290

This came up recently, where a user had taken the time to look at the link on a post and realized that the post wasn't just off-topic, it was actually half-decently-disguised spam. So he flagged it as spam, and an overworked mod declined it, giving the questioner the benefit of the doubt. And we probably want mods to be giving the questioner the benefit of the doubt.

We also want to get rid of spam ASAP, and make the lives of mods easier. I'd like to suggest a feature the mod in question mused might be useful: An optional textbox where you can add a note saying why you think it's spam, so the mod can see it easily. In this case, the helpful user who went the extra mile could have written "Site linked doesn't relate at all to the question, instead solicits BitCoins" or similar.

One can leave a comment on the question, but:

  1. People sometimes leave...unhelpful...comments on spam questions, and we don't want mods to have to take the time to read through them hoping for something useful

  2. It allows the user flagging the spam to do so and say why without calling attention to themselves publicly


Update: We may well want this for all flags, but in particular I had reason to want it for an NAA flag, so I've posted this other feature request for it.

migrated from meta.stackoverflow.com Mar 18 '15 at 20:48

This question came from our discussion, support, and feature requests site for professional and enthusiast programmers.

  • 57
  • 9
    Additional support for this idea: meta.stackoverflow.com/a/288085/19679 – Brad Larson Mar 16 '15 at 18:28
  • This would help. Although I've had a spam flag declined when the OP, an unregistered uses, had a link at the bottom of their post saying "come to me for SEO" (unrelated to the question). Something like that, I'm not grown-up enough to be able to look at the (later) deleted answer. On the other hand, I still have a custom flag outstanding on a self-promotion/borderline spam. So, it won't be perfect (we make mistakes), but it will be useful. If we get the feature, would the advice be to use it, unless absolutely blatantly 100% obvious? – Bill Woodger Mar 16 '15 at 21:54
  • 26
    I want this for all flags. In particular not-an-answer often requires an additional explanation. – CodesInChaos Mar 17 '15 at 12:47
  • 2
    Isn't this what the "other" flag is for? It already has a text box to describe the reason. – TylerH Mar 17 '15 at 13:34
  • 13
    @TylerH: No, "other" doesn't have the same automation associated with it, nor does it appear in the same places for mods to look at. – T.J. Crowder Mar 17 '15 at 13:36
  • 2
    How can it be made clear that a mod will often not read this text, as the processing of spam flags is mostly automated. E.g. don't include anything other then explaining the spam flag. – Ian Ringrose Mar 17 '15 at 15:04
  • @IanRingrose, as far as I know/have found, mods get it on their dashboard first after 1 flag and only after 6+ flags is it handled automatically. For making it clear that it should only be explaining the reason for the spam flag, perhaps just have placeholder text displaying when the user is presented the input, something like "(Optional) Provide specific details of why this post is spam". – Turnerj Mar 17 '15 at 22:29
  • 2
    @CodesInChaos NAA goes to LQP, so I usually just leave a comment on the post for the reviewers (and presumably the mods will look at it if they handle it manually). So if we're going to add optional explanation, it needs to show up in LQP too. – Jeffrey Bosboom Mar 17 '15 at 22:32
  • 1
    Being able to explain yourself couldn't be a bad thing in any circumstance. – Chris Mar 18 '15 at 3:50
  • 1
    @Ian Ringrose: The system could implement this by way of creating both a spam flag and a custom flag. Custom flags will remain pending on posts even after they are auto-deleted by spam flags. – BoltClock's a Unicorn Mar 18 '15 at 3:56
  • 1
    @Duncan every meta is monitored by SE employees (even the little ones) and they all are appropriate for system-wide support. – corsiKa Mar 18 '15 at 19:57
  • @corsiKa: That's useful to know. – T.J. Crowder Mar 18 '15 at 20:04
  • @BoltClock'saUnicorn I hate you for unleashing a link to tvtropes on unsuspecting readers. There goes 3 hours I'm never getting back. – Ben Lee Mar 19 '15 at 23:20
  • 3
    Being able to add in custom text to a spam flag would be useful for when a user has information promoting their product and their profile but fail to disclose that when they post an answer. This would give the ability to let the moderator handling the flag to check the users profile or other source as needed. – Joe W Sep 21 '15 at 15:21
6

The more general case came up recently: We should always be able to include explanation with moderator flags, even if they're not a custom situation. I think Shog's answer applies here too. If it's not clear that a post is spam, you might want to use a custom flag instead of the simple spam flag. The downside, of course, is that you aren't contributing to 6 flags required for the community to delete spam automatically.

