Smokedetector: Leave a comment on autoflagged posts

Created on 24 Mar 2017  Â·  7Comments  Â·  Source: Charcoal-SE/SmokeDetector

SmokeDetector could leave a comment on posts which have triggered the autoflagging threshold to further increase visibility, and also to guide the occasional user who wonders what they should do.

A link to a pertinent meta question might be useful to include.

Rough sketch for what to put in the comment:

This post was identified as probable spam. To expedite its removal, please flag as spam (don't vote to close). If the spam diagnosis was incorrect, please reply to this comment.

The last sentence would obviously require some additional logic to set up something to look for replies to these comments.

Background (though nothing more than in this ticket really): http://chat.stackexchange.com/transcript/message/36253120#36253120

Most helpful comment

I'd be against implementing this. It sounds like noise for any post we would autoflag - for the purposes of this, we have a 100% tp rate. When we do have a false positive, this could change it from an internal discussion to an external one.

And even though we had a featured mSE post for a week or so, we all know that not everyone saw this. This is a step in the direction of maintaining PR with 150+ sites. We don't have the resources to handle more than a couple 'why is some robot flagging stuff here' posts, so I don't think we should go provoke people into posting them.

Also, while I've heard people complain about folks handling spam 'wrong', I've never seen conclusive numbers (or any numbers, really) on how much of an issue it really is. If it's a problem, it's one that should be solved in the flag dialog, not by a project like ours.

All 7 comments

Angus suggested replying to any posted comments instead but I'm still partial towards commenting as soon as we can. http://chat.stackexchange.com/transcript/message/36253259#36253259

Alternate idea: rather than posting a comment on all autoflagged posts, watch for comments on the posts, then respond to ones which say stuff along the lines of 'I'm voting to close...' or 'I'm voting to delete this answer as NAA because it's spam' with some sort of automatic reply

As detailed in the chat, that seems less useful to me for a number of reasons.

Replying to comments is more difficult, because we have to watch for them and detect which to rely to. I'd be in favour of leaving a comment as soon as we detect a post, and asking people to come to CHQ if it's incorrect.

This post was identified as probable spam - please flag as spam to destroy it quickly if this is correct. If this is a mistake, please let us know.

I'd be against implementing this. It sounds like noise for any post we would autoflag - for the purposes of this, we have a 100% tp rate. When we do have a false positive, this could change it from an internal discussion to an external one.

And even though we had a featured mSE post for a week or so, we all know that not everyone saw this. This is a step in the direction of maintaining PR with 150+ sites. We don't have the resources to handle more than a couple 'why is some robot flagging stuff here' posts, so I don't think we should go provoke people into posting them.

Also, while I've heard people complain about folks handling spam 'wrong', I've never seen conclusive numbers (or any numbers, really) on how much of an issue it really is. If it's a problem, it's one that should be solved in the flag dialog, not by a project like ours.

Agreed with @Undo. This sounds like it is adding noise to posts that will
be removed shortly anyway.

On Mar 24, 2017 9:06 AM, "Parker Erway" notifications@github.com wrote:

I'd be against implementing this. It sounds like noise for any post we
would autoflag https://metasmoke.erwaysoftware.com/flagging/logs - for
the purposes of this, we have a 100% tp rate. When we do have a false
positive, this could change it from an internal discussion to an external
one.

And even though we had a featured mSE post for a week or so, we all know
that not everyone saw this. This is a step in the direction of maintaining
PR with 150+ sites. We don't have the resources to handle more than a
couple 'why is some robot flagging stuff here' posts, so I don't think we
should go provoke people into posting them.

—
You are receiving this because you are subscribed to this thread.
Reply to this email directly, view it on GitHub
https://github.com/Charcoal-SE/SmokeDetector/issues/617#issuecomment-289031798,
or mute the thread
https://github.com/notifications/unsubscribe-auth/AGKb5yu3eFO3ojxLoAHSuEFErL76UEiEks5ro84AgaJpZM4Mnyp2
.

Closing as discussion on this seems to be dead. Additional comments welcome.

Was this page helpful?
0 / 5 - 0 ratings

Related issues

j-f1 picture j-f1  Â·  3Comments

angussidney picture angussidney  Â·  4Comments

user12986714 picture user12986714  Â·  7Comments

gparyani picture gparyani  Â·  3Comments

SulphurDioxide picture SulphurDioxide  Â·  9Comments