-229

Thanks for your input here and elsewhere on Meta.

Please see the results of this discussion on the update post:
Updating the Hot Network Questions List - now with a bit more network and a little less "hotness"!


Some things happened yesterday that caused a need for us to (quickly) remove a site's eligibility to contribute to the list of hot network questions. For those of you not familiar with what we've come to call the 'HNQ' list, please visit the link; the list of questions shown as 'hot' on stackexchange.com are displayed in the sidebar of most Stack Exchange sites. Here's an idea of what it looks like:

Screen shot of hot network questions in the sidebar

It's a list of titles, with their origins vaguely identifiable by the presence of the 'favicon' element that identifies each site. We're extremely proud of the quality of content that can be found on every single site in our network. The titles, however, when shown out of context, can be problematic. Anything out of context can be problematic.

This isn't a new problem. We've dealt with trigger words in titles that occur naturally through good-faith questions on sites like Parenting, Gaming, Christianity, Islam and more. While we probably deserve some egg on our face for relying on regular expressions as a go-to quick fix, they've been sufficient for ensuring trigger words and overt clickbait was kept out of the list for the most part.

Then we launched a site that tended to have 3-4 hot questions asked each day (by any way one could conceivably measure velocity and good reception). The titles of these questions, out of context, are just too likely to be inappropriate on our professional and academic sites, so we took the site out of the contributing list. Still with me? Thank you, because there's no way to open this discussion without background.

This is one of many times we've talked about problems with hot network questions. This isn't a new problem. What escalated the problem is that we had to take an entire site out of the mix, which isn't ideal, and I'll explain why.

What goals do we have for the hot question list?

  1. They help with discovery. User interviews continue to reveal that folks simply don't often discover just how big of a resource we've become collectively. The list boosts engagement in quite a few ways (we'll admit some of them aren't optimal)

  2. They often remind communities to have discussions about scope. If folks see a clear demand for certain kinds of questions, it's worth revisiting conversations about how valuable finding ways to meet those needs might be.

  3. Hot questions are extremely great fodder for evangelists, the kinds of people that reach thousands of folks that are likely to identify with the way we do Q&A.

  4. Let's face it, they're a jackpot and a curse. A little patch of chaos that grows pumpkins and poison ivy - just usually more pumpkins.

What's broken?

The list of questions is generated through a very expensive query, cached, and refreshed on a set schedule. Giving moderators or even community managers the ability to 'evict' a problematic title while keeping the system in check with what we consider performant and sane are .. mutually-exclusive goals.

Now that we have gotten to the point where we've gone from saying "Let's ignore titles with these words.." to "Let's ignore sites with these titles..", we're seeing a pattern of scale that starts to suck if you think about it protractedly.

What do we need from you?

We want some ideas on ways we can continue to meet our goals of getting the right sites in front of the right people who are likely to appreciate them and participate.

Automatic selection simply will not scale if we continue to push the limits of subjective topic spaces, and we need to continue to do that for the network to grow and remain relevant to the issues folks want to use our software creatively to solve.

We don't expect to solve this problem in one post right away.

Let's say we agree to take out the hot network questions entirely. But, we want something else in its place that continues to pull in some of the benefits, without being such a mindless pain in the neck.

In order to find out what that could look like, we need to consolidate what we want out of it. And that could be "Just turn the darn thing off and be done with it", or "I found my favorite site through that list, you gotta preserve that somehow!" -- all of this input at this stage will be equally good.

Suffice to say, the HNQ (as we know it) is now looking at numbered days, but that number is probably kind of high because it's going to take some work to figure out what voids would be worth filling if we yanked it as we know it now.

So, let's talk about it. And, if this seems like it's a continuation of a conversation, it is, from the moderator chat room. Even though the scope of the discussion is still rough, I wanted to get it out 'in sunlight' early, this isn't something to be settled in an all-mod room.

Thanks for your time, we look forward to finally (at least committing) to settle on something we can all like a bit more than what we've currently got.

