385

Some of you may have noticed that we've been making some changes to the Hot Network Questions on the back end over the last week or so. I'm here announcing our first round of changes to how the HNQ works and give you some ideas of why we're starting here and where we're planning to go in the future.

Before I get to that, I want to thank all of you who have participated in this discussion either on Tim's request for input back in October or over the last few years here on MSE. I spent a lot of time working through the suggestions and they really did a lot to help me prioritize what we should work on changing. Also, a huge thanks to Adam Lear who did the actual work making this happen.

Here are the big changes:

There's now a history event created the first time a post appears in the HNQ list.

This allows us to see which questions have been featured and look at data of how being featured impacts the questions and their answers so that we can continue to improve on the HNQ in the future. You can find this information in either the question's timeline (/posts/[postid]/timeline) or the edit history (/posts/[postid]/revisions).

Right now we don't have an event for a post dropping off the list. The concern with this is that questions right on the edge of the list may pop on and off the list several times, as frequently as every fifteen minutes, thus cluttering up the history. We'll look into whether we can find a neat way of achieving this but for now there's another feature that will help with this:

Questions will age out of the list after being on it for 72 hours.

There's a sense of fatigue with some of the questions that live on the HNQ list for long periods of time. To combat that, we'll remove questions that have had their day in the sun (or three days, anyway) to make way for new questions. Three days after a question is added to the HNQ list, it becomes ineligible and will be removed. This won't create a history event for now but it's something that we'd like to add in the future.

Each site can only have a max of five questions on the HNQ list at any given time.

This is a big, much-requested change and we may reduce the number in the future even further. We're starting it higher than some might want (suggestions went as low as one per site) because sites that have had a lot of exposure through the HNQ may see a dramatic drop in visits, so we need to be careful to find the right number here and possibly do some testing at different levels.

This number is also configurable on a per-site basis, so if a site wants to reduce their HNQ footprint, we can lower it even more, even to zero if a site wishes to be excluded entirely. Sites will need to go through a meta discussion before requesting this change and it will be up to the site itself to request a change rather than having the limit imposed upon it (unless we lower the maximum for the entire network). So for example, Stack Overflow can't vote to kick Movies & TV off the list entirely because they don't want to see spoilers for the last season of Game of Thrones, but Movies & TV can ask that fewer of their questions be shown so that they can devote sufficient time to those that are.

User preference to remove HNQ sidebar section globally

Your user profile settings tab got a bit more crowded today. You now have the option to turn off the HNQ List widget in the right hand sidebar. If you don't want to see the HNQ list, you don't have to! So, if this is what you've been dreaming of for years, head over to the sidebar section of your preferences and change it! The HNQ list is viewable by default for all users.

Hide Hot Network Question preference on user preferences page

There are a few reasons we're starting with this solution:

  • It'll work on every site and every device you use.
  • It's something we could implement relatively quickly and get a reliable result.

There are a few things it doesn't do:

  1. Doesn't let you show or hide the HNQ list on different sites or devices.
  2. Isn't an option for non-logged-in users.
  3. Doesn't let you customize which sites you see by either a whitelist or blacklist.

While I'd like to solve 1 and 2 by making it possible to collapse the section in the sidebar on the page and saving that state as a cookie, that will take a bit more design work so we're putting it on hold for now. Solving #3 is likely very complicated to implement so we don't have plans for it for the time being.

Title blacklists will be configurable on a per-site basis.

Right now we rely on regex to blacklist words from all sites. This is troublesome when words have different meanings depending on context. So, we'll be able to adjust blacklists both across the network and on each site individually.

Moderators have the ability to remove questions from the HNQ List.

There are times when the hotness formula selects a question that a site would rather not have featured. Up until now, the only recourse that was available was to close the question (which may be appropriate anyway but isn't ideal when done purely to manage traffic), or to do nothing. We're putting the power in the hands of our moderators to remove questions that don't set a good example for their sites. I recommend each site have a meta discussion with guidance for moderators about when - if ever - a question should be removed.

Once a moderator excludes a question, it can't be selected again, so don't think of this as a temporary "hide this question" option. In general, we recommend that you exclude questions that attract negative attention to your sites, that is, questions that are controversial, start large amounts of debate or arguments or even edit wars. Removing a question should not be a substitute for fixing it! Remember that it may take several hours for a moderator to respond to a flag so do what you can, first:

  • If the title seems click-baity or doesn't adequately describe the question, edit it!
  • If the body is full of spelling or grammatical errors, fix them!
  • If the body contains unnecessary detail or salacious content, see if it can be removed without impacting the question.
  • If the question is unclear or broad, vote to close it. In most cases it will be better to close a question and wait for it to be improved rather than asking for it to be removed.

This tool is a big gun and should be used sparingly. Don't reach for it if you think the question can be fixed.

When moderators kick a post off the HNQ list there will be a delay of up to fifteen minutes or so as the list is cached but the question will be removed the next time it runs. In addition, an event will be logged in the post timeline and edit history that indicates when it was removed and by whom. This will help us understand what sort of posts are unwelcome in the HNQ list on different sites.

At this point in time, mods can only use this tool on a post currently in the HNQ list - they can't use it preemptively to prevent a question from being added to it.


What we aren't doing, at least, not yet

There's a lot that we're not getting to in this round of changes or that we're hoping will become less of a problem with some smaller adjustments than some of the big asks we've gotten in the last few years. We know that there's still work to do but these changes will address the most immediate needs and give us the information we need to make changes in the future that are based on data rather than anecdotal evidence.

The data will help us see how effective changes to the formula are as we make them. You'll notice that we're not changing the formula itself for the time being and that's intentional. There have been some really great and interesting ideas floating around for a while, and we may try them out and see how they impact which questions are selected and how long those questions stick around.

We investigated having a review queue for this and, for now, this won't be possible. There are technical problems but also a concern that only sites with active reviewers will ever have posts approved to appear on the list. We'd like to see how much we can improve things by following the path we're currently investigating first. Some other adjustments, like setting a minimum question age so that users can have time to edit or close a question before it is eligible to appear on the list or preventing locked questions from appearing are other options we're considering for the near future.


That's the bulk of it. There's absolutely ideas we're not implementing I haven't mentioned due to wanting to keep this post reasonably short. If you have any questions about something - whether it's a feature I've mentioned here or not - feel free to ask about it in an answer.

Yes, I've avoided mentioning IPS in this question. If you're interested in that site in relation to the HNQ, please see their meta. The TL;DR there is: if they want back in, they can get back in but it's up to them.

46
  • 43
    "Stack Overflow can't vote to kick Movies & TV off the list entirely because they don't want to see spoilers for the last season of Game of Thrones" Spoken like a M&TV regular ;-) Mar 11, 2019 at 20:26
  • 51
    There's now a history event created the first time a post appears in the HNQ list. - I smell a hat trigger, maybe even a set of badges. Mar 11, 2019 at 21:02
  • 4
    @MathieuGuindon make that an answer as a FR!!
    – scohe001
    Mar 11, 2019 at 21:15
  • 3
    "we'll look at some stuff and see if it makes sense or would have prevented questions that were good fits for the HNQ list based on the site history" - what data are you thinking about here? The past five years' worth of HNQ hits? That type of historical analysis will take a significant amount of time before it becomes reliable (and, given the substantial amount of unrecorded history in the formative periods of existing sites, may never really get there). Or just at well-received posts, regardless of how they got that way?
    – E.P.
    Mar 11, 2019 at 21:26
  • 59
    Moderator removal + every individual user can now hide the list probably solves 95% of the problems. This is great!
    – jscs
    Mar 11, 2019 at 21:40
  • 7
    Will the "Achieved HNQ" have an attribute that's exposed via the Data Explorer? If not, I'll need to be making a request for that.
    – user287266
    Mar 11, 2019 at 22:32
  • 5
    @WebHead Yep. Should be there Sunday.
    – Catija
    Mar 11, 2019 at 22:37
  • 9
    For the longest time I have avoided joining Meta SE, but today I join to update this post. Thank you for these changes, they are extremely welcome
    – Tas
    Mar 11, 2019 at 23:21
  • 9
    As someone perpetually infuriated by HNQ, this seems... pretty good. Thanks.
    – hobbs
    Mar 11, 2019 at 23:48
  • 33
    Finally an action by SE that I can support... Good changes here.
    – user245368
    Mar 12, 2019 at 0:43
  • 3
    Thank you for this; it's great. Only killer feature missing is a per-user blacklist/whitelist option, but hiding it globally is an OK mitigation step while that is missing.
    – TylerH
    Mar 12, 2019 at 0:56
  • 11
    Right now we don't have an event for a post dropping off the list. ... neat way of achieving this ... Questions will age out of the list after being on it for 72 hours. --- So wait 72 hours, figure out when it was last on the HNQ list, and make a belated history edit (jammed into the list at the correct location) saying: "This question was last seen on the Hot Network Questions list" - any re-entering and exits restart the 72 hour delay, then the entry is made. IF it rejoins the list so be it (truly deserving), absolute worst case is there will be a few entries each week.
    – Rob
    Mar 12, 2019 at 3:17
  • 10
    did you consider status-completed for this old feature request: Allow mods or gold tag badge holders to prevent question from being on hot network questions list?
    – gnat
    Mar 12, 2019 at 6:58
  • 3
    Thanks! 1) So the HNQ sidebar isn't going away in the near future? 2) Can there be more cases of sites removed from the HNQ due to bad content reported by people? (i.e. is there something to prevent it from happening) Mar 12, 2019 at 7:41
  • 7
    "There's now a history event created the first time a post appears in the HNQ list." Only the first time? judaism.stackexchange.com/posts/100595/revisions shows three such events.
    – msh210
    Mar 12, 2019 at 8:48

24 Answers 24

125

Is it possible to have a moderator inbox notification of questions making it to HNQ? If some mods feel that would be too much, allow them to opt out or expect them to be able to ignore it. If this is too noisy, maybe collect all HNQ events from the past 24 hours into a single inbox item or something. Perhaps we could just have a single "New items on HNQ" notification which contains a link to a list and is bumped whenever needed, never leaving more than one item in the inbox. Or maybe the HNQ entry event could trigger a flag. There are various implementation alternatives (like a chat feed as mentioned in the comments), but I would like some kind of a notification for the site mods.

