8

Why I even ask

I'm over in Super User review and see an item pop up in suggested edits that appeared to be vandalism with hate crime type speech in it of prejudice based on skin color with the word "die" really.

It turns out it's a trick/test question so I look at the main post, and it has no history of it ever being vandalized in the past either, but this language still populates here with such words.


What I saw

Typically I'd quote the referenced material, but I'd prefer not for what this question is about specifically (the speech/words used) and who is responsible for that.

Hover over the screenshot below to see the offensive content:

Enter image description here


Question

I guess my question would be to let the community know who specifically is responsible for content or otherwise the design of such an algorithm to use language of this nature in a test review audit?

This has never been edited (e.g. vandalism history) so this could imply someone arbitrarily put this language in this test question/review audit item or into the algorithm of words to use, etc.?

Enter image description here

22
  • 4
    Suggested edit audits are randomly generated, probably using a Markov chain. That text wasn't put in there on purpose by anyone. Although the algorithm could be modified to filter that out, I'm not sure it's worth it -- I've never seen anything offensive before in a suggested edit audit, and I've seen much worse things in real edits.
    – NobodyNada
    Commented Oct 3, 2017 at 5:03
  • Probably doesn't sound too reassuring tho. Commented Oct 3, 2017 at 5:05
  • 3
    FWIW, there have been cases in the past where offensive posts flagged as spam were used as review audits. There's some debate there about whether or not offensive content should be filtered from audits.
    – NobodyNada
    Commented Oct 3, 2017 at 5:09
  • 2
    Rejected suggested edits with the n*-word in them.
    – Glorfindel Mod
    Commented Oct 3, 2017 at 6:01
  • 1
    Which indicates its presumably a useful review audit Commented Oct 3, 2017 at 6:32
  • 4
    @JourneymanGeek I don't think so. Just because evil stuff happens doesn't mean the system should reproduce it. The reviewer who sees this is a person who might be distressed by it, just like any other site visitor. I think it's a bug that the system reproduces or propagates offensive content in any context
    – Zanna
    Commented Oct 3, 2017 at 6:50
  • @Zanna please provide a complete and indisputable list of offensive content to aid the system in detecting and subsequently hiding said content.
    – rene Mod
    Commented Oct 3, 2017 at 7:21
  • @rene At the very least if the system already knows it's offensive (for example, words that cause flagging a comment to delete the comment instantly, content that has been deleted as rude/abusive), then the system shouldn't reproduce it.
    – Zanna
    Commented Oct 3, 2017 at 7:43
  • @Zanna the whole point of audits is to include rude/abusive content based on flags. Excluding all words that were in comments or posts that were flagged will overtime exclude all English words from being used in audits. Although I'm not unsympathetic to the request made here, I fail to see how your suggestion is practical implemented, without the expense of many human resources to decide if single words are rude/offensive or not.
    – rene Mod
    Commented Oct 3, 2017 at 8:05
  • 6
    @rene o.O that's not what I meant, I meant, the system already knows that some words are offensive (because if a comment contains them, flagging it will cause it to auto-destruct). And that's not the whole point of audits I believe - the point of audits is to check the reviewer is actually paying attention - we can do that without offensive content. Nobody awake will approve a suggestion to add the n-word to a post, but there are less unpleasant ways to check someone is awake
    – Zanna
    Commented Oct 3, 2017 at 8:09
  • @Zanna my point stands: provide a list of offensive words. The system can't do that.
    – rene Mod
    Commented Oct 3, 2017 at 8:28
  • 1
    IMO this is not a big deal. The audits are supposed to represent REAL bad edits that you could see on a site and trust me, I have seen and had said to me, a lot worse. The issue is, if some words start getting removed, then someone else will want another removed and then another. As what is offensive is highly subjective, who decides what is offensive? Offence is always taken, never given and if you are botherd by a string on a screen, computer generated at that, that is really your personal issue, not SE's.
    – Mark Kirby
    Commented Oct 3, 2017 at 10:58
  • 1
    A simple solution for those who are eaisily offended, block the words: addons.mozilla.org/en-GB/firefox/addon/foxreplace
    – Mark Kirby
    Commented Oct 3, 2017 at 11:00
  • 3
    @rene Eh, "we can't be 100% perfect so we may as well do nothing at all" isn't a great stance either. To OP's and Zanna's points, we have a blacklist in place on a number of words we don't want appearing in the HNQ titles. I think we should look into reusing that for the suggested edit audit generator. I'll bring it up for discussion internally.
    – Adam Lear StaffMod
    Commented Oct 3, 2017 at 16:03
  • 1
    @Facebook I'm all for common sense, but I give you this: once that list is created we get people that request all kind of words they find offensive should be on that list. I'm not arguing what you find isn't offensive, I do think the solution is non-trivial and only visible for those users that already kind-of opted in to expect offensive content in the first place. I would much more like if the dev team could spend their time on the DAG priorities instead of these kind of corner cases that happens once a while. Though luck, review accordingly and call it day.
    – rene Mod
    Commented Oct 3, 2017 at 16:19

0

You must log in to answer this question.

Browse other questions tagged .