Comment by lolinder
1 year ago
What these are is evidence of your parent comment's point that this isn't direct moderator action, rather a combination of algorithms and user flags.
Most likely, people flagged the Germany story because it has a sensational title and they likely aren't from Germany and so wouldn't have context to know whether it's overblown.
I'm confident that Vx.dev got flagged by a bunch of people because they're tired of LLM stories (as repeatedly attested in this thread).
Based on the ratio of comments to upvotes, I suspect the Open Source Builds and Ford discussions ran afoul of the overheated discussion detector. Usually when the ratio gets too lopsided the software automatically drops the post off the front page, because that's an indicator that a lot of people are arguing in the thread without actually reading or enjoying the article.
I think generally it works well- when there are actual major events like early COVID or Ukraine - HN managed to inform we way ahead of mass media with various interesting sources. But I’m happy to have a “news” thing pop up only a few times a year. You’re gonna have someone be mad about every instance when you moderate
>LLM stories
Does that mean stories about LLMs or by LLMs?
Serious question.
I am one of the (few? many?) people (devs) who haven't look into LLMs or even tried out ChatGPT yet :), except to make jokes about it here once in a while.
There ought to be a time-based flagging limit, so that people don't abuse the system. I've already raised this earlier.
If Company A makes a killer product announcement, rival Company B could simply get its employees to spam down votes on and flag that post. Company A gets less visibility, and dang won't be able to come on time to stop it.
This is an easily plausible hypothetical, which may already be happening.
Flagging requires high HN karma. You get that by being a positive member of the community. Most such people, if a company even has one, would find it against their personal ethics to do that. And dang can see the karma ratio and unflag any actually worthwhile announcements.
I think as people have become more and more aware that flag negatively weights items for rankings, and isn't just a "hey have the mods look at this rule breaking thing", more people have started using it as a downvote button. It was my understanding that HN originally didn't have a downvote feature to avoid the kind of issues that the flag usage is now causing.
1 reply →
I think you're probably generally correct, but "blaming the algorithm" sure smells to me like a whole lot of camouflage for censorship, which we ought to know by now has as much to do with 'quality' as it does 'shaping the narrative'
Generally speaking HN is a good site and a case study in successful community moderation, but you have to wonder 'who's watching the watchers' these days as the Overton window on free speech continues to be narrowed, almost entirely at the behest of big tech.
The simple solution would be to display a log of all removed/flagged/shadowbanned posts and comments, like Wikipedia does.
Preventing the site from being taken over by incessant meta debates is one of the moderation goals of the site.
https://hn.algolia.com/?dateRange=all&page=0&prefix=true&que...
(In many places there, obviously a lot of that is about Meta the company).
Periodic threads like this one are, I think, allowed as a sort of escape valve for pent up meta energy. Emph. on "periodic".
If you want a site that makes the opposite call here, Lobsters has a public mod log. You might like that system better!
Enable showdead in your profile if you want to see dead posts. You can't see deleted posts as the author deaded or asked HN to delete the post. See the HN FAQ [0].
[0] https://news.ycombinator.com/newsfaq.html
[flagged]
I don't care if its a right-wing talking point and I don't care if a company has given themselves permission to censor themselves.
My interest is broadly about the attitudes towards speech as they continue to be shaped by corporations who continue to privatize the public square. See Taibbi's 'Twitter Files' for ample examples.
2 replies →