As a mod, I'd be happy to be notified so I can check that the question in question is well moderated and generally presentable. The most heated discussions I remember seeing on the site I moderate have been initiated by newcomers, many of them brought in through HNQ. I can manage that amount of heat, but I would appreciate being forewarned. The questions that need most cleaning up are among the ones that have been in HNQ.

I imagine this would be easy to implement and would have an effect somewhat similar to an HNQ review queue. If this is too hard to implement, a mod-accessible (or why not generally accessible?) list of current and recent HNQ questions would be nice, although I would prefer active notification myself.

22
  • 10
    I think this would be too noisy -- my site (scifi) generates enough HNQs that my mod inbox would probably be dominated by these notifications.
    – Null
    Mar 11, 2019 at 21:57
  • 6
    @Null That's why I mentioned the chance to opt out. On Latin (where I moderate) this would not be an issue, but I know other sites are different. I edited just now to add the option to combine various HNQ events into one mod inbox item. Mar 11, 2019 at 21:59
  • 3
    A daily compilation of HNQs would probably be more useful. The question would then be: when would the notification go out? At a certain time all around the world? It would probably be more useful if individual moderators could configure that time (e.g. for the time an individual moderator normally first logs into the network for the day). For that matter, though, just an easily-accessible list of the site's HNQs from the past 24 (or 72?) hours would work just as well.
    – Null
    Mar 11, 2019 at 22:08
  • 1
    @Null Maybe the earlier notifications could be merged into the new one? Or there could be a single notification "There's a new HNQ thing" with a link to a list, and the notification is just bumped whenever needed. Mar 11, 2019 at 22:27
  • 61
    Maybe just a dull "Current local HNQs" in the mod dashboard. I smell a possibility for a userscript here... Mar 11, 2019 at 22:55
  • 13
    @Null Remember that one of the announced changes is that you'll only have 5 questions on the list at any given time.
    – jscs
    Mar 11, 2019 at 23:22
  • 14
    Maybe add a feed to mod room for questions that hit HNQ .
    – A J
    Mar 12, 2019 at 2:52
  • 1
    @AJ There is such a feed in this room, which was hacked together before these updates.
    – rob
    Mar 12, 2019 at 10:50
  • 4
    @JoonasIlmavirta this answer which is based on this answer explains about HNQ feed on chat room. Mar 12, 2019 at 11:04
  • 1
    @rob Does that feed actually get all of them? I had one added to the A&C chat room but it was only able to grab the top 50, not the top 100.
    – Catija
    Mar 12, 2019 at 13:11
  • 2
    In response to this answer and especially @Gallifreyan's followup suggestion, I've bountied this old question in the hopes of getting a new official response. Mar 13, 2019 at 12:26
  • 3
    @Randal'Thor I mean... you can. But I've already said that some sort of indication is on the table, we just need to figure out how... but that's not going to get an answer within the week.
    – Catija
    Mar 13, 2019 at 21:47
  • 1
    @gnat That's probably not impossible but what I'd rather do is just ignore questions that have 3+ close votes. That wouldn't require a permanent kick in the chance that the post was fixed and the votes were retracted or the post was reopened.
    – Catija
    Mar 14, 2019 at 15:11
  • 2
    @Catija I think that would be OK, too - penalize hot questions having 3-4 close votes :)
    – gnat
    Mar 14, 2019 at 15:14
  • 5
    We already have a mechanism for alerting mods that there's something they should look at on a post: flags. If "post enters HNQ" raised an auto-flag, then (a) it would alert mods and (b) when one mod handles it the notification is cancelled. Mar 22, 2019 at 14:20
94

Thank you for this.

All of these changes are between positive and extremely positive, and they show the change of stance that we were asking for last year.

There's still much more to be done, as well as missing historical data that will make it harder for SE communities to evaluate the effects of the changes, but everything here is a step forward.

I'm looking forward to seeing how this feature evolves, and to the next round of engagement for further design iterations!

66

In practice, how does the "at most five from a given site" work?

I can think of two possibilities:

  1. Once a site has five HNQs, no more can hit HNQ until one of those five drops off.
  2. The "hottest" five questions at any time are able to go to HNQ.

I guess option 1 is much easier for you to code, but option 2 is much preferable. If your Q&A gets a lot of quick votes and would be HNQ eligible, it shouldn't have to wait 3 days just because some other questions got there first. So it would be nice to know which of these is what actually happens.

12
  • 7
    I would imagine it is actually option 2 and it is recalculated every time the job runs. However, the problem, which is largely unavoidable, is that once a questions goes to the HNQ it is going to get hotter faster than one that isn't there, generally. So you're still going to end up with a situation that is largely similar to option 1 anyway. Mar 12, 2019 at 10:15
  • 2
    The hottness formula of a question are described here.
    – llrs
    Mar 12, 2019 at 10:31
  • 2
    @llrs I know; that's not what I'm asking about. Mar 12, 2019 at 15:02
  • 50
    @Randal'Thor It's the second. The formula picks a set of questions, which we then narrow down - throw away questions picked > 3 days ago, remove questions with MathJax in titles and questions that aren't in English, then trim per-site options down to 5 hottest, and grab top 100 for HNQ out of that. Note that the list is cached (and, under normal circumstances) recalculated every 15 minutes. Any given question isn't guaranteed to spend 72 hours in the list.
    – Adam Lear StaffMod
    Mar 12, 2019 at 16:03
  • 1
    @AdamLear Cheers, that's what I'd hoped for (but wasn't sure if it'd be too expensive in calls/processing time/whatever you programmers call it) :-) Mar 12, 2019 at 16:11
  • 2
    @AdamLear Thanks for the clarification, but there's an argument to be made that, in practice, it's really the first one, since the incumbents already have the usual feedback loop pushing their scores up. (This also happened before, but the competition was a network-wide run for 100 places, now it's site-wide for five.) So that's probably an important aspect of the data to keep an eye on, to see if the new caps don't make it more important to age the hotness score faster.
    – E.P.
    Mar 12, 2019 at 17:48
  • 2
    @E.P. Possibly, yeah. And all this is further complicated (in the sense that it's harder to establish stable causal relationships) by the fact that what shows up in the sidebar on Q&A sites is randomized. We'll have to see how it goes. The upshot is that we have a better setup now that allows making changes to either the algorithm or the filtering easier, if we need to go that route.
    – Adam Lear StaffMod
    Mar 12, 2019 at 17:51
  • @AdamLear I suspect that the missing historical data from the days before the cap was implemented is a much bigger problem there. But it's good to know that the platform is more robust now.
    – E.P.
    Mar 12, 2019 at 17:54
  • 1
    @AdamLear So in other words, anyone with edit permissions is able to unilaterally knock questions off of HNQ on any site that supports MathJax? That's a little silly. I suppose it'll come in useful though... Mar 21, 2019 at 22:43
  • @forest This is how the HNQ has worked for years, not a new development.
    – Adam Lear StaffMod
    Mar 21, 2019 at 22:47
  • @AdamLear I'm aware. It's just a "feature" I hadn't known about until now. It's still silly. Mar 21, 2019 at 22:47
  • 1
    @forest yep Mar 22, 2019 at 8:49
52

First of all, these changes sound great, so I just want to say thanks!

Having said that, could we also consider adding a feature, perhaps one that can be enabled on a per-site basis, that would temporarily increase the rep threshold to cast an upvote to the same as that required to cast a downvote while a question is on HNQ? Or, alternatively, not count the association bonus when computing the rep requirement for upvoting while the question is on HNQ?

Particularly on the sites that tend to cover a lot of controversial topics (Politics and Skeptics come to mind, for example,) there has been a longstanding problem of votes changing from meaning "This answer is useful, well-researched, and answers the question that was asked" to meaning "This answer agrees with my biases" when a question hits HNQ. The flood of users who have never contributed to the site and aren't familiar with how it operates can't cast downvotes, but they can (and do) cast upvotes because of their association bonus.

The upvotes coming in from 101 rep users who aren't familiar with the site can drown out the downvotes attracted from the site's users on a poor-quality answer, since the people coming in from HNQ off of massive sites like SO often vastly outnumber the regular users of the site who can downvote.

