In a recent thread, a spam answer containing pornography was uploaded (it's now deleted).

When I saw it, I immediately flagged it, but the post remained up for more than half an hour. Then, another user with editing privileges decided to edit the offending post to remove the pornography. They were castigated by another user who said that we should use the "flag" feature and not edit offending posts.

I strongly believe the user who edited the post was not in the wrong, since the post was extremely graphic, and moderation took its time to answer the flag.

According to the rules of this site, though, who was in the right?


3 Answers 3


I think it is fine to edit the post to remove the offending content even though it comes at the cost of the auto-detect algorithm having one less example. Ideally it should auto-delete when it accumulates enough flags, but occasionally they slip through and we'd be the manual exception handlers to take down the post.

To me, it is more important to immediately protect the community and delete/edit the offending content than it is to train the algorithm to do it in future.

(It is my opinion that if there is a present threat, it is more important to deal with the threat with the present resources than to wait and hope that additional resources can be obtained from the current threat to deal with future threats.)

According to the rules of this site, though, who was in the right?

The rules of the site say what can or cannot be posted but do not say what is or isn't the right way to deal with offending posts. It is reasonable to remove the offending post directly, and it is also reasonable to train the algorithm to deal with it. After reviewing the exchange of comments on the (now deleted) post, I do not think that the comments rose to the level of castigation. One user felt that it was more beneficial to delete it as soon as possible, while another user felt it was beneficial to let the algorithm handle it, and they simply made direct and neutral comments to this effect.

Now I do not know how the algorithm works, and it is unlikely they will reveal this information because it will allow others to circumvent it after studying the algorithm, or how or when it takes information from the offending post to learn. If it reads the deleted post as-is, then what we can do is to delete the post or offensive content first, then after it is deleted we can restore the post behind-the-scenes away from the public view, so that it the flags stay on the non-visible deleted post with the offending content as-is for the system to learn how to deal with similar posts in future. Alternatively we need a more robust learning method for the system to also study the post edit history, and we can eat our cake (which is to delete the content) and have it (which is to still let the system learn how how to auto resolve flags even though it has been edited away).

Regarding the issue of how the flags may become misleading towards other moderators if the offending content is deleted, I do not think it is an issue since I feel that mods or high-rep users can access the post history and check that it was indeed offensive.

  • 4
    If the automatic algorithms don't have access to previously reported and deleted offensive content, then there is a major problem with its implementation and not something that should affect how we, as users, deal with offensive content.
    – a20
    Commented Feb 16, 2022 at 14:32
  • 1
    I agree. It is not our responsibility to adapt to the algorithm. Machines should work for us and around what we naturally feel is a reasonable course of action.
    – Flaw Mod
    Commented Feb 19, 2022 at 16:03
  • I'm not sure if "training the algorithm" is really relevant here. There is probably a threshold of offensive flags before a post is auto-deleted. The question is, whether there is less harm done by editing out the offensive content (essentially sparing other users of seeing it) and waiting for the moderators to delete/retract the content, or by trying to leave it as is to gather as many offensive flags as fast as possible. There is an obvious interference between these two strategies, but I don't see immediately which of the two would lead to less people seeing the content.
    – Earthliŋ Mod
    Commented Feb 21, 2022 at 12:21
  • So I guess the question of "total harm" partially depends on how many non-registered users frequent the site (seeing the offensive content, but not flagging it) and how fast the local moderators react to the flag (where if edited out, the offensive content can still be seen in the edit history).
    – Earthliŋ Mod
    Commented Feb 21, 2022 at 12:22

There's actually a third approach that is much more effective. Go to Charcoal HQ and run !!/report <post URL>.

Editing offensive content out of a post makes it hard to continue accumulating the flags it needs to get auto-nuked. On the other hand, unlike spam where editing it out makes it hard without helping anyone, editing out NSFW content can avoid disturbing people or getting them in trouble.

If you report it in CHQ, it goes to the Smoke Detector project, and if you mention what the problem with it is, people who have signed up to review posts and would be okay with seeing NSFW can review it and give it enough flags to be automatically deleted.

You need to be a privileged user in CHQ but you can just paste the link and ask someone else to report it or ask an admin for how to gain privileges.

Finally, an alternative is to edit the image into a spoiler instead of removing it entirely. That way, people will still notice an issue with it and either be prepared or be able to not see it, but can still review it if they are willing to open the spoiler and give it a flag.

  • This is interesting. What exactly is Charcoal HQ? Is it simply a place where a lot of people with mod privileges can have their workloads automated?
    – jogloran
    Commented Feb 16, 2022 at 7:42
  • 2
    Since the edit history is available, why would it be more difficult for it to get autonuked? Surely the automatic algorithms, as well as humans, can look at the history of a post and not just the most up-to-date version (just like they can open a spoiler)? Personally I was in my office when I saw it... fortunately no one but me was around at midnight.
    – a20
    Commented Feb 16, 2022 at 14:29
  • 3
    @a20 When I say auto-nuke, I mean via six people flagging it with Rude/Abusive or Spam. If the image is hidden away in the revision history, it will be harder for people to notice and flag it. If the post goes into review, it will also appear okay to the reviewer unless they choose to look at the history which most don't because usually that doesn't make sense. Editing a note like "NSFW content was removed; check the edit history" would help with that. I don't think editing the post is wrong but it does have drawbacks, that's all I'm saying. Commented Feb 16, 2022 at 16:18

The two existing answers are great, and are both from moderators in the network. As a mod myself on another site, I'm adding my two cents to supplement theirs.

There seem to be certain things missing in the discussion so far. I added a comment replying to your comment under that post, but soon after the post was deleted, so you might not have seen it.

The spam/abuse flags are not only for moderators. I'd go as far as to say it's usually not for moderators. As you note, the mod team takes time to respond, but moderation is not limited to the mod team. The SE network is in large measure community-moderated. You probably didn't notice it, but the commenter who called on people to refrain from sanitizing the post was not a regular member and probably not even a member of the Japanese SE community. (as a sidenote I think castigation is too strong a word to describe anything that took place in regard to that spam post and the spammer) I imagine that user followed a Smoke Detector warning and came to help spam flag spam posts and keep the community free of spam. Those warnings are pushed to various chat rooms. In fact the two spam posts yesterday were picked up by Smoke Detector: See for example this and this from our chat room.

I left a comment response to yours, explaining that if we modify the post on sight, we run the risk of confusing networkwide visitors following Smokey's warnings here. Of course they could always check the edit history, but then again, everybody's a volunteer and people who are not invested in or familiar with our particular community (Japanese SE) may not go that extra mile.

When the number of red flags (spam/abuse) a post amasses reaches a certain threshold, the system will automatically delete it. The current threshold should be at 6. Please refer to this Meta SE post for more information. That's why the site is community-moderated. (would be nice if we could rid our community of the evils of the SE company)

To sum, I don't think you need to sanitize posts like yesterday's. As you've said, they (there were actually two, not one) were pretty graphic. Such posts almost certainly will get picked up by Smokey and pushed to various chat rooms as warnings. It usually takes a matter of minutes. You know what people say: let nature take its course. Well, nature is us.

On the flip site, I have also seen malicious flag rings in operation—that's a term I picked up from people working on and with Smoke Detector: six socks of the same user viciously attacked another user's post, throwing spam flags on it and sinking it within a minute, right in front of my eyes.

You must log in to answer this question.

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