36

Today, there was yet again an anti-semitic question on Stack Exchange, about Nazi Germany's actions against the Jews.

What do I mean by "yet again"? Well, in this Skeptics meta question from February this year, I noted that half of the "helpful" flags on questions on Skeptics.SE were about the holocaust, and approximately half of the remainder were about Jews or Nazi Germany. And the majority of the questions I flagged for holocaust denial were around for more than two hours, ranging from 6 minutes to 8 hours, 20 minutes.

And is it just Skeptics.SE? Nope, here's a quick sample from History.SE. One, two, three, four, five - and that's not going all the way back in my flag history. There have been some problematic content on Politics.SE, but usually it's less blatantly antisemitic, apart from this one.

I don't think it's a problem with diamond moderators. They've tended to do the right thing, and the fact that it happens on more than one site indicates it's probably not an issue with a particular moderator. I also believe that the smoke detector tries to deal with this.

I don't think it's an unsolvable problem - it seems to be the a very small handful of people or groups posting the same hateful garbage again and again every few months, for example compare Were Nazis favorable to zionism? and https://skeptics.stackexchange.com/questions/44757.

For the avoidance of doubt, I want Stack Exchange to act on all bigotry, and I've flagged on a variety of bigotries (people who are bigoted against Jews are also bigoted against other groups), but this is the particular ongoing problem I've faced in my experience as a non-diamond-moderator high-rep user.

  • 9
    In general, there's a practical technical limit here -- six humans who know how to flag as abusive need to actually read those kinds of posts before they can be removed. Smaller sites, it's gonna be harder to remove stuff like that quickly, within the system as designed. How long do these posts tend to stay up, in your experience? – Aza Oct 13 at 9:00
  • 7
    I don't have 10k on History, but the Skeptics posts you linked were handled by the community within hours of posting. What would you like to see? That those posts are handled within minutes, or that they are (semi-)automatically prevented from being posted in the first place? – gerrit Oct 13 at 9:30
  • 10
    What kind of "more"? Examples: They are deleted! On History the kind of primitive AS content as Qs is usually gone in minutes, usually disappears before I can take action. The more problematic thing is when AS posts remain (and some do; Holocaust denial, relativism, primary and secondary antisemitism) Since my comments, flags etc on some of these were denied or resulted in 'helpful' non-action, I'd like you to find examples you identify as such that are still standing. (Sidenote: I'd be almost for these posts remain visible, but with oblivion votes next to them) – LаngLаngС Oct 13 at 9:41
  • 1
    +1 Thank you for bringing this issue to meta! It's something that really needs to be addressed. And I think it's not only a problem with obvious trolls, but also how antisemitism of established users is handled (or not handled). Just as an example: the user complaining on your linked politics.SE post that Jews control the media, that Hitler wasn't an antisemite, and that Jews brought the Holocaust on themselves is not banned. – tim Oct 13 at 10:39
  • 17
    They are almost all deleted already so where's the problem exactly? – Christine H. Richards Oct 13 at 12:02
  • 9
    @tim The problem with politics.SE is that mods allow users with years-long problematic behaviour to continue to crap over the site. See e.g. Why are people who unapologetically and repeatedly post racist drivel not banned for extended periods of time? IIRC they were banned for *a week*(!) after denying the holocaust on that Q. I eventually just deleted my account as the platform has become one of misinformation and toxicity. It's the worst site on SE by far, and has been for many years. It should be killed and (possibly) rebooted. – Martin Tournoij Oct 13 at 12:24
  • 2
    Andrew - what, exactly, are you requesting? Flags need a human at the other end to address the issue. Whether it's a mod to delete the post, or multiple members whose cumulative flags/delete votes will get rid of it as well. The bots do ok, but still not at the level of humans. FWIW even if individuals get banned, it's little effort for them to get another IP address, and new member ID. – JTP - Apologise to Monica Oct 13 at 12:24
  • 1
    Either way, the response on Politics.SE seems unique to just that site; and the responses on other sites has been significantly more appropriate from what I've seen. – Martin Tournoij Oct 13 at 12:25
  • Martin, that's another story, I don't spend enough time at Politics to see that. Those accts should be deleted, but the person will still return. – JTP - Apologise to Monica Oct 13 at 12:25
  • 3
    @ChristineH.Richards The high volume combined with a sometimes slow response time means that skeptics.SE for example has at least hosted Holocaust denial for 75 hours in the last couple of years (just taken from OPs and my flags; the actual number is likely considerably higher). This is getting to a point where it's not acceptable anymore, and where a technical solution might be helpful. – tim Oct 13 at 12:26
  • 7
    @tim 75 hours in 8 years of existence? Less than 0.1% of Skeptics entire existence. I call that a success history. Despite the obvious target one would expect more denialism. – Braiam Oct 13 at 12:45
  • 1
    @JoeTaxpayer No, sorry for the misunderstanding. I didn't mean that one question was up 75 hours, but combined the 26 questions me and OP flagged were up 75 hours (with an average of 3 hours and a maximum of 8 hours per question). – tim Oct 13 at 12:48
  • 4
    @tim - 3-8 hrs makes sense, no? The humans are doing fine. The real question is whether the bot can be better trained. – JTP - Apologise to Monica Oct 13 at 12:54
  • 3
    @Mark I think that screenshotting isn’t the best of ideas, but I’m not fully opposed to it. However, I’d like you to reflect on the fact that they’ve been marked as helpful by moderators, have been deleted, make up a large proportion of my helpful rude and abusive question flags, make up a large number of questions overall, and have been described as a problem by other users, and decide whether it’s plausible to attribute it predominantly to over-sensitivity to questions which were asked in good faith. – Andrew Grimm Oct 14 at 3:11
  • 3
    @AndrewGrimm Well, even granting the points you list, it's still plausible they were asked in good faith, and that needn't imply "over-sensitivity" (to use your phrasing) on behalf of anyone involved; perhaps (I speculate blindly) the mods validated the flags simply because they were awful questions, or because they were bad questions specifically on sensitive issues likely to cause offence. But even if they are bad-faith, that could just as well mean either disingenuously naive questions about sensitive topics, or manifestos about wiping out the Jews, and I'm still curious about which. – Mark Amery Oct 14 at 8:44
68