The reason that I suggest this being site-specific is that I don't think this is a major problem on the sites that don't get a lot of controversial content and, indeed, the extra upvotes might even be helpful on those sites. For example, a large portion of my SE usage has been on Aviation and I don't really see this as a problem there. But I've seen it again and again and again on sites that frequently cover more controversial topics.

8
  • 1
    This is a great suggestion. While it is unimplemented, regular users who observe this kind of mis-voting should be encouraged (via local metas) to cast a flag requesting that the question be manually un-HNQ-ed.
    – rob
    Mar 12, 2019 at 11:08
  • 11
    I'd like to see this. It seems like many (not all) HNQ visitors vote based on Truthiness (and often whatever answer is already at the top) without necessarily having the expertise to judge whether those answers are actually workable, accurate, suitable, etc, hampering the ability of other up-and-coming better-working answers to gain comparable traction. Mar 12, 2019 at 11:22
  • 12
    @doppelgreener I think casual users (like myself in the past) also have a tendency to vote based on interesting/entertainment value instead of quality/correctness on HNQ. Nowadays, I don't vote on any HNQ at all, unless I also have an expertise on that site. Mar 12, 2019 at 11:32
  • 5
    That fractures the Q&A engine - you effectively have two different types of system behaviours depending on whether a question is on HNQ or not (and a bunch of users won't understand what that even is, or why the system behaves inconsistently). But you can just disregard association-bonus-only upvotes when calculating the hotness score, without touching the Q&A engine at all, and it will have the same effect. For more details on that version, see this previous proposal, as well as the links here.
    – E.P.
    Mar 12, 2019 at 12:25
  • 8
    @E.P. There's already things which introduce two different types of system behaviours depending on a binary switch: consider Protected status for example. If visibility is an issue, voters can be informed that since the question's on HNQ, only people with some rep on the site can vote on it. Mar 12, 2019 at 15:26
  • 7
    @E.P.: Removing these low-quality upvotes from HNQ hotness helps, but leaves the answers with low-quality, inaccurate scores, and thus dilutes the value of SE expert Q&A. So it would be best to prevent them entirely. Mar 12, 2019 at 19:50
  • @NathanTuggy That may well be, but it's a much more drastic change. If SE are OK with that kind of scope (particularly including all of the design work to make sure it's clear to all users why the system's behaviour changed), then all the better. Given the years of neglect, keeping a tight scope made much more sense in 2016 - but maybe SE is more willing to consider cutting up parts of the engine now? It's hard for me to tell whether the current effort is willing to go that far or not.
    – E.P.
    Mar 12, 2019 at 20:16
  • 1
    Relevant: meta.stackexchange.com/questions/317418/…
    – Tim
    Mar 14, 2019 at 19:35
47

There is a race condition between

At this point in time, mods can only use this tool on a post currently in the HNQ list - they can't use it preemptively to prevent a question from being added to it.

and

questions right on the edge of the list may pop on and off the list several times, as frequently as every fifteen minutes

If a moderator discovers (via a flag or whatever) that a question is suffering from HNQ problems and should be excluded, but happens to examine the question during a fifteen-minute interval where the question has been bumped off of the HNQ list, then the "remove from HNQ" tool is not available. So the moderator goes to bed, the question works its way back onto HNQ, and the next day there is a mess on the question that could have been avoided.

It would be nicer if the "remove from HNQ" tool was available on all questions that (a) have an "added to HNQ" event in their history (b) which is younger than the 72-hour time cutoff.

4
  • 32
    Wait, you don't like playing hide-and-seek with it?
    – Catija
    Mar 12, 2019 at 20:34
  • You don't actually need condition b). Although selecting the "remove from HNQ" tool will be pointless if it applies. Mar 14, 2019 at 15:47
  • Actually, you could really just get rid of condition a), and have that checkbox on all questions younger than 72 hours.
    – Cullub
    Apr 1, 2019 at 1:44
  • 3
    @Cullub The time limit is from the first time the question reaches the HNQ, not from the time the question is asked.
    – rob
    Apr 1, 2019 at 13:09
43

If someone notices that a question's attracting "negative attention", they may well suspect that it's on the HNQ list: they shouldn't need to search that whole list to confirm that it is, prior to excluding it or suggesting its exclusion. (The edit history will only show whether a question once appeared on the HNQ list & whether it was excluded.)

If a question is currently on the HNQ list, it would be useful to have some visible indication of that on its main page.

13
  • 15
    Finding some way to indicate which questions are on the HNQ is on our list of things to look at, whether that's for the mods only or for everyone. There are a bunch of ideas - a [HNQ] title marker like [protected] is an idea. We could also have a list similar to the bounties tab that only exists when there's something on the list... we just need to figure out a reliable way to do it. Mods can actually tell if a question's on the HNQ by checking whether they can kick it off or not. If the button can be clicked, it's on the list.
    – Catija
    Mar 12, 2019 at 13:34
  • 5
    @Catija SOX has an option to add an indicator to show if something is on the HNQ or not, I imagine they scrape the HNQ page to do it but might be worth seeing how that works. Mar 12, 2019 at 13:35
  • 5
    @Catija: (1) If non-mods are going to be asking mods to exclude questions from the HNQ list, then it'd be a good idea for everyone to be able to see the indicator - to save time & avoid mistakes. (2) Thanks for the tip. Mar 12, 2019 at 13:52
  • Yep! Before I shortened my comment because it didn't fit, I'd stated that there's no reason for it to be mod-only. :P And you're welcome!
    – Catija
    Mar 12, 2019 at 13:53
  • @TheLethalCoder what is SOX?
    – LShaver
    Mar 12, 2019 at 13:57
  • @LShaver Stack Overflow Extras a userscript that adds some extra functionality to the network. Mar 12, 2019 at 13:58
  • 1
    @Catija How about just some discrete text "Hot Network Question" on the top of the right-hand sidebar, immediately below the asked / viewed / active indicators?
    – E.P.
    Mar 12, 2019 at 14:50
  • @E.P. May be better under the voting arrows/favorite star. That right sidebar disappears on some viewports and we're already trying to figure out what to do with that content, so we probably wouldn't put it there. :)
    – Catija
    Mar 12, 2019 at 14:52
  • @Catija Everything there is (primarily) actionable, not informative. It would be some pretty jarring design to mix those two functions.
    – E.P.
    Mar 12, 2019 at 15:04
  • 3
    Why not list it as a post warning "This question is currently in te HNQ list, and may see increased activity/views...etc." Mar 13, 2019 at 11:24
  • @Pureferret That's not a bad idea (though the copy would need some work) but I worry that it might be a bit too visible... also, the post notice world is a bit complicated and many HNQs get protected, which then puts two post notices on the thing which starts to take up a lot of space... if you get where I'm going?
    – Catija
    Mar 13, 2019 at 22:04
  • @Catija I do get it. Just a suggestion :-) Mar 14, 2019 at 0:10
  • All these indications would make hot network questions even hotter, and I don't know whether that is desirable.
    – mathlander
    Oct 28, 2023 at 0:55
37

Right now we don't have an event for a post dropping off the list. The concern with this is that questions right on the edge of the list may pop on and off the list several times, as frequently as every fifteen minutes, thus cluttering up the history.

Maybe make every question only have at most one "Achieved HNQ" event and one "Knocked Off HNQ" event?

So if a question has been on the HNQ list and then bumped off, it'll have both Achieved and Knocked Off events. If this same question somehow winds back up on the list, instead of creating a new event, why not remove the Knocked Off event? It'll make the history look cleaner and for all intents and purposes will be mostly correct (every post with an Achieved event will eventually have a matching Knocked Off event).

I think this is more informative than only having an Achieved event and not so crazy that it'll confuse everyone.

4
  • 4
    Rather than deal with deleting events (which doesn't make sense if an event causes your phone to buzz with a notification—can't unbuzz the phone!), there could be one event for when a question has achieved HNQ, and another event for when a question cannot return to HNQ, either because its 72 hours are up, or because it was closed, or because it was mod-removed from HNQ, or whatever. Mar 12, 2019 at 0:18
  • 5
    @SqueamishOssifrage This sort of history event doesn't have a notification associated with it. :)
    – Catija
    Mar 12, 2019 at 1:23
  • 7
    There's definitely an event for mods removing a post - In addition, an event will be logged in the post timeline and edit history that indicates when it was removed and by whom
    – Catija
    Mar 12, 2019 at 1:24
  • 2
    @Catija oops! I must've totally missed that.
    – scohe001
    Mar 12, 2019 at 11:53
36
+50

Cut down on the HNQ effect by reducing the number of answers that go into the hotness equation.

Right now between a laser-focused question with one good answer and a broader, more controversial question, the broader question will stay hotter longer because as more answers are added, the hotter the question gets up to a limit of 10 answers.

So a question will go hot and as people come in from HNQ and add more answers it creates a positive feedback loop that makes the question that much hotter.

Reducing the number of answers in the hotness equation would even the playing field between the narrow and the broader questions and I think would encourage better and more tightly focused questions and answers.

