What are we assuming about utility functions? — LessWrong