Hacker News Reader

Comment by xhrpost | Hacker News Reader

Except these posts seem to regularly get taken down or flagged. This isn't even on the front page anymore it seems.

It set off the flamewar detector. We monitor that list and eventually cancel the penalty for threads that aren't actual flamewars, but it takes a while. If anyone wants to make it faster, emailing hn@ycombinator.com is usually a forcing function.

Edit: Also, posts without URLs get a downweight by default. I've turned that off too, and merged the other threads into this one, since it was the first one.

How does one programmatically detect a flamewar? :)

By proxy, which is why there are false positives.

> How does one programmatically detect a flamewar? :)

An NLP/Machine learning model would be a great way! Probably also taking in other features as input such as the number of comments per second, and how deeply nested those comments are.

Let's build one!

> It set off the flamewar detector.

Why? There do not appear to have been any posts remotely resembling a flamewar in the comments.

That's answered in my sibling (well, cousin) comment here: https://news.ycombinator.com/item?id=25921757. It can't test for flamewars directly so it ends up guessing wrong sometimes.

I'm glad I'm not the only one noticing this. I could usually rely on HN to not be so heavily "filtered" for a lack of a better term but I have noticed the slide into a more controlled/gamed PR outlet.

Reply on news.ycombinator.com