6
  • 23
    cut off at 10 answers is an extremely outdated parameter in hotness formula. It only made certain sense many years ago when the list was heavily dominated by Stack Overflow, with its unique scale, scope and moderation style (and even at SO is is considered rather troublesome). As of now, this setting only encourages trolling, broad discussions and big lists - that is, things diluting the value of Q&A model...
    – gnat
    Mar 12, 2019 at 9:28
  • 8
    ...the only sensible reason I can think of for it to remain untouched in this round updates is desire to avoid making too many changes at once. I can understand that - but I would be much surprised if it doesn't change to something more reasonable (like 4 or 5) in the next update
    – gnat
    Mar 12, 2019 at 9:29
  • 10
    Yeah. I'd like to look at the suggestions here that weight answers a bit to exclude zero-score answers from impacting the hotness score. And yes, @gnat we're avoiding too many changes but also wanting to see the effect of those changes.
    – Catija
    Mar 12, 2019 at 12:45
  • 3
    speaking of zero score answers @Catija - I wouldn't rely on these because inexperienced askers can break things with indiscriminate upvotes out of gratitude. This isn't a problem in regular questions because of 15-rep requirement for voting, but in HNQ this barrier naturally breaks
    – gnat
    Mar 12, 2019 at 21:17
  • 2
    @gnat The actual formula recommended in that link does more than just ignore zero-score answers, that's just how I quickly phrased it. :) If two answers have 20-30 score each and a dozen have 0-4 each, they're probably not worth counting as much... so they're given lower weight based on their score.
    – Catija
    Mar 12, 2019 at 21:20
  • 1
32
+100

Have the "top dogs" concede part of their HNQ slots to one of the "much smaller dogs" once a week

Having 5 slots each is very nice as it allows for more candidates sites to reach it. Yet... I'm guessing that a lot of the much smaller beta sites won't be seen on it for some weeks/months.

I'm guessing that out of the 140+ sites, some four categories can be made:

  • HNQ regulars (see the answer to Which sites appear most often in the Hot Network Questions list?), who probably don't have much to worry about;
  • a couple of sites who aren't "top dogs", yet do appear rather regularly (Movies, Arqade come to mind);
  • sites with the occasional HNQ (rather beta sites, let's say Chess and Literature for instance);
  • and sites who never, ever go there (in one year of using SE, I'm not sure I've ever seen a Coffee question on HNQ).

As part of the goal is to give the whole network a chance, it would seem sporting that the "regulars" offer a chance to the fourth category to get some more attention, say with a workflow like the following:

  1. At the end of each week, out of the 10 sites that had the most questions on HNQ during this week, 3 (let's call that parameter N) get selected. For the first 3 (charityDays) days of the following week, they'll have 4 slots max
  2. An algorithm determines which 3 (N) sites haven't been on the HNQ for the longest
  3. the "free" slots are given to, say, the current hottest question on the sites from point 2., they're being granted a minimum amount of time on HNQ (grantedTime)
  4. when said minimum amount of time has expired, the question is given an "offset" amount of hotness points, and is now treated as a "normal" HNQ question. If it fades out, it fades out, it had its chance!
  5. In the meantime, algo from 2. re-determines the lowest sites, puts them on the now free special slot, and repeat until charityDays has passed.

For the minimum repartition to be fair, I'm guessing charityDays has to be a multiple of grantedTime. Obviously, N, grantedTime, charityDays are open to suggestions, 3 days might be too long, the figures are just here for the example. (and if the wording of "charity" days offends anyone, feel free to change it)

Of course all the sites can still make their way to HNQ the "normal" way during that time, may they be of the scale of Stack Overflow or Augur.

The top dogs would probably not have much harm from this, as they're healthy sites; there's the "random" selection which partly ensures Worldbuilding doesn't have to concede a slot four weeks in a row; and they get their normal chance back after charityDays have passed.

It sounds like a win for the "little" sites, who are kind of ensured to be on the list at some point, and the "sites who haven't been on HNQ for the longest" should shuffle itself as having been on HNQ during week n means you don't get a "free spot" the following week.


I do realize this is overseeing some of the "bad" effects of the HNQ, for instance, the touchy subjects IPS/The Workplace have to handle. This suggestion treats the HNQ as some "chance" which would help the target sites to get some needed traffic; as my guess is that they're more niche-y subjects with less dire real-life consequences. Obviously, that's partly a generalization.


Possibly, some of the "top dog" communities won't like it, and feel it's unfair, that their righftul slots are theirs, etc, etc. I think it's only sporting to give the chance you have to someone who doesn't, and I'll quote this page of an absolutely great comic:

They didn't realize that their freedoms​ were astonishing luxuries that the rest of us could never dream of. Their baseline decency is an incredible paradise for most of us.

4
  • 11
    You could simplify a little by just lowering the default max from 5 to 4 and/or reserving a few slots for sites that haven't been on HNQ in a while. Would still need special casing to handle the foreign language sites which are intentionally excluded because of the "must have 75% of the title be English" requirement (e.g. SO in <another language>, or Russian language questions asked in Russian, etc.).
    – Troyen
    Mar 12, 2019 at 1:06
  • @Troyen much simpler indeed... I guess I got a bit carried away ahah :)
    – Jenayah
    Mar 12, 2019 at 7:21
  • 6
    this is probably related: New Resources for Our Stack Exchange Network "Help smaller communities get some exposure in the Hot Network Questions list... top 5 sites (Workplace, SO, SFF, Worldbuilding, Code Golf) get 50x to 200x(!) more clicks than bottom 10 (Hinduism, Android Enthusiasts, Latin, Chess, French, Law, Board Games, Russian, Mi Yodeya, Christianity, Motor Vehicles). It gets even worse if you take into account that mentioned snippet lists only 70 sites of total 130+, meaning that about 60 sites get even less views..."
    – gnat
    Mar 12, 2019 at 11:38
  • Perhaps a second list called "needs some love" or something with questions that are low on views and are lacking answers from all around the network, preferably at least couple days old?
    – ave
    Mar 14, 2019 at 12:44
26

What about the other Stack Overflow sites, the ones not in English? As two recent questions have mentioned, these sites have started to show up in the Hot Network Questions. (This could also apply to the Russian in Russian site, although it's very unlikely to happen in practice.)

I heard that this is because of a change; instead of looking at the body and title to see if something is in English, only the title is looked at. Programming languages are mostly based on English as are the errors so it's all too easy for the title to be mostly English when nothing else in the post is.

Because nothing else in the post (or the site) is in English, it's a long shot that anyone coming from other sites would be interested in it. The HNQ list doesn't even show up on the non-English Stack Overflow sites! Are these sites supposed to be on the HNQ list? Or would they have to elect to remove themselves from it? Ideally I'd be hoping for something more customizable, so that people who speak these languages can see them (and maybe also limit them to people who are interested in programming too?).

11
  • 27
    Story time! The way English detection worked in the past is that we literally had a 4MB text file with "English words" in it that we checked posts against. As part of this work, we opted to stop using that and switch to a language detection library. The downside there is that it has a tendency to mis-identify mixed-language posts (e.g. "I could say X as Y or Z in this language, which is correct?" on language learning sites). So we thought "hey, if the title's in English... that's probably good enough." Cue narrator voice: "it wasn't". :)
    – Adam Lear StaffMod
    Mar 11, 2019 at 21:51
  • 10
    The detection library does allow some amount of configuration... it's on my list to find a reasonable compromise there, but I didn't want to delay the launch of everything else on that.
    – Adam Lear StaffMod
    Mar 11, 2019 at 21:51
  • 4
    @AdamLear That sounds absolutely horrifying. Why not do it properly and determine HNQ eligibility based on the site itself? Mar 12, 2019 at 7:09
  • 7
    @forest At least the site I mod has both questions in English and questions in Italian (and we usually ask for the answers to be in the same language as the question). I suspect this is the case for most language sites. Mar 12, 2019 at 8:15
  • 3
    @AdamLear I can see the dilemma here... and this is maybe my selfish request, but until the user preference's on site customization for HNQ is implemented, I think removing localized sites (yes, localized, not the [language].SE one) is better for now... Mar 12, 2019 at 11:21
  • 6
    @AdamLear Why not leave them on? Some of us are Spanish, Russian, Japanese, or Portuguese-speaking programmers who would be as interested in the localised Stack Overflow HNQ posts as we are in our English Stack Overflow HNQ posts. Many of us won't be interested / able to understand the topic, but that's the case for a lot of HNQ posts anyway depending on the reader. Mar 12, 2019 at 11:31
  • @Bookends I need to beef up language detection either way to make sure [language].SE sites don't get left out, so that will probably effectively exclude international SOs.
    – Adam Lear StaffMod
    Mar 12, 2019 at 15:14
  • 5
    @doppelgreener I'm sympathetic to that argument, but I also think there's a difference between not being able to understand a question due to its subject matter vs not being able to read it in the first place. I personally don't feel super strongly here either way, but we do get a solid stream of complaints every time localized sites end up in HNQ, so ¯\_(ツ)_/¯
    – Adam Lear StaffMod
    Mar 12, 2019 at 15:17
  • I guess dealing with events that generate streams of complaints is a good reason to not leave them on. :P Mar 12, 2019 at 15:19
  • 9
    @AdamLear there are many, many questions in HNQ with titles in English which are incomprehensible without specialist background knowledge so having a few in other languages, even other writing systems, hardly seems a problem.
    – mdewey
    Mar 12, 2019 at 16:52
  • A suggestion to beef up language detection was already offered and rejected by the community.
    – Rob
    Mar 16, 2019 at 10:23
23

Each site can only have a max of five questions on the HNQ list at any given time.