The catch here is that it's hard to know beforehand that your spam flag will be misunderstood. If you were a 100% certain the moderator wouldn't see how this is spam, you would always leave a custom flag. If you were certain the spamminess was obvious, you'd never bother. So how likely are spam flags to be wrongly declined?

All-time on Stack Overflow, the statistics are:

N      Name       
------ ---------- 
221606 Helpful     
 17027 Declined    
  9309 Disputed    
   860 Self Clear  

(For those who, like me, get confused about what a disputed spam flag is, it means the post was rolled back or undeleted and a moderator dismissed the flag. I've also excluded Smokey's autoflags.)

So if every declined spam flag represented a misunderstanding, there would only be a 7% chance adding a message would help. But I'm pretty sure that most of the declined spam flags are the result of flaggers not understanding the purpose of the flag. It seems like the cases where a custom comment on a spam flag would be useful are vanishingly small.


Another question suggested that not providing an explanation wastes moderator time. If this were a problem for them, we'd certainly heard about it. But there's an even better reason to recommend subtle spam should be flagged 'in need of moderator intervention': human task switching. Spam flags are usually quite easy to clear as it's generally random crap unrelated to the site's topic. (For examples, check out the autoflag logs.) Self-promotion spam is a very different beast. Often it's not malicious, but just a little unclear on what the rules are. Most spam users can be deleted without thought, but people who talk about their own stuff should be contacted by a moderator more often than not.

In other words, we talk about spam to mean two very different things:

  1. Mindless attempts to plant links somewhere (anywhere!) on the internet.
  2. Calculated attempts to answer questions with subtle advertisements.

Despite the shared name, these are very different problems with different remediations. So they should be flagged differently.


It occurs to me that we can estimate how many spam flags would benefit from having comments by looking at a sampling of posts that received a declined spam flag:

  1. C++ Program - Linked Data Structure for Sparse Matrices (Image link I guess?)
  2. HTTP ERROR 500 given when adding php script to website login/register feature (Links to own site, but it's not worth spamming yet as it's just a shell. Also, OP disclosed their affiliation.)
  3. been hacked phone has been cloned i dont know how to speak up to your standards fairly new using this kind of technology (Email?)
  4. Create multiselect lookup in salesforce using apex (Deleted since it's just a link to a GitHub repository. Might be spam, I suppose, but only one of the 7(!) flags was of the spam variety.)
  5. Getting Error Code "-504" when trying to update Facebook via Play Store (Of the 8(!) flags, only one is for spam. No link or anything.)
  6. the localhost:4200 is not working (Image? IDK.)
  7. zoomable user interface in python? (No link or any reason for the spam flag.)
  8. Page navigation buttons (Link to blog, supposedly. But the domain is not registered.)
  9. RTSP command for move Head of CCTV using JAVA JDK 1.8 (IP address in the code?)
  10. what can change extension .txt to .HTML files with ruby (I got nothing.)

So of the ten most recent rejected spam flags, maaaaaybe 2 (#4 and #8) are plausibly self-promotion and none are the typical fake designer handbag type. Personally, I'd decline all ten flags. My guess is some of them are mis-clicks and the rest are misunderstanding how flags work here.

