Alignment with argument-networks and assessment-predictions — LessWrong