So wait, unless I'm missing something, home SFF currently has six questions on the list (warning, some may be Captain Marvel spoilers):

I'd blame caching, but the way I always understood it, the list is refreshed every fifteen minutes, so I thought the hasRemainingSlotsOnHNQ() check is done at this time, but maybe it isn't? In any case, the "five max, any time, any place" doesn't seem to apply here.

To clarify, I'm just wondering if maybe a parameter wasn't set right, or, if it is indeed caching, shouldn't an additional check/reduce cache time/something be done, to make sure to respect the "at any given time" wording?

11
  • 21
    shouldiblamecaching.com Mar 11, 2019 at 20:49
  • 4
    @Randal'Thor "I'd blame caching..." :P
    – Jenayah
    Mar 11, 2019 at 20:49
  • 7
    Looking into it. Thanks for mentioning that.
    – Catija
    Mar 11, 2019 at 20:54
  • 5
    Looks like this is "fixed" but let me know if you see it happen again so we can look into it more closely.
    – Catija
    Mar 11, 2019 at 21:49
  • 5
    Five is at least two too many.
    – shoover
    Mar 12, 2019 at 0:28
  • @shoover for SFF or in general?
    – Jenayah
    Mar 12, 2019 at 0:30
  • 2
    @Jenayah In general, I'd say.
    – TylerH
    Mar 12, 2019 at 0:56
  • 4
    @TylerH then that should probably be an answer, or a comment under the main question, I guess :)
    – Jenayah
    Mar 12, 2019 at 0:58
  • also, I fixed the spoiler :)
    – KutuluMike
    Mar 12, 2019 at 2:34
  • @Jenayah In general.
    – shoover
    Mar 12, 2019 at 2:57
  • 7
    The 3rd one is clearly a Hoth question, I don't know about the others (hides).
    – Lundin
    Mar 12, 2019 at 14:19
18
+200

These are good changes. Thanks for making them!

One of the factors in determining whether a question is "hot" is the combined score of the answers. Questions without answers aren't eligible for HNQ.

With a history event indicating when a question went "hot," it becomes possible to consider the scores on pre-HNQ and post-HNQ answers separately. I propose that downvotes on answers added after a question has been promoted should be given some extra weight in determining whether a question remains "hot." A question which gets promoted because of an exceptionally good answer, but then attracts a bunch of negative-scoring or controversial answers, is a question that is perhaps not benefiting from the broader exposure.

An alternative version of this proposal would be to give all downvotes more weight than all upvotes when determining "hotness." That would give the local community more influence than newcomers and passersby in determining which questions remain hot, since newcomers will generally not have the downvote privilege, without necessarily requiring someone to manually cast and respond to a flag.

4
  • 5
    First suggestion is too complex, IMO. Second suggestion is more simple and makes sense, yes. Mar 12, 2019 at 11:51
  • 1
    @ShadowWizard How so? This is a complex calculation, sure, but it's only being performed on ~100 questions on a once/15min basis.
    – E.P.
    Mar 12, 2019 at 12:20
  • It's adding extra complexity to already complex query, while the second suggestion won't. :) Mar 12, 2019 at 12:22
  • I had the second idea while I was writing the first. I'm starting to like the second idea better myself, but don't have time to edit for now.
    – rob
    Mar 12, 2019 at 13:45
14

Right now we don't have an event for a post dropping off the list. The concern with this is that questions right on the edge of the list may pop on and off the list several times, as frequently as every fifteen minutes, thus cluttering up the history. We'll look into whether we can find a neat way of achieving this but for now there's another feature that will help with this:

Questions will age out of the list after being on it for 72 hours.

Adding history entries every time a question bounces on or off the list could be noisy, as you say. For the sake of SEDE, could you add an event once when a question becomes ineligible? You're already doing that for one of the two ways this can happen -- a moderator kicks it out. Can you add the other one, when the question ages out? Yes we can figure it out ourselves based on dates, but having an explicit "exited HNQ" entry for every "entered HNQ" entry means query-writers don't need to do conditional logic (if there's an exit event use that, else do date math). It might also help people scanning a post's history on the site (timeline or revision history).

4
  • 2
    How is this really different than this answer? Mar 13, 2019 at 1:03
  • 5
    @SupaMegaDuckyMomodaWaffle that answer proposes updating for every event but only keeping the latest (editing history). I'm proposing that we keep first on and last off, only, and that we always add the latter. Mar 13, 2019 at 1:05
  • 4
    So, there's kinda a reason I wouldn't want to do this... Eventually I'd like to know if a question actually ages off - meaning it was still on the list when it hit the 72 hour max and had to be kicked off. There's a lot more I'd like to do here for tracking. I worry that if we do this, it might clutter that data up. So, we'd have to be careful about the event essentially differentiating aged out but wasn't necessarily still on the list from removed and still on the list.
    – Catija
    Mar 13, 2019 at 1:59
  • 5
    @Catija oh, I hadn't thought about that case. I guess what I'm describing is a "no longer eligible" event, as opposed to an "exited" event, but that might not be all that useful either. What I think we both really want is, at the 72-hour mark, to add the actual exit time to the history, but that's more work. Mar 13, 2019 at 2:32
13

Thank you for the changes! May I suggest also allowing the OP to remove their post from the HNQ?

Not all posts need to be under a spotlight as they may attract excessive or undesirable attention and sometimes the OP may be aware of that.

5
  • 3
    "Not all posts require undesired attention" What do you mean? What is "undesired attention"? Why does one asks a question if they don't want attention on it? Are you talking of some kind of meta effect? Mar 15, 2019 at 17:30
  • 17
    I understand your request and hadn't thought about this option. At this point it's unlikely that we'll add this but, as the OP, you can always flag your own post and request that your question be excluded along with an explanation. Be aware, though, that being the OP doesn't mean you necessarily have the final say. Removing it from the HNQ list also impacts the people answering the question.
    – Catija
    Mar 15, 2019 at 17:31
  • 3
    I was thinking along the lines of a question on Workplace that the asker would not want their boss to see. Nor is all the traffic from HNQs productive, espescially in the case of controversial questions. Mar 15, 2019 at 17:35
  • 11
    @MohammadZuhairKhan: Askers are not generally likely to be good at knowing which questions are going to generate productive traffic, so they should leave the final determination to those with actual experience figuring that out. As far as anonymity, if you're worried about your boss seeing the question, ask from an unrelated account and change all the names. Just hoping the question doesn't make it to HNQ is highly ineffective. Mar 15, 2019 at 17:37
  • 3
    @Catija Still, guidance from SE to moderators to give special consideration to take-me-out-of-HNQ flags from the OPs would probably go a long way.
    – E.P.
    Mar 17, 2019 at 21:54
11

The one feature that I have always wanted is the ability for me as a user to block certain sites from showing up. I have no interest in voting Movies & TV off the hot list, but I personally never want to see it there for several reasons that are individual to me. I want to see the hot list for the various technology sites, though, so I'd rather not have to turn the feature off entirely.

4
  • 3
    This is probably not going to be implemented: “3.Doesn't let you customize which sites you see by either a whitelist or blacklist. ... Solving #3 is likely very complicated to implement so we don't have plans for it for the time being.“ (from the question)
    – MEE
    Mar 17, 2019 at 18:48
  • 2
    Still, this can be implemented client-side with a suitable user-script - I don't have the link handy, but it's out there.
    – E.P.
    Mar 17, 2019 at 21:51
  • 1
    A custom stylesheet can do this: meta.stackoverflow.com/a/349111
    – jscs
    Mar 18, 2019 at 16:03
  • @MEEthesetupwizard Yes the question does say that. But I think that's a mistake as they may well have underestimated how many users would like that feature. I have often wanted the ability to hide specific sites from the list and this answer shows that I am not alone.
    – kasperd
    Mar 24, 2019 at 14:51
9

Make it easier for moderators to find flags cast on HNQ.

Make a visual indication or some way to filter / prioritize for moderators flags that are cast on hot questions - or more precisely on questions having HNQ entry in history, being open and eligible by age. This would help moderators handle such flags in more timely manner. (This very request is sort of follow-up to discussion of a particular flag that was handled slower than desirable because it was buried under flags on regular posts.)

It would be unreasonable and unfair to expect flaggers and moderators to do cumbersome legwork of checking and communicating if the flag is on the hot question when all the needed information is already available in the system and can be obtained and communicated automatically.

Worth noting that 100+ upvotes to this answer suggest that there is generally high demand to improve mod tooling with regards to maintaining hot questions.


I don't have strong opinion on whether to do this to flags on answers and comments or only on questions. That said, I am inclined to think that it would be better to have it for all kind flags because I regularly see reasonable questions (not worthy of any flags) polluted with troublesome answers and comments from HNQ passers-by.

