The Mystery of the Haunted Rationalist — LessWrong