←back to thread

Coming Soon to Hacker News: Pending Comments

661 points pg | 9 comments | 21 Mar 14 23:49 UTC | HN request time: 0.234s | source | bottom

A surprisingly long time ago (2013 was a busy year) I mentioned a new plan to improve the quality of comments on Hacker News:

https://news.ycombinator.com/item?id=6009523

Since I'm going to check out of HN at the end of this YC cycle, this was my last chance to get this done. I didn't want the people who are going to inherit HN from me to have to build it as their first project, because it interacts with so many different bits of the code in such subtle ways.

So I found time to implement pending comments this past week, and with any luck it will launch tonight. Since it's a big change, I wanted to warn HN users in advance.

Here's how it currently works. From now on, when you post a comment, it won't initially be live. It will be in a new state called pending. Comments get from pending to live by being endorsed by multiple HN users with over 1000 karma. Those users will see pending comments, and will be able to endorse them by clicking on an "endorse" link next to the "flag" link.

Someone who has a pending comment will have to wait till it goes live to post another. We're hoping that good comments will get endorsed so quickly that there won't be a noticeable delay.

You can currently beat the system by posting an innocuous comment, waiting for it to be endorsed, and then after it's live, changing it to say something worse. We explicitly ask people not to do this. While we have no software for catching it, humans will notice, and we'll ban you.

Along with the change in software will come a change in policy. We're going to ask users with the ability to endorse comments only to endorse those that:

1. Say something substantial. E.g. not just a throwaway remark, or the kind of "Yes you did, No I didn't" bickering that races toward the right side of the page and no one cares about except the participants.

2. Say it without gratuitous nastiness. In particular, a comment in reply to another comment should be written in the spirit of colleagues cooperating in good faith to figure out the truth about something, not politicians trying to ridicule and misrepresent the other side.

People who regularly endorse comments that fail one or both of these tests will lose the ability to endorse comments. So if you're not sure whether you should endorse a comment, don't. There are a lot of people on HN. If a point is important, someone else will probably come along and make it without gratuitous nastiness.

I hope this will improve the quality of HN comments significantly, but we'll need your help to make it work, and your forbearance if, as usually happens, some things go wrong initially.

Show context
beloch ◴[22 Mar 14 01:15 UTC] No.7446449[source]▶
My may concern with this system: Sledgehammer meets tack.

The comments on HN aren't perfect, but they're far from bad when compared to other sites of this nature. There has been a downwards trend most probably due to the increasing popularity of HN. A response is warranted. However, this system has the potential to silence a lot of high quality comments on any threads that aren't on the front-page for an extended period of time. Thus, you get a feedback loop. Good posts require quality discussion to stay on top, but must stay on top to get quality discussion going with this added approval lag.

I think you should ease these changes in as conservatively and gradually as possible. For example, apply it only to the top page at first, and reduce the number of endorsements required for display to 1. You might also consider merely greying out comments that have not yet been endorsed, as currently happens to down-voted comments. Another option would be to apply the endorsement system only after threads have reached a certain age so as to jump-start discussions. Additionally, I would recommend that authors of a parent post should be able to see all child posts regardless of their karma. Below, Babuskov raised the point that the endorsement system will obstruct useful back-and-forth discussions between sub-kilokarma users in buried threads that often takes the place of a private messaging system on HN. This would fix that more effectively than merely reducing the endorsement requirement.

You should not entertain any illusions that you can flip the switch and watch this system work perfectly, and that you will therefore be able to avoid confusing people with many changes over a lengthy period of time. Tweaking will almost certainly be required.

replies(13): >>7446750 #>>7446782 #>>7446797 #>>7446806 #>>7446997 #>>7447135 #>>7447144 #>>7447370 #>>7447425 #>>7447580 #>>7449121 #>>7449627 #>>7450193 #
1. minimaxir ◴[22 Mar 14 01:58 UTC] No.7446750[source]▶
For note, HN isn't necessarily increasing in popularity.

Number of HN submissions over 3 years: http://i.imgur.com/r9Ayvb1.png

Number of HN comments over 3 years: http://i.imgur.com/4FwglA8.png

replies(4): >>7446803 #>>7446963 #>>7447000 #>>7448566 #
2. yblu ◴[22 Mar 14 02:06 UTC] No.7446803[source]▶
But this is just one metric, how about the number of users? It certainly counts toward "popularity", doesn't it?
3. jonnathanson ◴[22 Mar 14 02:31 UTC] No.7446963[source]▶
It's possible for HN to be increasing in popularity while decreasing in # of submissions. For instance, if more users are coming to the site, the site is growing in popularity. If the X percent of users who normally submit the most frequently are decreasing their submission rates, submissions will go down despite the user growth.

Of course, this scenario assumes that # of submissions follows a power law distribution: a small percentage of posters make the majority of submissions. That pattern emerges on many content sites, especially sites with active forms of communal self-moderation. I have no idea if it's the case here, but it seems feasible.

replies(1): >>7446984 #
4. minimaxir ◴[22 Mar 14 02:36 UTC] No.7446984[source]▶
Whoops, it should be noted that the spam algorithm changed at around the peak, which also explains the decrease. (I made the chart)
replies(1): >>7447139 #
5. gruseom ◴[22 Mar 14 02:38 UTC] No.7447000[source]▶
Does that graph include comments under "submissions", or only stories? I bet the latter. There are only so many conceivably-appropriate-for-HN stories out there, and many of the rest get taken out as spam.

If the number of comments, on the other hand, hadn't gone up since 2012, I'd be shocked.

replies(1): >>7447048 #
6. minimaxir ◴[22 Mar 14 02:46 UTC] No.7447048[source]▶
Submissions only.

Here is a chart of the comments: http://i.imgur.com/4FwglA8.png

replies(1): >>7447434 #
7. mbesto ◴[22 Mar 14 03:02 UTC] No.7447139{3}[source]▶
This data isn't relevant to the discussion.

IMO, the quality has degraded due to more people viewing the site, not the submissions. If I get my post on the frontpage of HN I can expect roughly 1,000 hits per hour. This wasn't the case 4 years ago when I first starting lurking here.

There are two problems with this:

1. More people (of poor quality) upvoting stories that attract the most attention (link bait, stories that aren't relevant, etc)

2. More people (of poor quality) upvoting comments that are not relevant to the discussion.

No graph will be able to chart this - it's purely subjective.

8. gruseom ◴[22 Mar 14 03:56 UTC] No.7447434{3}[source]▶
Thanks for making that! I'm not shocked. :)
9. gahahaha ◴[22 Mar 14 10:51 UTC] No.7448566[source]▶
According to Google Trends Hacker News' interest seems to be be stable: http://www.google.com/trends/explore?hl=en-US&q=hacker+news&...