And in the spirit of causing tons of people to flood into a single discussion, I'm going to feature this; I just can't help it.

  • 157
    You are aware that by doing this you killed off some 75% of the site's traffic, yes? – Magisch Oct 17 '18 at 17:52
  • 91
    Can you please respond to the relevant meta question on the site in question, with what exactly those "things" were? – Sonic the Anonymous WizHog Oct 17 '18 at 17:52
  • 77
    Unrelated, but perhaps we should allow Meta Stack Exchange to be placed in the HNQ list. This can draw attention to important discussions here. – Sonic the Anonymous WizHog Oct 17 '18 at 18:18
  • 216
    @NicolBolas I had no idea what site it was, and so it made for a very annoying post. – Azor Ahai Oct 17 '18 at 23:44
  • 161
    I fail to see what's so problematic about the two posts in question. Sure, the topic does not seem to be particularly professional, but it's not vulgar, obscene, offensive, or anything like that. Is the problem that two posts in a row are related to sexuality? Are we really so puritan as to get offended by that? – forest Oct 18 '18 at 5:55
  • 227
    To clarify, I absolutely support a profile option to disable HNQ, or even disable questions that are not related to programming (I might not want to get sucked into politics while I am working, period), but to declare an emergency and disable it for everybody because someone on twitter who doesn't even use the site is a puritan who tries to see hidden subtext in everything they read? That's just silly. – forest Oct 18 '18 at 6:20
  • 74
    @E.P. That's not true in my opinion. Just because something is a trope doesn't mean it's sexist. Context matters everywhere, and the title on itself was nothing of the sort. "How do I tell someone that they might be autistic" is not an ableist trope, and suggesting it is is in and of itself quite ableist. "How do I stop people from flirting with me at a school I volunteer at" doesn't even mention either person's gender, and inferring any would be a biased inference and also sexist. I don't understand your line of argument here. – Magisch Oct 18 '18 at 12:55
  • 159
    @TheLethalCoder It is very very sad to see that People on Twitter who do not bother to read the posts they are complaining about are more important for SE now than Mods who care about their communities. That person didn't even mention DMs and yet there's SE employees scrambling to rescue her from the "trolls", who actually did nothing more than replying to her tweet to explain the post respectfully. If SE keeps doing that to become a brand, they will lose the users who made them the valuable product they're. All they will be left with will be the people who like to create fake controversies – NSNoob Oct 18 '18 at 14:25
  • 72
    @TheLethalCoder until we all collectively quit, SE employees will continue to act the same way towards us (as users) as they have in the past. That is to say, condescending and disrespectful. The bright side of seeing SE folks on Twitter is... well it's unfiltered. But the attitudes there reflect how I've felt that SE feels about me as a user for years now. – enderland Oct 18 '18 at 15:45
  • 241
    This is hilarious. During years of asking for changes to and/or mod tools for managing HNQs (e.g. during "Second Iteration of the Stack Exchange Quality Project" which ... just died), nothing happened. For Computer Science, the feature is, plainly spoken, crap, and we don't get any good traffic from it (afaict). Removal of the site from HNQs would have been an attractive option, even. But some Twitter outcry gets things moving in under an hour? Come on. – Raphael Oct 18 '18 at 16:49
  • 237
    I've downvoted this because users have been asking for years to make changes to HNQ and you've done nothing, while a single tweet is enough for you to take action. Way to spit in the face of the SE communities. – Stijn Oct 18 '18 at 19:21
  • 196
    @TimPost I mentioned this in TL (♦ only) and should also say here: One of the breakdowns here is that community members have been asking for change to this feature, and pointing out problems with disturbing titles, for years, but have been consistently met with silence and inaction. But it took one person tweeting and now there's suddenly lots of action. Why did no internal community feedback ever receive this response? That's worth introspecting on as staff, because there's an untenable problem to be found there. – doppelgreener Oct 19 '18 at 12:19
  • 128
    I read the related Twitter thread yesterday. Seeing how unreasonable and toxic the poster was being towards other people trying to calmly explain the situation at hand, I just muted that profile and hoped no one else would fall victim of a disrespectful exchange from there. Seeing that the company chose to listen and drive this entire Meta discussion out of that only makes me sad. – E_net4 is out of comment flags Oct 19 '18 at 13:16
  • 146
    I'm also downvoting this post despite agreeing with the need to work on HNQ because A) I really don't like how this just kicked a whole site without any real notice B) Ignores SE users for Tweets (or at least ignores channels to get feedback within SE) C) because you have again taken action then asked for ideas - ideas don't help when SE's minds are already set – JGreenwell Oct 20 '18 at 20:53
  • 93
    Why in the hell are you people listening to Twitter? Congratulations; you've just given the entire world an enormous lever. All someone has to do is put a post on twitter implying that your website is not inclusive or offends someone's delicate sensibilities, and they can get anything they want here. – Robert Harvey Oct 23 '18 at 16:12