9

Questions must now be at least eight hours old before they can be added to the HNQ list. ~Catija

Set a minimum age for posts before they're eligible for the HNQ list.

(this idea was already discussed elsewhere - posting it here because it looks like a very worthy addition to that "master list")

First time I saw it proposed by SE CM (Shog9) in a discussion of particular hot question at TWP meta:

delay non-SO questions by some significant period of time - a day or more maybe - to give folks time to perform the sorts of cleanup and moderation that necessarily take longer here than on SO...

Later I saw similar idea mentioned by another SE CM (Catija) at IPS meta and here at MSE:

This gives users on a site the chance to edit the post so that it's suitable for the potential network-wide consumption or close it if it's eligible for closure - and to even get it edited and reopened! It also gives the experts on site the chance to vote on - or even delete - answers. Essentially, more eyeballs will mean it's less likely that a problematic post will make it to the HNQ list. For your moderators, there's a reduced need for constant vigilance when you've got a, say, six-hour delay before a complicated question can start causing problems...


My own take on above suggestions is, while I agree with them in general, I think we could start experimenting with smaller delay, like 3-5 hours (and only increase it if it later turns out that it is indeed not enough). By design, questions get hot when they quickly start attracting more than average attention of site regulars so that we can reasonably expect that in the absence of "deafening" upvotes, answers and comments from sidebar passers-by this attention of active regulars can be also quickly focused on the needed cleanup actions.

Another thing I'm not comfortable with is strict focus on smaller sites. While I agree that waiting for a full day would make no sense for Stack Overflow (though I also can't see how it could be harmful), I think it still could benefit of some smaller delay like for a few hours. In particular, this would give holders better chance to close duplicates which sometimes pop up in hot questions from SO (and kind of incentivise asking and answering obvious duplicates due to extra upvotes coming to HNQ).

13
  • 4
    This is already in progress. We're doing some analysis of the current impact of the HNQ on sites and we'll be adding this once we have some idea of the impact we expect from this change.
    – Catija
    May 7, 2019 at 11:03
  • @Catija thanks for update, I saw you mentioned planning for this in other discussion but wasn't sure. Your concerns about impact are understandable, especially if you're considering extreme change like one mentioned by Shog, "a day or more maybe"
    – gnat
    May 7, 2019 at 11:12
  • @Catija implementation of delay appears somehow a bit leaking: I just noticed two questions newer than 8 hours shown in the sidebar. It's interesting that their revisions history doesn't show HNQ event: first, second...
    – gnat
    May 21, 2019 at 11:52
  • ...I think I saw something like this earlier today with this question: it first popped in sidebar without HNQ event in its history, then disappeared for a while and at 8 hours age entered HNQ again, this time with respective event in rev history
    – gnat
    May 21, 2019 at 11:56
  • 1
    Can you try to get a screenshot of the HNQ next time you see this? Is it in the main list or the sidebar (both?)?
    – Catija
    May 21, 2019 at 12:32
  • 1
    @Catija OK will try to catch it and make a screen shot, and if I find it will also try to check if it gets to main list or not
    – gnat
    May 21, 2019 at 12:41
  • @Catija there you go, I just observed this problem in sidebar when viewing this question - see red marks at sidebar screen shot and when checking the list (page 2) I discovered these questions in there, along with some others - see red marks at list screen shot
    – gnat
    May 21, 2019 at 14:51
  • Weird. I can't reproduce that. I don't suppose you have any scripts that may be impacting it? I've got someone looking but if you can get me more info, that'd be great.
    – Catija
    May 21, 2019 at 15:02
  • @Catija no scripts, nothing at all. Just viewing sidebar and hot list from plain Firefox 66.05 / Win 10
    – gnat
    May 21, 2019 at 15:04
  • ...with regards to reproducing, is has to be difficult because the issue is intermittent, just as I felt when I first asked you: I just re-checked the list and all troublesome questions are now away from it, there are currently only questions older that 8 hours in there. So it is like system breaks temporarily, and after a short while cleans itself up leaving no traces. It's particularly tricky since as I mentioned these breaks in functioning don't leave HNQ events in the history of involved questions @Catija
    – gnat
    May 21, 2019 at 15:14
  • 1
    Hmmm... I wonder if it's related to the caching... the list runs every 15 minutes but I think it takes longer than that to actually run the list (the last time I ran it manually it took about 17 minutes)... so maybe there's something weird in the cache but once the new list is actually finished, it cleans itself up?
    – Catija
    May 21, 2019 at 15:19
  • that sounds like a plausible explanation - thank you @Catija
    – gnat
    May 21, 2019 at 15:32
  • 2
    @Catija to add some info, my question was only 2 hours old when it reached HNQ, is that normal? May 23, 2019 at 12:35
8

Making the new ability to hide the HNQ list have a local setting for the computer/browser would be nice for people who like to hide it at work/school to limit distractions from questions that appear in the list but like to see it at home and other locations where the distractions are not as big of an issue.

10
  • 5
    It's one click away, can't see reason to spend time of this. Mar 12, 2019 at 12:32
  • 4
    @ShadowWizard It is more than just a click away and besides that is not the point. Currently someone who wants it off at work but would like to use it at home would have to toggle the setting twice every work day. Once when they get to work and a second time when they get home which adds up to a lot of work for the individual user. Currently this can be done with browser add-ons such as stack overflow extras. However since they are working on changes to the HNQ functionality and added the ability to hide it I figured it would be a good idea to ask for local settings as well.
    – Joe W
    Mar 12, 2019 at 12:56
  • 4
    At the same time please don't drop the possibility to hide it globally. In case cookies are cleared regularly it very annoying having to redo such settings again and again as it's the case disabling the responsiveness Mar 12, 2019 at 13:25
  • 1
    This is what I'm indicating in bullet point 1. I think the easiest way for us to achieve this is to have it be cookie based in the sidebar. Because it's cookie based, it will be specific to domain at some level, so the behavior on SE.com sites will be different than AU, SO, SF, etc... at minimum... but we could make it per-site, too (that might be too granular, though)... and it'd be per-device.
    – Catija
    Mar 12, 2019 at 13:49
  • @Catija I misread that when I was looking at your post. Thanks for clarifying with your comment, and I am looking forward to the changes even more now!
    – Joe W
    Mar 12, 2019 at 14:06
  • 1
    @ShadowWizard, if it were genuinely one click away (eg. a toggle underneath the "related questions" list), I'd agree with you. But it's not -- it's four clicks and three page loads away.
    – Mark
    Mar 12, 2019 at 20:15
  • use adblock instead, that can be controlled per location.
    – Kevin B
    Mar 13, 2019 at 18:04
  • @KevinB Why when Stack Overflow Extras handles it with ease (and improves other areas) or as it appears they are looking to be able to do it on a per site/browser basis.
    – Joe W
    Mar 13, 2019 at 18:39
  • Because it works right now?
    – Kevin B
    Mar 13, 2019 at 18:41
  • @KevinB Maybe I am confused here but what does a post about making changes to HNQ and an answer about some possibly new functionality have to do about how it currently works?
    – Joe W
    Mar 13, 2019 at 18:54
8

Since other feature requests are being made here, I'll go ahead and add this here instead of making a new FR question.

Factor bounties into the hotness formula.

I know that bounties aren't directly a sign of hotness -- they're not votes or views or any of that other stuff that feeds into the formula. But bounties are requests for attention, just like hot questions in the hot list. Further, with bounties somebody has paid for publicity, which usually means that it's a harder or more specialized question. (Easy questions get good answers quickly and people can move on, after all.)

For example, this question of mine currently has a 500-point bounty and requires specialized tools knowledge that many users don't have. (If you say that it should have been asked elsewhere, that would just shift the problem.)

The HNQ list is all about drawing attention, just like bounties. Can they work together? I'm asking that the presence of a bounty feed into the hotness formula, not that a bounty is enough for the HNQ. (Bountying that -10 question won't help.)

It might be reasonable to set a minimum bounty amount; on my sites there aren't too many bounties at a time, but larger sites might be swimming in 50-point bounties. If so, you might want to only factor in bounties of 100 or 200 points or whatever value the data suggests is a natural breakpoint. I think we can find a way to let bounties influence the HNQ without having them take over.

4
  • 1
    I ran a query and there are currently 40 bounties worth 250 + on the network and an additional 26 worth 200, 16 worth 150 rep, and 149 worth 100 rep... And that's when people are assuming that their bounty'd only get them featured on one site, not all of them. That's... more than the entire list and even the 250+ is almost half of it. If someone's paying for a bounty and they're expecting that to mean it shows up on the HNQ, we then have to explain why we prevent some of these questions from showing up or why we let mods kick them out after the fee is paid. Any thoughts on this?
    – Catija
    Mar 29, 2019 at 4:47
  • 2
    @Catija I don't mean that a bounty should guarantee a spot on HNQ; I mean that it should feed into the formula, just like upvotes and answers and views do. In particular, by the time a question is eligible for a bounty views won't help (because the formula counts velocity), so a bounty could compensate for that. Nobody should ever think a bounty is a guarantee, and removal from HNQ should work the same for any hot question. Mar 29, 2019 at 14:04
  • 2
    Bounties can only be added two days after a question has been asked, if it hasn't gone hot in two days, the odds are really good it never will Mar 30, 2019 at 5:58
  • @CharlieBrumbaugh one of the recent changes is to limit questions to three days on the hot list. That tells me that even now, some of them qualify when they're two days old. If, on top of that, bounties provided a boost, then we'd get some of the ones that weren't originally hot but are important enough for somebody to invest in. That seems like a win to me. Mar 31, 2019 at 2:07
7

I don't have ready access to the numbers and am uncertain if I can look it up, but I have observed a significant difference in apparent traffic on the site as a whole on weekends. As such, if possible, questions that spend most/all of their 72 hours over the weekend should get a minor extension to increase their exposure.

