1. “If people are so inclined to see the world through their tunnel vision, why suppose they are able/willing to be explicit about their biases?”
2. “If priors are to represent biases, shouldn’t they be kept separate from the data rather than combined with them?”
3. Lastly, putting aside questions 1 and 2 for the moment, doesn’t bias contaminate the Bayesian updating process altogether? For example, if I am biased in favor of X hypothesis being true, won’t my bias cause me to neglect or discount any evidence against my favored hypothesis?
Hat tip to Deborah Mayo, who posed questions #1 and #2 in her recent talk on science and statistics (check out slides 31-33 for a helpful summary of Professor Mayo’s take on the new “data journalism” generally). For our part, my wife and I posed question #3 in an email to Mr Silver long ago (December 2012) but have yet to receive a reply.
So, what’s the cure for bias?