67 Answers 67

2

We have review queues for questions and answers of dubious quality. The same thing would be good for individual SE sites, where people could review hot questions and decide if they are problematic either for their titles or content, as an advertisement for SE.

However, if users were to review this queue on their own site, then they would likely remove from the review queue questions that they felt weren't good quality according to their personal predilections. That's not what we are trying to catch. So I would suggest trialling two things.

  • Firstly, any hot question should be hot-question-flaggable for a set of specified reasons. (Not judged according to whether it is a good question for that site, does it show research etc). HNQ's would get a different flag list.

  • Secondly, the reason that site-specific questions might be problematic taken out of context is something that users of that site are less likely to see. They tend to see the inherent usefulness of the question in context, and in the context of their own site. However, users of other SE sites will be able to judge pretty quickly whether certain question titles, or question content might be problematic to a wider non-specialist audience. Because there are a relatively small number of HNQ's—only a few HNQ's per site—all questions for certain sites that are known to be problematic could enter a review queue to be passed or rejected by high-ish rep users of other sites.

I think that reviewing HNQ questions from different sites would be extremely interesting for most people, as these question tend to be more inherently interesting in the first instance. It would be highly participated in. It would certainly be more interesting than reading hundreds of poor and uninteresting questions that one finds in close-vote queues.

It would also be a good way to introduce a new incentive for getting higher rep on users' own sites. This is because you could raise a new privilege, for say 30,000/40,000 rep users, of being able to review HNQ's for SE-wide sites (not including their own).

If all questions on such sites had to go through ten pass onto HNQ votes, then even if some great questions got eliminated, there would still be an HNQ list of interesting questions for all sites.

2

HNQ is neat, but I can see why some of the questions might be a little odd in the context of, say, Stack Overflow. And why some Stack Overflow questions might seem a little odd to nonprogrammers too.

So I suggest: Don't display all of the HNQ on each site. Instead, use another expensive query to compute site/pair relation strengths based on question migration and/or shared user records and then display the HNQs from the most related ten sites, for some value of ten. And perhaps offer a per-user option to include all, and perhaps include some extra-bland sites.

1

I have been a member of Stack Overflow for nearly 5 years, and find the Hot Network Questions to generally provide great knowledge. Sometimes the knowledge is relevant to my work, and sometimes it is not. On many occasions, I have seen questions and stack sites which do not capture my interest at all. As we all know, this is the nature of the feature as it currently stands.

That being said, I have some suggestions about how we can improve the Hot Network Questions feature without outright removal of the feature and without the need to hide it by default when someone is visiting a stack site.

Regarding Non-users/Not Logged In Users

If a user has found their way to our site, do not make the Hot Network Questions available to them until they log in. There is no guarantee that a question title from many of the stack sites won't have NSFW content, so this would be the single best way to prevent a Hot Network Question from turning potential users off from the site.

Regarding Existing Users

While hiding the Hot Network Questions by default would solve the problem of an out-of-context NSFW question title from appearing upon initial page load, it would still be problematic when the user decides to open the list. Instead, the better alternative would be to put an opt-in and opt-out system in place for users, and, for new users, add it into the account creation process.

