To help moderators decide how consistent is planned action on particular item in queue with existing practices, indicate how similar items were handled in the past.

  1. when item is added to queue, check items that were in queue in the past
  2. from past items, select those similar to current one
    with similar content, tags, authors, flaggers
  3. get statistics for decisions made on selected items in the past
  4. display results to moderator in the form indicating
    how typical was particular decision for similar items in the past


Above suggestion is intended to aid moderator in their investigation but not to substitute their judgement.

  • In particular, it is expected that moderator will not bend decisions deemed correct even when these appear inconsistent with past statistics.
  • 9
    And let us all non moderators join forces in getting WTF at the top of the list...
    – yannis
    Commented Jan 14, 2012 at 18:04
  • 2
    Interesting idea. It could both help newly elected moderators and show moderators in general how consistent their actions are.
    – Adam Lear StaffMod
    Commented Jan 14, 2012 at 18:13
  • 2
    Well, we do already have several facilities available to new mods for learning how to make such decisions. The Moderator Dashboard allows any moderator to review any other moderator's decisions (on the same site). And the Teacher's Lounge chat room can be used to ask questions about moderator decisions in real-time; there are several moderators and SE staff members who are available there during most of the day.
    – user102937
    Commented Jan 14, 2012 at 18:22
  • 2
    @RobertHarvey I personally found digging through other mods' decision logs to be inconvenient, although there's definitely more information there now than there used to be.
    – Adam Lear StaffMod
    Commented Jan 14, 2012 at 18:24
  • Your image editing skills are a wonder; I almost missed it until I noticed the shadows were wrong Commented Jan 14, 2012 at 18:27
  • Shouldn't this be a feature-request? Commented Jan 14, 2012 at 18:41
  • I'm curious. What was the original picture? Russian election results? Commented Jan 14, 2012 at 18:45
  • 5
    The premise is initially flawed. No flag is expected to be treated with a "quick shot". Every flag is expected to be investigated thoroughly. A tool such as this could cause moderators to actually "blow through flags" which is something you really don't want them to do.
    – casperOne
    Commented Jan 14, 2012 at 18:52
  • @GoranJovic they say it is a fragment of TV translation from Russian elections. Can't be results though because total votes exceed 100% :)
    – gnat
    Commented Jan 14, 2012 at 20:46
  • @casperOne "blow through flags" already happen anyway, maybe it's a bit late to be scared of? For that matter, I feel like I can afford risk of blowing through cases like 100%-0-0-0-0-0-0 - especially if that leaves more time for mods to thoroughly investigate items like 58-32-23-19-9-1-0,59
    – gnat
    Commented Jan 15, 2012 at 4:47
  • @casperOne "quick shot" is replaced with "regular checks" in question text
    – gnat
    Commented Jan 15, 2012 at 5:49
  • @gnat Scared? I think not.
    – casperOne
    Commented Jan 15, 2012 at 6:45
  • @gnat that doesn't change anything, no flag is to be treated as "regular"; each flag is to be given a thorough investigation by the moderators. Your point is still moot.
    – casperOne
    Commented Jan 15, 2012 at 6:46
  • 1
    +1 for the image (requires speaking russian to understand) :D Commented Jan 16, 2012 at 6:48
  • @casperOne understood - thanks. "Regular checks" wiped out
    – gnat
    Commented Jan 17, 2012 at 11:31

3 Answers 3


I'm not a mod so I don't feel fully competent to judge this, but even if this is really possible to approach algorithmically: isn't this trying to replace something that a mod needs to develop by themselves - quick, competent and independent judgement of a situation? Are they not there to correct errors made by OPs, flaggers, or the system?

I can totally see the desire to use algorithms (that already work wonders for the site elsewhere) to unburden the mods a bit. But such statistics carry the high risk of leading mods into decisions without them really looking at the situation. (That is not a criticism of our moderators - it's just a human trait applying to every one of us. If you have a repetitive task, and the system allows you to, you get sloppy eventually.)

A mod needs to be able to interpret what they see, and based on that, make a decision. (That implies the occasional mistake.) A new mod needs to learn this, and automated charts don't really help do that.

  • If the moderators have too high a workload, we need more moderators and/or improvements in the UI to make moderating easier. (if the UI were a problem. I have no reason to assume it is. Just making the general point.)

  • If the necessary action for a question can be determined automatically (at least to some degree of certainty), the system should carry it out automatically (like the quality filter already blocks thousands of questions daily) or the community user should start casting flags himself, as a clearly recognizable AI entity acting based on algorithms.

  • 1
    +1 for criticizing the moderators
    – yannis
    Commented Jan 14, 2012 at 18:46
  • 2
    @ZaphodBeeblebrox Your comment, when flagged, would be classified in the "WTF" category (as in, I don't see criticism of the moderators in this answer).
    – casperOne
    Commented Jan 14, 2012 at 19:23
  • 5
    I'm a mod, and I approve of this answer. There are cases where new moderators (especially pro tempore on beta sites who may have little SE experience) may not be aware of all the possible actions, and we should work on this. But providing decision statistics doesn't help here, moderators shouldn't feel that they have to meet a quota of conversions, migrations and so on. What mods need to know is how to come to a decision based on all the data for a particular situation. Commented Jan 14, 2012 at 19:37
  • 2
    @Gilles I'm a mod, and I approve of your comment. =)
    – casperOne
    Commented Jan 14, 2012 at 19:40
  • 2
    A new mod needs to learn this, and automated charts don't really help do that that's a good point. A very good one, thanks. I intended the feature to support not replace quick, competent and independent judgement of a situation. I plan to update question text with this clarification as soon as I figure how to clearly spell it out. Maybe displaying stats after mod makes a decision will do the trick. Or maybe not - I need to think of that.
    – gnat
    Commented Jan 15, 2012 at 5:46

The part where I expect this to fail is in the magical program that selects the similar items that are used for the comparison. Correct me if I'm wrong, but I don't think that something like this is possible with an accuracy that would be actually useful.

If you actually managed to get that software useful enough to let it help the moderator's decisions, why not replace us all by that AI?

There are things that require mod attention that probably could be detected easily, like e.g. vandalism of posts. But many situations are more complicated, the new moderators should think about them, and ask the more experienced mods if they are unsure, they shouldn't rely on some software to make the hard decisions for them.


As others have said, it's going to be impossible to figure out what a mod ought to do in a new situation. The critical word in your process is "similar". This requires understanding the meaning of English text, and applying a judgement call. This requires a human.

However, your proposal to seems to also allow the moderator to 'grade' themselves. As Robert Harvey pointed out in the comments, that's already possible through the moderator dashboard, which provides stats on:

  • Time Last Seen
  • Flags handled
  • Comments deleted
  • Posts deleted
  • Posts closed
  • Edits
  • Mod messages sent
  • Votes cast
  • Posts created
  • Comments created

It also provides some grading in the form of 'hot' and 'supernova' highlighting for unusually poor stats.

It also provides links to the history of my activity and the other moderator's activity, where we can see the last actions a mod has performed in all of these areas. This allows a mod to check their performance, as well as letting checking each other.

You must log in to answer this question.

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