11
  • 4
    I like this a lot. Maybe every weekend hour only counts as 1/2 of a Hotness hour?
    – scohe001
    Mar 12, 2019 at 11:56
  • 4
    It's not just apparent traffic. This is the traffic history on Physics (available at 25k+ rep, in the site analytics within the 10k tools). The dips are centered on the Saturdays (with traffic on the GMT Friday already including some weekend portions for timezones ahead of GMT, and conversely for Sunday). Ask a 25k'er on your local site to confirm what it looks like there, and what the depth of the modulation is.
    – E.P.
    Mar 12, 2019 at 12:31
  • 11
    Though that said, I'd be rather opposed to this kind of uneven clock. Three days is enough to span a weekend. Time in the spotlight isn't a "right" that questions have, if they get less of it because they were posted on a Friday, then too bad.
    – E.P.
    Mar 12, 2019 at 12:33
  • @bruglesco Nobody questioned the fact that it's network-wide. I'm not sure why you're attacking hard data that supports your observations.
    – E.P.
    Mar 12, 2019 at 12:49
  • 1
    As for "fairness", there's a whole ton of other aspects that require attention before this, and query complexity is indeed a limitation. There's deep structural issues with the mechanism, whose fixes are likely to face significant design, coding and performance challenges before they can be effective. The changes proposed here are little more than window-dressing, adding some substantial design and coding complexity just because some questions might get a fractional modulation in traffic.
    – E.P.
    Mar 12, 2019 at 12:54
  • @E.P. Complexity outweighing worth is more constructive than "too bad". Thank you.
    – user384163
    Mar 12, 2019 at 12:57
  • 2
    @bruglesco The "fairness" argument is still misguided, though. HNQ is there to promote sites, not questions that have somehow (how?) "earned" their spot there. There absolutely is an argument to be made that smaller sites are getting an unfairly low exposure on the list. But it isn't "unfair" to a question that it got less network-wide traffic because of when it was posted - that's what I mean by it not being a "right". If you do think that questions "deserve" such "fairness", then please do tell what merits they have that warrant that, and why the hotness formula captures that so well.
    – E.P.
    Mar 12, 2019 at 13:05
  • 14
    So, part of the reason it's three days is to make sure that questions will be around for at least one non-weekend day... so this was a consideration. We're trying to balance simplicity with fairness. I'd hate to end up with an explanation for this that is reminiscent of the fun English "I before E" rule... "I before E except after C... or in sounding like ay as in neighbor or weigh.... etc, etc, etc." :P
    – Catija
    Mar 12, 2019 at 13:57
  • This is why you ask questions during the week to get more points.... Mar 12, 2019 at 16:32
  • 6
    Nah @CharlieBrumbaugh , you ask questions on Sunday night in the US so they're easier to get on the list for Monday morning in Europe and stay there. :P
    – Catija
    Mar 12, 2019 at 19:17
  • @Catija You don't want to ask too early or you will hit the rep cap and miss out on internet points Mar 12, 2019 at 23:10
5

Allow excluding specific sites from specific sites' Hot Network Questions list

(To be clear here, this isn't a user preference, but rather a site-wide preference. As stated in the question, the team is not currently planning to build a user preference for this purpose, so I'm not proposing that here, but a lesser version.)

After reading through some per-site meta posts on the Interpersonal Skills site, I came across an answer which says:

Many sites on the network are technical/professional sites, and some segments of their userbases don't want or need to see our questions.

According to a famous micro-study by ArtOfCode, many new or anonymous users may not know what "Hot Network Questions" is:

HNQ was one of the most confusing features mentioned. On a page full of questions about gardening, looking at the sidebar and seeing questions about Star Wars and Windows 10 next to each other confused my test group.

As far as I can tell, that was partly at play when provoking the original complaint that resulted in the site becoming ineligible to contribute questions there. Unfortunately, this resulted in a massive drop in answers and caused the site to remain much more stagnant than before. I feel that the previous configuration helped provide more diverse answers to questions, which is very helpful there, whereas currently, most of its answers aren't from as diverse perspectives as before.

In my opinion, it's important to try and strike a balance both ways: having that site listed on technical/professional sites may not be ideal, not only due to a lack of understanding as to what the list is among new users, but also because of irrelevant titles and other reasons mentioned there, but questions on the site should be featured on other sites where others may have an interest in answering (e.g. sites in the "life" category).

For these reasons, it should be possible for a site's HNQ questions to be included on specific sites or site categories, but excluded on others, i.e. for instance, not show HNQs from, say, Ask Ubuntu, on the list on Interpersonal Skills (just a random pick).

15
  • 2
    It would be great if I had explanations as to why this request is being disagreed, so I can modify it to address concerns. Mar 11, 2019 at 23:40
  • 5
    I like this, but even better would be per-user blocks, i.e., like I can have ignored tags on SO. Because I really enjoy HNQ but if I could arrange to never see another HNQ from IPS or PCG that would be sweet.
    – davidbak
    Mar 12, 2019 at 0:20
  • 2
    @davidbak True, but the question says they are not implementing that at this time, so I left it out. Mar 12, 2019 at 1:10
  • 1
    Oh, to be sure. But I just wanted to add my support...
    – davidbak
    Mar 12, 2019 at 1:13
  • 10
    Personally I would not want my experience of my HNQ change depending on which site I am currently browsing.
    – Federico
    Mar 12, 2019 at 10:01
  • 10
    I wouldn't want a site removed from the HNQ on a particular site just because some small subset of vocal users don't want it there. If this is done, which I strongly would not like, then the settings around what to exclude and where need to be very well thought out and you probably want input from both sites involved. Mar 12, 2019 at 10:29
  • 3
    @TheLethalCoder It's not "just a small subset of vocal users". Even an SE community manager and a former IPS mod have stated that it's not ideal to have potentially salacious titles in the HNQ on technical/professional sites (which people access at workplaces, mind you). Also, I refrained from proposing a user filter because the team said in the question that they're not planning on it at this time. Mar 12, 2019 at 10:33
  • 13
    @SonictheWizardWerehog I don't see how "Even an SE community manager and former IPS mod" refutes my point, mods and SE staff are obviously a small and vocal subset. However, my main point was that if this is implemented you need something more than a single meta discussion to decide this. Not everyone reads meta yet this change could greatly effect a lot of people that aren't even aware it is possible. Mar 12, 2019 at 10:35
  • 1
    @TheLethalCoder If someone really wants to see HNQ from a particular site, by the way, they can click the heading, which goes to the stackexchange.com homepage's "hot" tab. That lists all questions currently eligible for HNQ. Mar 12, 2019 at 10:36
  • 9
    I think you're missing the main point of my argument so I'll leave the discussion here. That said the same thing applies here, most people probably don't even know you can click that, I know I didn't until I accidentally did once. Mar 12, 2019 at 10:38
  • 6
    This exclusion proposal would enforce bubbles and echo chambers. The variety and 'in other news' aspect is a valuable addition to counter single-mindedness and even boredom. Seems that the 'more relevancy' reason is arbitrary (chances are nil, really) and the 'workplace decency' is a bit bogus as well: HNQ titles are (SE-internal) ads, how much control do we have over inappropriate ads being visible on a workplace computer? The confusing newbies part is worth addressing, but a subset of users seeing 'inappropriate' is probably not. Not even the next step after seeing: 'misclicking'. Mar 14, 2019 at 15:23
  • The best way to solve the problem described here is grouping. Suppose you're on Superuser. Hot network questions. From this site. Windows question, Networking question, Excel question. From other Technology sites. Apple question, Android question, Motor Mechanics question. From other sites. Relationship question, Workplace question, Puzzles question, Movies question. Then on, say, Movies, it'd be From other Culture sites. Sci-fi question, Anime question, Literature question. Etc etc Mar 18, 2019 at 22:32
  • 1
    @user568458 Good idea; you should propose that in a new answer so it can be voted on separately. Mar 18, 2019 at 22:40
  • @SonictheWizardWerehog The question may state that a per user setting won't be implemented. But I think we need to change their mind about that as I guess a lot of users would want that feature.
    – kasperd
    Mar 24, 2019 at 15:05
  • @user568458 I don't think grouping is appropriate as two sites being related isn't an equivalence relation. And it would defeat the purpose of the HNQ which is for users to learn about other sites which may be of their interest. Grouping means users would only learn of sites in groups they already know of, but users may never learn about the existence of other groups of sites.
    – kasperd
    Mar 24, 2019 at 15:08
5

Just for the record and to avoid losing this nice idea:

As user Mathieu Guindon suggested in a comment (with 44 upvotes at the moment):

There's now a history event created the first time a post appears in the HNQ list. - I smell a hat trigger, maybe even a set of badges.

