Reflective Probability Distributions and Standard Models of Arithmetic — LessWrong