[note: I am technically a mod but in practice that means I give the team my opinions and occasionally curate something. The following is my opinion. I think it has some overlap with team members’ opinion but I don’t know how much, or how much of that opinion has made it into policy]
A counterintuitive effect is that bad comments are often worse for SNR than bad posts. Bad posts seem like they’d be worse, because posts are more important than comments and are placed higher. But karma works better (on average) on posts: bad or mediocre posts get enough downvotes, or at least not enough upvotes to compete, and gently disappear. But comments’ views scale with the viewership of the original post, so a mediocre comment on a popular post will get lots of attention regardless of its karma. If a post gets enough comments that low karma comments can’t get much attention, they still compete with new high-quality comments, and cut into the attention for the latter.
And even if no one else sees a bad comment, they are still likely to be read by the author and annoy them. If this gets bad enough, authors may stop reading their comment sections or stop posting altogether.
If a post gets enough comments that low karma comments can’t get much attention, they still compete with new high-quality comments, and cut into the attention for the latter.
Seems like this could be addressed by changing the comment sorting algorithm to favor recent comments more?
I disagree. Negative comments often provide feedback to the author he wouldn’t get elsewhere. And if you are annoyed by it you can filter them out (settings → hide low votes).
No. There can be many means in between or different altogether.
But back to my original comment: It was about the not made explicit action of what to do with bad comments. I agree that the dynamic for posts and comments is different. But I disagree with what I saw was the push that negative comments should be stronger discouraged because they have higher weight.
But when rereading, I see that you don’t say what to do about these comments. You only point out negative effects. What is your proposal?
Note: I’m in favor of tending the garden and discouraging orcs and banning trolls. But I’m also in favor of critical and negative remarks. Reduce their visibility maybe, but don’t completely prevent them.
AFAIK there was a wave of rate limits, not bans. I think it’s a huge error to conflate those. Most importantly, you can complain on-site about being rate limited in a way you can’t complain about being banned.
I have complaints about implementation but the theory seems sound. I’d like the team to put more work into implementation or treat false positives as more costly, but that’s easy for me to say since I’m not the one that has to do it.
Complaints:
the combination of imperfect filtering and no communication seems bad to me. How are people supposed to know their ban was a mistake and asking will help, instead of annoying mods further.
“retroactive to a year ago” sounds pretty bad to me. But I don’t think that’s the right frame. I think the team meant to intervene and not rate limit people who’d had an issue 11 months ago but have been great since. habryka described at least one ban as a mistake in comments on this post, so sounds like this was inconsistent. But conceptually I think it was supposed to be “we have a new tool for detecting people who have been below standards this entire time” not “we raised the bar”.
[note: I am technically a mod but in practice that means I give the team my opinions and occasionally curate something. The following is my opinion. I think it has some overlap with team members’ opinion but I don’t know how much, or how much of that opinion has made it into policy]
A counterintuitive effect is that bad comments are often worse for SNR than bad posts. Bad posts seem like they’d be worse, because posts are more important than comments and are placed higher. But karma works better (on average) on posts: bad or mediocre posts get enough downvotes, or at least not enough upvotes to compete, and gently disappear. But comments’ views scale with the viewership of the original post, so a mediocre comment on a popular post will get lots of attention regardless of its karma. If a post gets enough comments that low karma comments can’t get much attention, they still compete with new high-quality comments, and cut into the attention for the latter.
And even if no one else sees a bad comment, they are still likely to be read by the author and annoy them. If this gets bad enough, authors may stop reading their comment sections or stop posting altogether.
Seems like this could be addressed by changing the comment sorting algorithm to favor recent comments more?
I disagree. Negative comments often provide feedback to the author he wouldn’t get elsewhere. And if you are annoyed by it you can filter them out (settings → hide low votes).
Shouldn’t one make a distinction between negative/critical and bad?
Sure, but literally bad ones will quickly get downvoted and the poster banned. This is about the less clearcut cases, right?
It sounds like you don’t think there should be any user-focused mod response between “nothing” and “banned”. Is that correct?
No. There can be many means in between or different altogether.
But back to my original comment: It was about the not made explicit action of what to do with bad comments. I agree that the dynamic for posts and comments is different. But I disagree with what I saw was the push that negative comments should be stronger discouraged because they have higher weight.
But when rereading, I see that you don’t say what to do about these comments. You only point out negative effects. What is your proposal?
Note: I’m in favor of tending the garden and discouraging orcs and banning trolls. But I’m also in favor of critical and negative remarks. Reduce their visibility maybe, but don’t completely prevent them.
Rate limiting. If I was pope I’d make a few tweaks, but I think the concept is fundamentally sound and the implementation good enough.
Then we agree about the general moderation of LW.
Did your comment also apply to the latest automated bans?
AFAIK there was a wave of rate limits, not bans. I think it’s a huge error to conflate those. Most importantly, you can complain on-site about being rate limited in a way you can’t complain about being banned.
I have complaints about implementation but the theory seems sound. I’d like the team to put more work into implementation or treat false positives as more costly, but that’s easy for me to say since I’m not the one that has to do it.
Complaints:
the combination of imperfect filtering and no communication seems bad to me. How are people supposed to know their ban was a mistake and asking will help, instead of annoying mods further.
“retroactive to a year ago” sounds pretty bad to me. But I don’t think that’s the right frame. I think the team meant to intervene and not rate limit people who’d had an issue 11 months ago but have been great since. habryka described at least one ban as a mistake in comments on this post, so sounds like this was inconsistent. But conceptually I think it was supposed to be “we have a new tool for detecting people who have been below standards this entire time” not “we raised the bar”.