And I agree, it would be nice to get hats in the winter bash, or even badges - based

  • on the number of one's questions that became hot
    • in total,
    • at the same time, or
  • on their duration on the HNQ list, for example.

I think it would be a nice way to acknowledge the contribution of content that is/was interesting for many people.

Of course, this has to be applied carefully to make sure it doesn't backfire. Maybe only award questions that stayed the whole 72h in HNQ. That should give the community/mods enough time to remove unwanted questions and thus only award the 'hot but wanted' fraction.

4
  • 4
    I feel like this would backfire horribly. A lot of controversial questions draw activity that easily puts them into HNQ, but those are often the types of questions sites don't want to promote. If there's an incentive to get a question onto HNQ and the easiest path is to create a controversial or clickbait question, won't that just encourage the behavior sites don't want? Relying on mods to delist won't help because the badge will have already been awarded and the damage done.
    – Troyen
    Mar 31, 2019 at 8:53
  • @Troyen So far, I had the impression that those unwanted questions are the exception. But I agree, a potential awarding system should take this info account. Maybe only award questions that stayed the whole 72h in HNQ. That should give the community/mods enough time to remove unwanted questions...
    – Arsak
    Mar 31, 2019 at 11:16
  • HNQ really isn't a measure of quality by any means, although that doesn't mean we couldn't make a hat or badge. But just keep in mind that HNQ generally just has questions that look interesting, which are often just controversial posts.
    – Cullub
    Apr 1, 2019 at 1:33
  • 1
    No, no, no. A hat based on getting questions into the HNQ or a high HNQ score will almost certainly backfire and cause drama.
    – ChrisF Mod
    May 3, 2019 at 13:57
3

Adjust hotness score to favor sites that didn't appear in hot list for too long.

Studying data provided by query referred here suggests that 20 sites didn't appear in the hot list for about 3 months. Not even once, go figure. For comparison, data for top site shows that about thousand of their questions have been in the list.

This doesn't look fair and I propose that we somehow account for cases like that.

For that, system could check when it was the last time that question from particular site entered hot list and if it's been longer than say, a week (or two weeks, or month etc) it would somehow boost the score for questions from this site. This would improve their chances of getting into the list.

(If it is technically not feasible to do suggested checks and adjustments in real time, this could be scheduled at some regular intervals, say once in a day or in a week etc.)


Related: New Resources for Our Stack Exchange Network

Help smaller communities get some exposure in the Hot Network Questions list...

4
  • To an extent that would help, but some of those 0 HNQ sites simply don't have many candidates. The Windows Phone SE is at 0.1 questions per day, and the stats look worse if you go to the questions tab - 5 asked total for the month of May. Beer is a site that has been struggling along with low activity for years. Ebooks has a few more questions but many unanswered. If you're trying to include some sites like those, that would require a hefty boost (and the hope someone asks a good question).
    – Troyen
    Jun 19, 2019 at 23:20
  • That being said, I do agree with the goal on giving smaller sites more exposure. But some of the sites at the bottom of the list might be beyond any help HNQ can give. You need a question (and ideally an answer) to even have a shot.
    – Troyen
    Jun 19, 2019 at 23:21
  • @Troyen that's a very good point about 0 HNQ sites, I was pondering about the same when preparing this request. Then I thought, without giving them a chance we can't reliably tell whether these "outsider" sites are genuinely of limited interest or it is merely a bad luck, a system mistake. For comparison, look at the sites showing 3-5 occurrences in these query results, these are kind of I feel good about - one can easily see that they were given fair chance and proven their limitations, that's okay...
    – gnat
    Jun 20, 2019 at 5:53
  • ...as for how much boost to give, I don't have strong opinion on that. As of now I am inclined to think that it would better be reasonably limited because I see the purpose of this request not to push the site into list at any price but more like reasonably improving their chances of getting there
    – gnat
    Jun 20, 2019 at 6:00
2

Moderators can remove from the list questions:

that are controversial, start large amounts of debate or arguments or even edit wars.

or

don't set a good example for their sites

Some sites are full of controversial topics, for example, Skeptics. Large amounts of debate happen all the time on controversial topics.

But we are only human and so are moderators. So, how can you ensure that this powerful tool is never misused*? You request that it's used sparingly, but would it be enough?

Also, when a question is removed:

an event will be logged in the post timeline and edit history that indicates when it was removed and by whom

  1. Why not ask moderators to explain the reasons why they used the tool and include it in the event? No canned messages.
  2. Will you add any other measures to prevent misuse?

*Although somewhat uncommon, I've seen moderation tools being misused by users and moderators (for example, closing questions) when they dislike a topic. Said questions were not close-worthy and later on were re-opened without substantial edits.

13
  • 11
    This is exactly why I suggest that sites have a discussion about what this means. When mods make errors it's usually because they're guessing what to do. If a site has guidelines for the sorts of things that shouldn't be in the HNQ, they'll be better able to match their actions with the expectations of users on the site. We don't really want to force mods to explain in detail for every use. If it's confusing or contentious, ask on meta and they can explain their thinking at that time but if everyone agrees, it's not really worth slowing things down.
    – Catija
    Mar 13, 2019 at 13:22
  • 8
    As to informing the asker... that doesn't really make sense. They're not owed time on the HNQ list. It's a bonus. If there's something truly problematic with a question, it should be pretty obvious because of a huge number of comments or answers so it's very likely that a mod may comments about it, anyway.
    – Catija
    Mar 13, 2019 at 13:23
  • 3
    "don't really want to force mods to explain in detail for every use" - Why not? If it's to be used sparingly there is no big waste of time. Btw, by "in detail" I mean the main reasons why it was rejected. No need to type a whole book :P I ll edit to clarify.
    – user
    Mar 13, 2019 at 13:25
  • @Catija Comments and long discussions under a question don't indicate whether it was banned from HNQ. By having it buried in the edit history moderation becomes less transparent. Perhaps notifying the poster is not the best action. Perhaps an automated comment "this question was removed from HNQ" would do (or a less "negative" analog).
    – user
    Mar 13, 2019 at 13:29
  • 18
    Why does the OP need to be notified? We don't even notify people when their questions are closed or deleted - which is much more problematic in my opinion. Those can both be changed. This can't. Why should we tell people about this? Part of your explanation here needs to be what the benefits of this are and all I see here is "mods might abuse this". We tend to trust our mods and then act if they're doing things wrong. We've made them accountable by having a history event for this. Please explain why we need this instead of continuing our trust through accountability?
    – Catija
    Mar 13, 2019 at 13:34
  • 8
    I don't quite get why OP should be notified or why it even should be publicised. HNQ is a fickle thing, and in the face of the new changes its now more transparent then ever. Previously CMs could already exclude a question from HNQ, and it was never publicly mentioned outside of adjacent conversation. I don't think there's value in fretting over and discussing what is and isn't on the HNQ.
    – Magisch
    Mar 13, 2019 at 13:36
  • @Catija you are both right. Notifying the OP about it when we don't notify for closed questions is excessive. I've edited.
    – user
    Mar 13, 2019 at 17:23
  • 2
    @Magisch the fact it wasn't transparent then doesn't mean it's ok to not be transparent now either. As I ve already explained transparency would make it less likely that a mod lightheartedly succumbs to his biases. I've seen it happen numerous times.
    – user
    Mar 13, 2019 at 17:27
  • 1
    @Catija "why we need this [instead of current system]" - because some mods that see eg. in Skeptics a Q titled "is this fact that is not politically correct real?" might ensure that the question is hidden because of their biases without thinking twice. I've seen it happen before with close-votes. Q was reopened since it was blatantly unjustified. Imagine what happens with a practically invisible HNQ-ban.
    – user
    Mar 13, 2019 at 18:08
  • 6
    If that's your explanation, put it in the post, not a comment. :) I'm still not really sure how it negates my point about accountability. Questions aren't necessarily owed time on the HNQ list. If they make a huge amount of work for mods and stir up a lot of drama, that's a reasonable explanation to exclude them. But a mod doesn't need to pre-explain by default. If you're confused about it, ask on meta. If you want to help mods know how to use it - or tell them not to use it at all - work that out on meta. :) Heck, you can even discuss a site policy that needs a comment explanation.
    – Catija
    Mar 13, 2019 at 18:27
  • 4
    But I don't think we need to require these things of all mods on all sites when, for the most part, we don't foresee this as being a contentious issue.
    – Catija
    Mar 13, 2019 at 18:30
  • 6
    @Catija Maybe add a "Recent HNQ knock-outs" to the 10k Tools front page?
    – E.P.
    Mar 14, 2019 at 11:07
  • A post not making hnq is the default behavior. Imagine if a mod unduly removes one from the list? Not much is lost. It's a pretty low risk action a mod is allowed.
    – user803868
    Jun 30, 2020 at 18:02

You must log in to answer this question.

Not the answer you're looking for? Browse other questions tagged .