Toy model #6: Rationality and partial preferences — LessWrong