Dinner with Bayes: On the revision of risk beliefs