Fig. 6.

A probability space {p₁,p₂} (with p₃ = 1 − p₁ − p₂; the prior with maximum entropy is at {1/3,1/3}). The figure illustrates what happens when probability noise is added to a given “meta-prior,” or population mean from which priors are drawn. Noise can move the prior in any direction, but it usually moves it in a direction that increases entropy.