zlacker

>>Robin8+(OP)
This is a good thing to be able to see, but I'm much more interested in identifying the soft censorship in comment sections.

This is obviously harder, because vote totals aren't publicly available for comment sections, but it is much more important as a tool. What topics are on the front page is much more clearly the legitimate domain of moderation than what commentary is made about them, especially when moderation of those comments contradicts the vote mechanism.

>>noqc+Pn
It's more boring than you'd expect. The comments that get most heavily upvoted tend to be either (1) indignantly rhetorical, or (2) generic. Mostly what we do is downweight those when they're at the top of the thread. The "algorithm" is about as simple as:

(1) if the top comment is indignantly rhetorical or generic, downweight it; otherwise go look at another thread;

(2) refresh the page;

(3) goto (1).

If I'm feeling diligent, I might do this for the top few subthreads, but that's about it.

This simple intervention turns out to be the highest-leverage thing we've figured out in recent years:

ttps://hn.algolia.com/?dateRange=all&page=0&prefix=true&query=by%3Adang%20highest%20leverage&sort=byDate&type=comment