From within their profile, users should be presented with the option to opt into the Hot Network Questions feature. After the user begins the opt-in procedure, they are presented with the list of existing stack sites along with a little description about what that site is about, and should be able to make a decision about which sites they might be interested in seeing additional questions from. Upon reaching the conclusion of the opt-in process, that user should also be presented with a disclaimer which states that "some questions may have potentially inappropriate or NSFW titles", and they must choose to accept this as a condition of accessing the feature.

After a user has opted into the feature, then the ability to update their list should be available at any time in the future until they decide to opt-out of the feature. As for when a question does have an inappropriate or NSFW title, perhaps an option could be added to the sidebar which allows the user to add the question to a personal block list.

Regarding the Hot Network Questions Query

It is stated that making changes to the Hot Network Questions Query (The Query) could prove difficult. I believe that we do not need to touch The Query for now. Once a user has selected the stack sites that they would like to view, then use that as a filter for the list returned by the existing query when the user loads the page. If they don't have any sites selected that appear on the current list, then they do not see any Hot Network Questions. If only two selected sites are returned by The Query, then the user only sees two.

Regarding Stack Site Visibility

I have seen some people disagree with hiding or removing the Hot Network Questions sidebar because it would lower a user's awareness to other stack sites, and they have a point. Many of the sites I have visited in the past, including IPS, I did not know of their existence until I saw them in the Hot Network Questions list. Since I generally do not tend to explore links on sites, I would never have learned of their existence without this feature.

To alleviate this concern, I think a new feature should be introduced. I think this feature would be beneficial to all stack sites, and is independent of decisions made to the Hot Network Questions sidebar. This new feature would be something like "Other Stack Sites You May Be Interested In", and it would present the user with a list of other stack sites. The query for this feature could be set up in different ways, but, if set up so that sites with less traffic over a set amount of time are preferred, it would bring smaller sites that are unlikely to show in Hot Network Questions into public awareness.

Regarding Stack Sites Which Do Not Want Hot Network Question Visibility

While reading other answers, some people mentioned that some stacks may not want to be visible on the Hot Network Questions sidebar. If the opt-in system is implemented, then administrators of individual stack sites could be given an option to allow or disallow their site to be opted into.

As an aside, if a site has disallowed users from opting in, this could potentially be worked into The Query to exclude these stack sites from the Hot Network Questions analyses and reduce the burden on The Query.

0

HNQ is not so bad. Just tune a little bit that query.

I think the real problem of the most loud HNQ dislikers is that posting answers to HNQ questions is an easy way to get many rep.

My only problem with the HNQ is that they are mostly not so good questions.

But you could easily solve this problem, simply refresh the list more often.

Simply tune the algorithm, too. For example, having a low upvote/visit ratio should be a strong signature that a question is a bikeshed.

But it was only a single example. What you could most ideally do, is if you would utilize some machine learning to fine-tune the fixed parameters of the HNQ query. You have data scientists there, in the company, and I think it wouldn't be a very big work for them.


For example, what I can see in the image of your example:

  • Latex SE, "Use counters...": good question, might not worth the HNQ upvotes but it is obviously a good one
  • Physics SE, "Conservation of...": very good question (was already renamed)
  • Music SE, I don't know, it is not my field
  • two Workplace SE questions: they are bikesheds
  • Politics SE, "Why is Elizabeth"... Particularly evil bikeshedding, it utilizes political sensitivity to attract visitors and votes. It should be voted down, deleted and sown with salt.
  • Electronics SE, "What kind of socket". Clearly on-topic and interesting, might be disliked by the Electronics SE community on cloudy reasons. Also this has a significant bikeshed taste.
  • Another MusicSE, I don't know
  • Film SE, "Did Neil Armstrong...". Particularly disgusting clickbait, just like the PolSE above. It is also likely off-topic.
  • MathSE, "How to tell if order matters..." Another disgusting, LQ clickbait.
  • ELL SE, "Is it ok to say..." I don't know, I think it is not so bad, but probably doesn't worth a HNQ.

