In standard models of rational learning from experience, prior uncertainties and disagreements recede smoothly as common evidence accumulates. However, this presumes that the underlying risks are relatively stable. Otherwise, rational learners need to sift random noise for signs that the trend...