This gave me an idea: how about you give the author in-editor feedback? As you type your message it runs the content through some sort of sentiment + content an

This gave me an idea: how about you give the author in-editor feedback? As you type your message it runs the content through some sort of sentiment + content analysis and tells you on a scale of 1-10 how civil or factual your post is. This way harmful content not only gets self-moderated, but with repeated, immediate feedback, users actually learn how to write better. I'm not saying that posts with a low score be removed, but perhaps they will receive less priority on the feed. Building that analysis layer is the difficult part.