Preventing the HNQ feedback-loop

Question

We all know that the HNQ causes good- or okay-questions to be flooded with activity, especially if they have an abnormal or controversial title. There’s no need to get into the mess which kicked off this latest round of HNQ concern here, so I’ll get straight to my feature request:

Create a new ‘HNQ protection’ type between ‘protected’ and ‘locked’, and apply this to all HNQ questions automatically for the duration of their time on the list.

Protection currently prevents a user without 10 reputation on the site from answering a question. This ignores the association bonus, which restricts people who don’t know the specific site from contributing.

The details

This new form of protection would prevent all interactions from users with less than 10 site reputation — answering, upvoting, commenting. Downvoting is already prevented due to the association bonus providing less than 125 reputation.

It was mentioned in comments that an answer from an outside contributor can often be good. As we already have protection, it could be that ‘HNQ protection’ still allows answers. A high reputation user / moderator could then chose to apply protection in addition to ‘HNQ protection’, if desired.

The downsides

A small number of new users who find this question naturally will be prevented from performing actions on the question and answers. I am confident this would do very little to harm the site; most users who aren’t coming from the HNQ list will have 10 reputation, and those who don’t can easily gain it. We already have a similar rationale for protected questions.

The benefits

It will do a great deal to help the question, by preventing people who don’t know the site culture from jumping in and blindly upvoting low effort, or even flat out wrong answers, from posting off topic comments which have to be cleared up, and from answering a question with quick, irrelevant answers.

This would break the feedback cycle of ‘popular question goes on list’ → ‘people upvote question and boost popularity’ → ‘good question stays on list’, and significantly reduce the workload of moderators, desperately trying to keep up with thousands of irrelevant comments across hundreds of answers.

It would also help to prevent some other negative effects — if an inexperienced user posts one thing which goes ‘viral’, they can end up with way more reputation than their post — and their experience of the site — deserves.

The privilege

This would be a system-only privilege. Only questions on the HNQ could receive this, and all would receive it automatically. However, a moderator would have the ability to remove this form of protection if it was deemed unhelpful, and, once removed, could reenable it at any time, until the question was no longer on the list. This would be a moderator only privilege, to be used only in those situations where votes are being artificially inflated by external visitors from the HNQ, Reddit and Hacker News, and potentially other social media, or when comments are getting out of hand.

Answer 1

There's clearly a real need, and proven high demand for an action something like this going back years, with two very similar proposals in 2014 and 2016 (not dupes) for solving the same problem a similar way getting positive responses, with discussions around exact details.

It's time to see some action.

At this point, based on these four five years of discussion of proposals for how to solve this particular problem, we can conclude that this should be taken forward, the SE devs and PMs should look at the three proposed variants and their discussed pros and cons, and should give feedback on which can be implemented most cleanly with fewest side effects, on what sort of time scale.

The proposal described above, or either of the two slight variants below, would be a big, immediate improvement to HNQs that would have few side effects and can be done relatively quickly.

It can be done while we discuss and decide on any other, deeper, more fundamental changes needed to HNQs. Let's not allow the perfect to be the enemy of the good (or, the fixed be the enemy of the less-broken).

Here are those two earlier variants on this proposal:

• From 2016 (currently +59, -5) - Breaking the HNQ feedback loop on bad questions. Similar to this but it suggests allowing votes from all users with sufficient rep, but not factoring them in to the HNQ algorithm unless the user has 100 rep from the site in question. Fewer side effects (and no user-noticeable side effects), but doesn't also solve the problem of many curious non-experts voting up bad answers.
• From 2014, (currently +141, -27) Prevent questions on Hot List from being upvoted by casual visitors (only rep is from association bonus) Very similar, with slightly different details around the amount of rep needed, the way it is applied, whether it applies to comments or not, and whether it is applied automatically or manually.

---

This has been proposed and discussed in 2014, 2016, and now 2018, each after many incidents where a question getting stuck on HNQs caused problems. Let's not spend another 2 years debating the exact details and speculating, let's get implement something that we can test for real and iterate on.

