HN is actually one of the most homogeneous public communities I've seen.
And judging by how few downmodded posts are around, self censoring.
But then the next day a different post on the same topic could have an entirely opposite hivemind, often no discernible reason.
I'm not so sure this is a huge effect. I often express "incorrect opinions", and get some amount of downvoting or ridicule, but not usually to a significant degree. When I've had my comments dogpiled, I can see that it happened because of how I stated my opinion more than because of the opinion itself.
Sometimes the same story will be flag killed one day, front page the next.
Look in a thread where China is mentioned and you'll see the real face of many HN'ers - if they aren't removed before you see it. Luckily the MOD'ing is pretty good.
Unfortunately since most people aren’t experts on most things, this turns into a bunch of fugazi. People who are financially or academically successful in business or engineering or marketing think they understand politics too.
This is what happens with topics like China. I don’t care about China much. So I end up defending it because the negativity towards it by the Global North is staggeringly inconsistent.
This doesn't seem obvious to me at all. It seems not only plausible that this could be the case, but unsurprising.
Almost nobody here reads every story in the feed. Any given story on HN is going to be read by a collection of people that differs from story to story. And most of those people won't make any comment whatsoever. It seems unsurprising to me that a story on, say, capital punishment could get a large number of people who are passionate about that topic, and a story on pedophilia could get a large number of people who are passionate about that topic, and the majority of commenters in both could hold opposite opinions without even one of them being hypocritical.
FTR I personally think that training commercial ML algorithms on bulk data produced by people that haven't released it to the public domain or licensed it for such use is ethically dubious (regardless of the legality) whether the data consists of code, art, metrics or anything else.