  • 9
    How many custom flags are there currently that could have been spam flags if spam flags had an explanation box? – John Dvorak Aug 15 '15 at 13:41
  • Agree with @JohnDvorak . What is the worst, no-one has access to the figures to prove anything with them. – ivan_pozdeev May 10 '18 at 0:47
  • According to meta.stackoverflow.com/questions/367645/… (10k needed) and meta.stackoverflow.com/questions/367645/… , a large portion of spam flags come as a result of automated actions by "SmokeDetector" (even though they seem to be double-checked by people), so the figures as they are are skewed, and the conclusion about the percentage is wrong! – ivan_pozdeev May 10 '18 at 13:32
  • @ivan_pozdeev: This answer is a little out of date. When it was written, I don't think SmokeDetector even existed. – Jon Ericson May 10 '18 at 22:46
  • @JonEricson Well, Shog9 has closed my feature request as a duplicate of this. If this Q&A is outdated, my request is not a duplicate because the changed circumstances have made the justifications here no longer apply. Commented there to this effect and VTRed. – ivan_pozdeev May 11 '18 at 1:17
  • If your previous all-time data had 5% rejection and the current one has 7%, that means if you only take recent data, the percent will be even higher. Comparing your two results, the percentage for the last 2.5 years is 11.4% and ~10.4k posts. – ivan_pozdeev May 11 '18 at 12:37
  • 1
    @ivan_pozdeev: That surprised me too. So I took a small sampling of recent posts that have declined spam flags. It looks like we can blame flagger error not lack of information for all 10. – Jon Ericson May 11 '18 at 22:20
  • 1
    There's another benefit to custom mod flagging... The mods will always see it, even if the post is deleted. If the post gets deleted as something other than spam - say being VLQ - it may need to be spam deleted to actually trigger the spam penalties. – Catija May 11 '18 at 22:24
  • The "two types of spam" model sounds like a workable compromise. The flag note will need to be updated though. – ivan_pozdeev May 11 '18 at 23:52
1

(Disclaimer: this post is obsolete now that @JonEricson updated his to address the issue.)


According to Spam flag declined on link from Founder & CTO, dozens of similar posts (10k needed to see the answer1) and @gparyani's comments in the same Q&A, a large portion of spam flags come from a source other than general SE users – as automated detections by the "SmokeDetector" software, allegedly double-checked by its volunteers. Naturally, they all come without evidence.

So the observed percentage doesn't reflect the UX of SE users, and @JonEricson's proof that these are "rare cases" is wrong.

Since no-one has access to the data, we cannot adjust the figures ourselves accordingly and (dis)prove anything with the results.


1Won't provide a screenshot: the author deleted the post shortly, so I suspect that they felt they disclosed too much. So it makes sense to limit exposure to 10k users who can see it anyway.

  • I could re-run the query with automated flags excluded, but it's unlikely to change the results much. Even if cases like the one you cited were more common now, it would still pay to use a "in need of moderator intervention" flag to explain the problem. In those more subtle cases, it's likely a moderator will need to contact the user anyway. – Jon Ericson May 11 '18 at 1:47
  • 1
    @JonEricson That wouldn't hurt in the least, especially if you mention what time period the data is for -- so one can get an idea if your upper bound of 5% and a few thousand posts is good or bad. I challenged your reasoning regardless of that though. – ivan_pozdeev May 11 '18 at 2:14
  • I'd say gparyani's MSO answer is a honest one. Furthermore, Charcoal is not some company. It's a voluntary organization that helps SE fight off spam. I changed your wording a bit to prevent it from diverting the meaning to the wrong side. – iBug May 11 '18 at 3:14
  • 2
    @ivan_pozdeev: I updated my answer. The new data suggests spam flags are more often rejected now than they were when I wrote that answer. (Smoky autoflags account for ~10k flag on SO.) I also added an argument that comments on spam flags are a bad idea even if this is a common problem. – Jon Ericson May 11 '18 at 6:26
  • 1
    I deleted the answer because other Charcoal participants wanted me to do so. (FYI I'm gparyani on Stack Overflow.) – Sonic the Anonymous Hedgehog May 13 '18 at 22:15
  • @JonEricson It might be useful to re-run the query with only flags that were handled by moderators, excluding those that were automatically deleted upon receiving six flags from normal users. – Sonic the Anonymous Hedgehog Jul 5 '18 at 3:20

You must log in to answer this question.

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