Having established some philosophical properties of Bayesian statistics, it is time to apply some Bayesian insights to current directions in the physical sciences. The first such insight relates to the greedy top-down method of experiment selection (section 5.2.)
This provides a Bayesian model of Poincaré's argument, articulated by Pirsig [55], that scientists choose to measure processes very far away in space or in time, because these provide the highest probability of a result other than the expected one; in other words, because they have a high value of the Shannon entropy, or of some similar function, over the marginal likelihoods.
This incentive for scientists to measure processes on very large length scales also applies to very small length scales, a process that Pirsig [55] identifies in biology. This provides two directions in which science has [10] moved, in the attempt to draw more robust inferences about the truth of theories: to larger length scales, and to smaller length scales. The latter is [58,25], according to Heisenberg's uncertainty principle, equivalent to moving to higher energies.