---

It's also worth considering some of the ideas under the 2013 (currently +94, -44) proposal, The association bonus should not enable users to vote on every site, which is more controversial and would have more side effects, but would also solve this problem and would be even easier to implement. There are many interesting suggestions in the answers, like delaying the vote privilege a couple of days after the association bonus is given, or requiring rep from the site you're voting on to vote on protected questions or questions seeing high traffic.

There's more than enough good ideas and discussion to get a PM involved and do some testing.

Answer 2

As we already have protection, it could be that ‘HNQ protection’ still allows answers.

Here's one possibility to make this work without requiring protection:

• If the question is on HNQ and the user does not have 10 rep earned on that site, then the system will allow them to post answers, but those answers will not become publicly visible until they are reviewed.

• The user can see their answer, but it has a banner on it that says something the lines of

  Thanks for your answer! This answer will be visible only to you until it is peer reviewed.

  much like suggested edits from users below 2k rep.

• The answer gets sent to the First Posts review queue, where the review banner text clearly indicates that the answer was posted on a Hot Network Question.

• If the answer gets five Looks OK votes, then it becomes publicly visible. If it doesn't, then it gets deleted.

This runs on small modifications to a bunch of existing codebase, UX and community moderation elements, and it allows site communities to kill bad HNQ answers dead in their tracks while still letting through any gems that get posted by outsiders.

Answer 3

There was a recent change that would make the original proposal possible: One can now see in a question's revision history whether and when it was chosen for hot network questions list.

In the linked post, Catija states:

This was made possible by the work Adam did recently to move where we house the HNQ list in our code. This will only indicate when a question appears - there's not a similar event when the question drops off the list, so don't use it to judge whether something is currently on the list, you'll need to visit the actual list for that.

According to this, there is now the information that allows the system to know whether a question was ever on hot network questions.

I would propose that the kind of protection suggested in the question is enabled together with the addition of this revision history entry.

The protection should expire after some 30 days or so because it is quite unlikely that a question is having a lot of attention for that much time.

Moderators (and maybe high-ranking community members) should also be able to add this protection to any question if this question is for example linked in Reddit.

If the protection is applied to a question, a banner similar to this one should be shown as post notice:

marked as hot by Community♦ 2 mins ago

This question is featured on the hot network question list or in another channel. To ensure voting reflects the subject expertise of this community, this question has been protected.

You need to have earned at least 10 reputation on this site to comment or vote on this post.

Answer 4

As the current HNQ algorithm needs work – this seems to be the opportunity to add another aspect that should be included when and if the system is about to change:

One aspect of the feedback loop is that quick question and answers go hot –– even if they are "bad" in a few ways. That would mean that way too many low quality questions make the list, being not representative of the site they originate from.

On one site I was told when discussing the unwanted distortions HNQ effects bring along was:

The proper way to prevent this is to put the question on hold. This is something the community has tended to be reluctant to do, even when the same users leave comments requesting changes or clarifications. Hence, the present issue.

Therefore, the community needs to be proactive about putting subpar questions on hold. Fix the question, nominate for reopening, and then answer it.

But that means that initially bad questions with a good core that just need editing into shape will never go hot. Because they are by then too old to get a neat hotness score.

The same goes for any other old question. It may be a brilliant question, without answers. Then some time later, sometimes years later, this brilliant question also gets a brilliant answer. And both get only a couple of votes in the process because the network wide attention and voting is focused on the substandard hotsauce making the round.

If the algorithm is changed it should be changed to promote quality content. That content may have an age that's bigger than the arbitrary number of hours as it is calculated now. But the biggest problem is that the quality of the content is currently not meaningfully correlated with speed of votes, answer and number of answers. I do not have a readymade better algo for that now, but I feel that only tweaking the current one is beating a fundamentally wrong horse for the job into more speed done equally wrongly.