Expansive translations: considerations and possibilities — LessWrong