123

Consider:

  1. User posts an answer.

  2. Community autoflags it as "low quality (auto)." (This is frequently a false positive.)

  3. User edits answer after 5-minute grace period. (Or perhaps someone else edits the answer.)

  4. Community sees the edit and obediently marks the "low quality (auto)" flag as helpful.

  5. Community applies an automatic downvote to the post, triggered by said edit.

Users are, therefore, penalized for improving their answers. This is wrong, wrong, wrong.

If a user posts an answer which gets autoflagged and someone else edits the answer to fix a typo, it absolutely should not come with a downvote. This makes no sense at all.

Possible solutions, in increasing order of preference:

  • Implement something like this, where the downvote is revoked if the post is later upvoted (this is what happens with "very low quality" flags, but not "low quality (auto)" flags). This is... not ideal, since the downvote shouldn't be there in the first place.

  • Do not apply an autodownvote if the flag was cleared by an edit. This is also not perfect, as it's an ugly special case and requires extra complexity. Which leads to...

  • Do not clear any type of flag on edits. This doesn't really make any sense in the first place. Edits are not necessarily guaranteed to improve a post; indeed, sometimes edits make a post worse. And using edits as a trigger to clear flags can be actively harmful, too, because perhaps the edit did not remove the content that the answer was flagged for, but the auto-flag-clearing prevents a moderator from seeing it.

  • Do not apply automatic downvotes when "low quality (auto)" flags are marked helpful. In theory, it's a good idea, but I often reflexively mark any Community ♦ flag as helpful just because it's less clicks and it doesn't make a difference... except in this case. It's hidden, confusing behavior, and I can't count how many times I've accidentally "downvoted" perfectly valid answers (indeed, it took me quite a while to finally catch on to what was happening). This problem is exacerbated by the fact that this specific flag has a very high false positive rate (on the sites I moderate, at any rate).

  • Both of the above (don't clear flags on edits, and don't autodownvote when flags are cleared as helpful, since neither are at all intuitive).

9
  • 1
    "The system will now revoke the automatic downvote if, after the flag is marked Helpful, the post later gets an upvote. Remember, the automatic downvote essentially exists because folks use this flag in cases where they should be downvoting. As such, it serves an important role... But should still be considered weak compared to actually downvoting the post in cases where there is conflicting information..." (Community downvotes posts that were just fixed)
    – gnat
    Commented Feb 29, 2016 at 12:42
  • 7
    @gnat I'm talking about the "low quality (auto)" automatic system flag (as I mentioned repeatedly in my question), not the "very low quality" flag.
    – Doorknob
    Commented Feb 29, 2016 at 13:19
  • 1
    I think same reasoning should apply to auto-flags: upvote to post forces Community user revoke downvote. If this doesn't happen, I would say that's a design mistake and linked post explains why the other way is correct
    – gnat
    Commented Feb 29, 2016 at 13:57
  • 1
    Wait, auto VLQ flags are still a thing? I haven't seen one in.. over a year?
    – ɥʇǝS
    Commented Mar 3, 2016 at 20:29
  • @ɥʇǝS We get them nearly daily on PPCG (last one was Low answer quality score [80] just yesterday, in fact).
    – Doorknob
    Commented Mar 3, 2016 at 20:36
  • @Doorknob I'm looking at this request - are you seeing this more on PPCG or Puzzling? Asking because you're a mod on both.
    – Taryn
    Commented Mar 3, 2016 at 21:08
  • 1
    @bluefeet You mean the undesirable behavior? It happens on both. Or do you mean auto low quality flags in general? They're very common on both sites as well (probably just because short answers are common due to how Puzzling and PPCG work).
    – Doorknob
    Commented Mar 3, 2016 at 21:45
  • @Doorknob Both. I'm going to pull some numbers on it so I wanted to know where to look initially. :)
    – Taryn
    Commented Mar 3, 2016 at 21:46
  • 12
    Holy ****. Is this why I see downvotes on crappy answers almost immediately after I edit them? I kept thinking it might be coincidence, but it's happened so often...
    – muru
    Commented Mar 4, 2016 at 14:07

1 Answer 1

32

Update: August 15, 2019

The system no longer automatically downvotes posts as a result of automatic flags anymore.

