optima. E.g., I'd imagine it would be hard to climb out of the theist energy well once your world view were based on it.