I've been thinking about flags for a long time, and recently I turned my attention to comment flags.  Out of sheer curiosity, I ran some numbers for the past year of comments flags across the network - to see how many helpful/declined flags we get in each flag type. In 2016, we had **354,440** comments flagged, of those **332,237 were marked helpful**. The breakdown is:

    +-------------------+-----------------+------------------+-------------+-----------+
    |   Comment Flag    | # Helpful Flags | # Declined Flags | Total Flags | % Helpful |
    +-------------------+-----------------+------------------+-------------+-----------+
    | Obsolete          | 158,527         | 2,575            | 161,102     | 98.4      |
    | Too Chatty        | 63,040          | 1,747            | 64,787      | 97.3      |
    | Not Constructive  | 76,314          | 8,832            | 85,146      | 89.63     |
    | Rude or Offensive | 21,180          | 4,835            | 26,015      | 81.41     |
    | Other             | 13,176          | 4,174            | 7,350       | 75.94     |
    +-------------------+-----------------+------------------+-------------+-----------+
    | Total             | 332,237         | 22,163           | 54,400      | 93.75     |
    +-------------------+-----------------+------------------+-------------+-----------+

Both 'Obsolete' and 'Too Chatty' have a higher helpful rate when compared to the other flag types, and [when reading the help center about comment flags](https://meta.stackexchange.com/help/flagging) the guidance is very similar. In my experience, those flags are typically very easy to process, and involve the least amount of effort in deciding if it's accurate.  

For the other 3 comment flag types, you typically need to spend a bit more time deciding if the flag is correct.  But I'd also guess that the decline rates on 'Not Constructive', 'Rude/Offensive' or 'Other' flags is from those flags being used incorrectly (more on this later). 

The reason I've been thinking about comment flags is because in August 2016, [we implemented a new auto-flag for moderators](https://meta.stackexchange.com/a/284058/164200) to point out users who consistently have comments flagged as rude or not-constructive. In looking at how accurate these auto-flags are, I've come to the realization that they stink. It's not because the auto-flag is bad. It's because users are flagging comments with the wrong reasons, moderators are told to process comment flags quickly, and spend very little time thinking about comment flags.  As a result, we have poor comment flaggers, with no correction of the poor flagging, which then leads to noisy auto-flags being thrown at the moderators to handle. 

The current flag dialog is, well, a bit terrible.  The options aren't very clear and there is zero guidance in the dialog.

I've seen comments like:

> You're welcome, glad i could helps.
>
> Updated with the latest spec.
>
> Thank you very much

all flagged as Not Constructive, when really they should have been flagged as too chatty.  

Our options to fix this are:

1. tell the moderators to spend more time on comment flags, and decline stuff flagged wrong to train the flaggers,
2. adjust the rude/not-constructive auto-flag trigger to minimize what is included,
3. or, revamp the comment flag options to try & capture what people are really flagging for

I chose option 3. Searching many Meta sites, you'll find posts about comment flags being declined, or confusion about what flag to use.  Let's just make the dialog, and options a bit clearer. 

I fully agree with minimizing or consolidating the flagging options, as suggested in the question. However, I'm concerned about the category 'Noisy'.  Using 'noisy' to include things that have little value, aren't needed, or are unnecessary, sounds ok, but I'm unsure how many people will understand that definition of the word.  I went in search of another [word or phrase](https://english.stackexchange.com/q/363535/28544) for it and got a lot of really good suggestions.  I also like some of the suggestions by [Robert](https://meta.stackexchange.com/a/253133/164200), but I think they are getting too far [into the weeds](https://en.wiktionary.org/wiki/in_the_weeds).  The comment flag categories, in my opinion, should be easily recognizable to minimize any confusion. 

I think we can consolidate the current options into 3 categories similar to your proposal but maybe alter the 'Noisy' option just slightly.  I recommend the following:

**Rude or Offensive**

Keep this category as it falls in line with the [Be Nice policy](https://meta.stackexchange.com/help/be-nice).  When handling these moderators will be able to [Delete] or [Dismiss] from the flag queue and if necessary, they'd message a user with a suspension if required. 

**Irrelevant or Superfluous** 

I'm aware this is just quibble over a word, but I don't think 'Noisy' is as clear-cut of a name that you'd like.  [Irrelevant](https://www.merriam-webster.com/dictionary/irrelevant) or [Superfluous](https://www.merriam-webster.com/dictionary/superfluous) keep the spirit of your original suggestion by encompassing the no value, not needed, or even obsolete that we're replacing.  I agree with your options available to handle these comments: [Move thread to chat and delete], [Delete] and [Dismiss].
 
**Other**

This should remain in the event a comment doesn't fit into any of the previous categories.  

Besides adjusting the categories, I'd also suggest we add a bit of help to the users inside of the comment dialog box.  This dialog should be updated to be similar to the post dialog - offering some guidance to the users.  A very rough draft could be:

[![comment flag dialog][1]][1]

I think that by offering more guidance to the users and minimizing the number of categories, we might make comment flags a little less stupid.  This would also allow the new auto-flag to solely use 'rude or offensive' flags as the indicator that moderators might need to address a problem user. 

I'm open to leaving more categories if we don't want to reduce this to 3 flag comment types, but I think having some high-level buckets with guidance will help considerably. 


  [1]: https://i.sstatic.net/3iHLG.jpg