Update: August 31, 2017

Internally, we discussed several ways to implement this, but in the end we made them work the same as normal VLQ flags. That means, if the flag was marked helpful, but later receives an upvote the automatic downvote is revoked. This is available as of the latest build. Thanks to the devs for implementing this change, and making the behavior the same across both VLQ flag types.

I agree that this probably shouldn't happen, if someone edits a post to improve it, punishing the OP after a post was edited isn't the best experience.

I was curious how often this happens so I pulled some numbers for the last 365 days. Stack Overflow is excluded because it doesn't get low quality (auto) flags, the data below is for all non-meta sites that had at least one helpful flag:

Total Flags Total Helpful Total Helpful By Edit Pct Validated By Edit Site Name
5252 4582 441 9.62 Super User
1678 1229 91 7.40 Server Fault
18 18 0 0.00 Stack Apps
898 721 42 5.83 Cooking
855 801 52 6.49 Home Improvement
407 369 34 9.21 Game Developers
5331 4955 377 7.61 Gaming
923 884 56 6.33 GIS
349 280 36 12.86 Mathematics
363 348 25 7.18 Photography
701 658 44 6.69 Statistical Analysis
1221 1195 91 7.62 Web Apps
320 287 23 8.01 Webmasters
4548 4086 205 5.02 Apple
25 24 2 8.33 Theoretical Computer Science
3541 2664 407 15.28 English Language and Usage
262 258 5 1.94 Personal Finance and Money
172 157 16 10.19 Role-playing Games
879 796 135 16.96 TeX - LaTeX
393 314 40 12.74 Ubuntu
2101 1735 444 25.59 Unix and Linux
1227 1144 111 9.70 WordPress
218 211 17 8.06 Bicycles
136 135 9 6.67 Programmers
4766 4595 272 5.92 Android Enthusiasts
125 109 9 8.26 Board and Card Games
1287 1197 81 6.77 Physics
50 37 3 8.11 Homebrew
545 532 34 6.39 IT Security
96 89 5 5.62 Writers
1244 1218 78 6.40 Electronics and Robotics
799 640 43 6.72 Graphic Design
719 645 77 11.94 Database Administrators
789 659 89 13.51 Science Fiction
1 1 0 0.00 Area 51 Discussions
341 331 28 8.46 Code Review
229 166 78 46.99 Code Golf
76 68 8 11.76 Quantitative Finance
28 26 0 0.00 Project Management
197 197 3 1.52 Skeptics
849 735 75 10.20 Drupal Answers
103 101 10 9.90 Fitness and Nutrition
426 385 26 6.75 Motor Vehicle Maintenance and Repair
173 173 3 1.73 Parenting
1063 959 62 6.47 SharePoint
240 217 9 4.15 Musical Practice and Performance
118 112 17 15.18 Software Quality Assurance and Testing
148 128 22 17.19 Jewish Life and Learning
74 73 2 2.74 Astronomy
163 126 30 23.81 German Language and Usage
108 63 11 17.46 Japanese Language and Usage
136 127 12 9.45 Gardening and Landscaping
97 96 9 9.38 Philosophy
48 46 4 8.70 Personal Productivity
633 584 28 4.79 Travel
88 86 6 6.98 Cryptography
107 107 14 13.08 Signal Processing
136 100 22 22.00 French Language and Usage
308 308 12 3.90 Christianity
255 249 13 5.22 Bitcoin
54 53 3 5.66 Linguistics
120 120 2 1.67 Biblical Hermeneutics
32 31 4 12.90 Economics
166 159 7 4.40 History
110 95 8 8.42 LEGO®
151 142 22 15.49 Spanish Language and Usage
42 38 2 5.26 Computational Science
909 889 66 7.42 Movies
235 123 16 13.01 Chinese Language and Usage
222 220 7 3.18 Biology
40 38 4 10.53 Poker
165 149 23 15.44 Mathematica
55 54 2 3.70 Cognitive Sciences
67 59 3 5.08 The Great Outdoors
32 32 0 0.00 Martial Arts
136 135 5 3.70 Sports
269 255 15 5.88 Academia
132 122 13 10.66 Computer Science
121 119 7 5.88 The Workplace
551 524 12 2.29 Windows Phone
547 538 34 6.32 Chemistry
132 108 19 17.59 Chess
520 422 23 5.45 Raspberry Pi
104 77 5 6.49 Russian Language and Usage
363 361 11 3.05 Islam
770 648 56 8.64 Salesforce
