I've realized over the years that most negative comments either don't originate from good faith, or are interpreted as bad faith, and that a majority of moderation is probably just reacting to these comments. In addition, it's less likely we will perceive comments as good faith if we can't identify with the commenter. Seems like there is still a big opportunity to iterate on these properties, like a system that highlights similarities between commenters, or reframes positions and backgrounds to be more in line with good faith.