I am sure that you can do it better. Have a clear concept and follow it.

enter image description here

  • 3
    Some people in the industry try to pretend that data science is magical fairy dust that makes all problems go away; the folks who are gullible enough to fall for that are in for a sharp disappointment when they find out it's not. OP explicitly sets performance (on the extremely simple formula currently in use) as a limiting factor - how would adding in a ton of ultra-expensive processing help? – E.P. Oct 24 '18 at 21:54
  • @E.P. No, it is not fairy dust, it is only the future of the IT. The future is not magic dust. Your argument would be strong, if I would see an SE fighting to tune its AI algorithms; but not this is what I can see. I see an SE using exclusively primitive algorithms and elementary school math. I can't name a single, for us visible feature of the software where they would use anything more intelligent. Never. And yes, it is enough in many cases, but not now. About the speed: the algorithm doing the real-time calculation can be much faster, if its fixed parameters aren't hardcoded, as always, – peterh Oct 24 '18 at 22:23
  • @E.P. everything is hardcoded in the SE software, but they are optimized off-line with machine learning. Furthermore, nothing makes for the SE obligatory to use that MSSQL everywhere. You know the SEDE, you have some queries, not even trivial ones, I think you know its limits well. But you know also the search function, which uses elastic search. Using some different than MS-SQL is not unheard even on the SE. – peterh Oct 24 '18 at 22:27
-5

I like HNQ as it is, quite a lot. EXCEPT for the damned spoilers coming from Movies and TV. What I would want to do with this thing is either choose a list of sites that are allowed in my HNQ, or at least a blacklist of ones that are not.

However, I recognize that this does not address the performance issues you're talking about.

So in the future, I'd like to see something that functions in a similar way: highlighting "interesting" content from across the network. It has that "get lost on Wikipedia" feeling. But I want some amount of control over what can come up in it.

As a developer myself, that does sound like a lot to ask for. But I'm asking for it anyway. :)

  • 8
    I wanted to post something along those lines, but what kept me off is the "my HNQ". While I'm all for customizing my HNQ (goodbye, tags I really don't care about), it wouldn't work for logged-out users, which I guess represnt a lot of traffic too, and the base problem for "aw bummer, there's NSFW content on the right, visible by every coworker peeking above my shoulder" issue. – Jenayah Oct 18 '18 at 15:16
  • The performance issues can be addressed client-side if it's simply unfeasible to do on the server (which I find doubtful: the page already filters on user tags; another filter within a selection of 5-30 links is unlikely to add that much processing time); if done correctly the client-side overhead will be extremely minimal. – E.P. Oct 18 '18 at 15:23
-6

Prevent reputation abuse: If we let communities choose posts for the HNQ, maintain an element of automation in the selection process

Questions on the HNQ list get a lot of views and therefore a (usually) disparate number of votes compared to other questions on the site. Giving a site's users (or worse, only a subset of a their users) the ability to decide which of their site's questions are on the list, we invite abuse of the reputation-generating effects of the HNQ.

Keeping the selection process in the hands of an algorithm prevents this from happening. While it may be desirable/necessary to give communities the ability to select questions for inclusion on the HNQ, I believe we must maintain an element of automation for the purpose of mitigating abuse of the reputation HNQ questions inevitably gain.

