Probabilistic reasoning for description and experience — LessWrong