On expected utility, part 4: Dutch books, Cox, and Complete Class — LessWrong