55 49 0 0.00 Patents
236 231 15 6.49 User Experience
28 28 4 14.29 Genealogy and Family History
41 41 0 0.00 Robotics
68 63 3 4.76 ExpressionEngine
48 46 2 4.35 Politics
97 86 5 5.81 Audio-Video Production
731 659 81 12.29 Anime and Manga
1255 1067 80 7.50 Magento
793 677 81 11.96 English Language Learners
22 22 0 0.00 Sustainable Living
17 17 1 5.88 Tridion Stack Exchange
51 41 6 14.63 Reverse Engineering Stack Exchange
262 210 23 10.95 Arduino Stack Exchange
214 198 6 3.03 Network Engineering Stack Exchange
42 40 4 10.00 Open Data Stack Exchange
22 22 2 9.09 Freelancing Stack Exchange
384 353 30 8.50 Blender Stack Exchange
193 178 8 4.49 MathOverflow
91 90 4 4.44 Space Exploration Stack Exchange
112 104 5 4.81 Sound Design Stack Exchange
124 117 5 4.27 Tor Stack Exchange
139 138 2 1.45 Pets Stack Exchange
23 19 0 0.00 Amateur Radio Stack Exchange
30 26 3 11.54 Italian Language Stack Exchange
872 750 102 13.60 Stack Overflow em Português
197 181 9 4.97 Aviation Stack Exchange
32 31 2 6.45 Ebooks Stack Exchange
43 40 0 0.00 Beer Stack Exchange
189 186 15 8.06 Software Recommendations Stack Exchange
68 65 2 3.08 Expatriates Stack Exchange
20 16 0 0.00 Mathematics Educators Stack Exchange
49 47 1 2.13 Earth Science Stack Exchange
127 119 13 10.92 Joomla Stack Exchange
41 40 4 10.00 Data Science Stack Exchange
435 366 107 29.23 Puzzling Stack Exchange
54 43 5 11.63 Craft CMS Stack Exchange
59 57 11 19.30 Buddhism Stack Exchange
206 204 5 2.45 Hinduism Stack Exchange
6 6 0 0.00 Moderators Stack Exchange
31 28 2 7.14 Startups Stack Exchange
207 201 15 7.46 Worldbuilding Stack Exchange
256 165 24 14.55 スタック・オーバーフロー
47 37 10 27.03 Emacs Stack Exchange
20 19 1 5.26 History of Science and Mathematics Stack Exchange
239 222 23 10.36 Lifehacks Stack Exchange
55 52 5 9.62 Engineering Stack Exchange
27 22 4 18.18 Coffee Stack Exchange
25 20 4 20.00 Vi and Vim Stack Exchange
62 55 8 14.55 Music Fans Stack Exchange
32 24 3 12.50 Woodworking Stack Exchange
66 50 4 8.00 CiviCRM Stack Exchange
36 36 0 0.00 Health Stack Exchange
2659 1844 602 32.65 Stack Overflow на русском
453 249 139 55.82 Русский язык
11 10 0 0.00 Mythology Stack Exchange
32 31 0 0.00 Law Stack Exchange
6 6 0 0.00 Open Source Stack Exchange
165 154 22 14.29 elementary OS Stack Exchange
29 23 7 30.43 Portuguese Language Stack Exchange
3 2 0 0.00 Computer Graphics Stack Exchange
10 10 3 30.00 Hardware Recommendations Stack Exchange
47 44 13 29.55 Stack Overflow en español

Besides the Russian Language site, Code Golf has one of the highest percentages so I understand why you noticed this.

I'm answering because I'm looking at it and will try to figure out the best way this should be implemented. To be clear, any change will only impact the auto flag, it will not impact VLQ flags placed by users because those have been seen by a human and that means a higher signal of a problem.

3
  • 15
    it will not impact VLQ flags placed by users because those have been seen by a human and that means a higher signal of a problem. That sounds like the downvote should be added immediately, like spam flags -- not like the user should be punished when they fix that problem.
    – user154510
    Commented Mar 16, 2016 at 4:44
  • Stack Overflow на русском has the biggest impact on that, seems legit...
    – EKons
    Commented Aug 3, 2016 at 8:03
  • 2
    @ΈρικΚωνσταντόπουλος Perhaps Cyrillic isn't being treated as text by LQ(a) bot.
    – wizzwizz4
    Commented Dec 25, 2016 at 19:32

You must log in to answer this question.

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