On the other hand, I do think companies like Facebook and Google and the news sites (or whoever makes their comment systems) can do a lot about "the impact of their algorithms and their ability to shape popular sentiment in our society," as he alludes to in the article but fails to explore in any depth.
What if there were simply richer tools for users to rate things? For instance, to tag a post as "+1 nuanced" or "-3 overly divisive" or "-2 unsupported by evidence" or "-3 inappropriately political" or "-5 bigoted", and then have algorithms (and user interfaces) that deal with this additional information in ways that actually are effective while also being careful not to discourage those who don't like getting downvoted? (e.g. only show downvotes to users a month after they appear so the user is less likely to emotionally respond, but still gets feedback as to why their microphone is getting the volume turned down)
Then of course give users tools to control what they see....e.g. hide (or suppress) divisive political content, etc.
There are any number of things that can be done to tone down the hateful divisive rhetoric that pervades online social spaces, and lets the insightful, nuanced content float to the top. Is anyone doing this? Are they even experimenting with it? Are they so scared that users will run away if there are too many options? (you know, you can always put them behind a "show all ratings options" setting that by default is off)
This isn't censorship, this is just putting into place things that have in place in the real world for millennia, but that disappear in naive approaches to bringing conversations online. It won't be perfect initially, but it can at least be a lot better.