I have a theory that, individually, people fundamentally disagree less than they think they do. And in real face to face meetings, they are willing to disagree about a topic without killing each other. They talk around points of disagreement. Eg: me, when I'm talking to my grandmother.

The problem is that online people tend to polarize into factions during complex discussions. The more heated the discussion, the more polarized they become. Eventually it becomes impossible for either side to be self critical, or to cede a point to the other side no matter how true it may be.

"B" says "the sea is made of saltwater". A moderately reasonable "A" agrees. Mistake! The "A" group accuse them of being pro-B or anti-A. Suddenly, the idea that the sea is saltwater becomes an "B dogwhistle". A cunning trick! At this point, the "A" group adds 'saltwater sea' to their list of unacceptable opinions, and the policing of everyday language and opinion is in place. That's when the notion of defending free speech gets questioned. How can you police language if people defend free speech?! It won't do! So now free speech is a dogwhistle too. Add it to the list! There is no possibility for constructive dialogue, no matter how sensible, kind and cool-headed it tries to be. Disagreement with dogma or talking to the opposition are offences. New rules: "Don't follow an A on Twitter". Guilt by association. Next up: let's use the abuse/reporting system to ban individuals. Eventually, online systems that were supposedly designed for communication and conversation have become tools for suppression and virtue monitoring.

While anonymity itself will help people to own up to an opinion that they might be otherwise afraid to voice, unless there are no usernames at all it will be possible to track an individual across the system and figure out "which side they're on". Then you'll be able to figure out if their comments are acceptable or hatespeech without having to properly consider them.

Don't get me wrong, I'd love to see this happen, but I think it requires a lot of work. Deduplication of points made, automatic detection of logical fallacies via natural language processing, personal dictionaries with auto-translation of terms (eg: an acceptable word to you might be a slur to someone else - if the word has a genuine meaning, auto-translate to a non-slur). That would avoid having people point and scream and claim moral victory because of a word infraction, instead of hashing thing out properly.