RE: Topcomment weekly curation report (week 21)

I'm really interested in that! I'll read your posts later!

what makes a great comment, and whether I can algorithmically "score" good comments.

We use some rules to decide what we consider a good comment. For instance it must show that the commenter has read the blog itself. They must reply to that subject (not reply just to tell their story or point of view) and they should have put effort into the comment.

I think this will be difficult to score this algorithmically. I use a script that collects all comments of 'a certain length' and sends them to our curation channel in our Discord server. From there we do a manual check whether there is real engagement. AI comments are also a thing recently. We've been spotting quite some 'long' comments that do engage with the author of the blog, but are AI generated.
That is why manual curation is important for rewarding comments.

I am also keen to see if there's a correlation between good content and good comments.

Yes, I think there is. Although some subjects are more likely to get discussions in the comment than others.
We do see that larger stakeholders often get more comments. This could be either because they regularly upvote comments or that people want to be seen/followed by these stakeholders.

Also the interaction with the commenters by the author could be interesting to look at.

my overall goal is try and drive better comment quality, not only in the use of good language (but also in terms of comment depth) - a comment is not very valuable if it doesn't generate discussion.

I can only agree with this. As you have seen I started following you since I find this an interesting matter. Keep me (or us) in the loop!
If you have Discord you can find us in our own Topcomment Discord server:
https://discord.gg/u7ebA2QKCd