  • 2
    That...is a problem of how the HNQ (indirectly) works. Solving it by refusing users control of it isn't the solution, fixing the HNQ is. Part of this might be just toning down that heavy reputation-skewing effect instead. This would also have other advantages, like not having users gain close-voting privileges after posting just a single question, something that e.g. IPS is genuinely struggling with. – Christian Rau Oct 18 '18 at 11:43
  • @ChristianRau To clarify I'm not advocating that we deny users control of the HNQ, but specifically the process that gets questions on the list. I'm all for giving us the ability to remove questions. – Twisty Impersonator Oct 18 '18 at 11:44
  • Users don't self-select posts for the HNQ, do we? I thought HNQ selection was already algorithm-based. Or is your answer proposing to keep the status quo? – Lawrence Oct 18 '18 at 11:45
  • Ah, that makes a little more sense. – Christian Rau Oct 18 '18 at 11:46
  • @Lawrence You're correct. But since everything's up for discussion, I want to draw attention to the reputation consequences of being able to pick which questions appear on the HNQ. – Twisty Impersonator Oct 18 '18 at 11:47
  • 4
    @Lawrence You're right. The answer is concerned about a possible future change of the HNQ system, though, and tries to prevent it going in this direction. – Christian Rau Oct 18 '18 at 11:47
  • I don't understand what changes you're actually proposing. Could you be more specific in your post? – E.P. Oct 18 '18 at 11:53
  • @E.P. See edit and let me know if that helps. – Twisty Impersonator Oct 18 '18 at 11:58
  • 1
    Not at all. I still have no idea what you're proposing. Are you proposing that the selection be done by "an algorithm"? Which algorithm? What would be its selection to the algorithm that's currently running? What data would that algorithm use, and what would be the relationship of that data to the site's userbase? – E.P. Oct 18 '18 at 12:02
  • 1
    @E.P. The selection is currently done exclusively by an algorithm. At least one other answer purpose proposes allowing a community to select questions for the HNQ. If this were granted, I'm concerned it would invite abuse for reputation gain, therefore, whatever changes get made, I'm advocating that an algorithm still be used in the selection process. – Twisty Impersonator Oct 18 '18 at 12:09
  • 2
    I visited and re-read the answer you referred because I generally dislike the approach of letting users "feature" questions to HNQ. And yeah, I didn't feel any urge to support it and it didn't look appealing to me. But frankly, what is proposed over there still looks better than current "fully automated" (and totally braindead) algorithm that allows a handful of clueless passers-by from Stack Overflow (armed with bonus rep) force network-wide advertising of any troll-bait that just didn't yet manage to get 5 close votes from smaller site regulars... – gnat Oct 18 '18 at 12:36
  • 1
    ...I recall a wile ago there was an experiment at SE.SE meta that had an effect somewhat similar to proposed featuring in that it pushed many involved questions into HNQ. And while I wasn't entirely happy about that, it still felt so much better compared to current "fully automated", "100% unbiased", "clean algorithmic" way. Guess I would probably prefer automated / algorithmic way if it was not as brainless. But compared to what we have now even manual featuring looks better – gnat Oct 18 '18 at 12:37
  • 1
    To be fair, it's not even that hard to predict the HNQ effect on sites you're active in. I essentially "gamed" my top answer because while reading the question, and seeing it had gotten several votes quickly, I knew it was very likely to show up on HNQ. It wasn't really abuse; but even without sites being able to pick their own, it's usually fairly apparent which type of questions are worth investing time in if the goal is to get easy reputation. – JMac Oct 23 '18 at 14:52
-11

On and on we go about HNQ again.

There are two sites that really need to be banned from HNQ altogether.

1) Interpersonal. The exact reason given in the question is good enough. Mostly the salacious questions end up on HNQ.

2) Skeptics. The site rule at skeptics.stackexchange.com about the level of backing of answers that is required would be overthrown if everybody coming over from HNQ voted on it. I have heard the repeated complaints of the moderators over every time there's a particularly hot question this fight breaks out again.

  • Welcome to me being controversial. This post is negative but I got positive rep anyway. – Joshua Oct 23 '18 at 2:43
  • 1
    "The exact reason given in the question is good enough." - to me this reads more conformist than controversial. With regards to the question, 200 votes down I observe on it (as of now) may possibly somehow influence score of this answer which refers it to justify IPS ban – gnat Oct 24 '18 at 8:42
  • @gnat: This is the more like the tenth time. That one got voted down (probably due to the meta effect) but the others didn't. – Joshua Oct 24 '18 at 14:20

You must log in to answer this question.

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