A voluntary organization consisting of passionate Stack Exchange users have built a bot named SmokeDetector, aimed at these posts that should be red-flagged (spam or abusive).

We're aware of this "holocaust" troll and this very word has been on our watchlist. Once the SmokeDetector bot reports such posts, we'll come and flag them to death quickly.

If you see such posts in the future, you're free to inform us in Charcoal HQ chatroom and we'll handle it.

  • 1
    Thanks for adding "Haavara" to the list today upon my message in the chatroom. – Andrew Grimm Oct 13 at 9:31
  • As a heads up, the post I was mentioning wasn't detected by the bot. – Andrew Grimm Oct 13 at 10:04
  • 10
    @AndrewGrimm Yes. A bot is a bot, and it's been missing borderline posts from time to time. That very post didn't contain any "sensitive" keywords (like holocaust), so it slipped through our checks. We rely on users reporting posts manually in this case, and once we notice them, we take actions accordingly. – iBug says Reinstate Monica Oct 13 at 10:42
  • 1
    @AndrewGrimm We're happy to get additional ways to detect spam/trolling, including what's being discussed here. Additional suggestions for things (e.g. text patterns) the SmokeDetector should be looking for are appreciated. If there are more complex detections you think would be of benefit, you're also welcome to submit PRs to the SmokeDetector GitHub repo.. A significant issue for us is that we only have so many people volunteering their time. If you, or anyone else, wants to help pitch-in and review posts which SmokeDetector reports, that would be quite helpful. – Makyen Oct 13 at 13:19
  • 1
    I don’t think it should be incumbent on ordinary users to check for new keywords to supply to SmokeDetector. In addition, some ordinary users don’t have the rep to see posts they’ve flagged once they’ve been deleted - I didn’t have the rep required to see content in either history or politics, I can’t remember which. Maybe paid staff ought to be doing this. – Andrew Grimm Oct 13 at 21:25
  • @AndrewGrimm don't worry too much. If it gets too noisy, the knobs can be turned to look for words combinations. Also, SD has a dashboard where we can review accuracy of the rules used. – Braiam Oct 14 at 0:46
  • Is there any statistics available on how long these posts live on average? – Trilarion Oct 14 at 6:46
  • 2
    @fredsbend Thanks. Hours feels still a bit long to me for such offensive content. I would very much like to support any initiatives to shorten the time even more. When I flag something for moderation I usually get an action within minutes (on StackOverflow). Of course Skeptics is much, much smaller. – Trilarion Oct 15 at 8:11
  • 2
    @Trilarion the median is slightly above 2 hours when I checked in February and posted about it in the Skeptics meta post I linked to in my question. – Andrew Grimm Oct 16 at 20:07
12

We've already hashed this out on Skeptics Meta, so readers here should check it out there first.

One thing I think we need here is perspective to see what kind of problem may exist that we can solve. These trollish hate things will always be posted. The point of trolling is to be offensive while not breaking the rules. They get posted and will probably always be. That leaves us with repeatedly downvoting, closing, and deleting. All these hate posts get handled in this way quickly. But when we see repeated human action our thoughts should lean toward efficiency and automation. Can we make a computer accurately imitate a human in these instances, without causing adverse effects? Can we save ourselves some time and trouble? I'm all for automating human tasks, but only if data supports the need and it can be designed without actually creating more exceptions and need for human intervention.

The Smoke Detector is a neat tool, helping to speed up human response. I worry however that in these types of things, nuance in language and intent is near impossible to condense into a script, so keeping review human is almost always preferable.

If SE could actually put some data together we can see if we should try to automate in the first place (ie determine if it would reserve human effort). How to automate is another question, and one I'm not convinced we could even do. It depends on what the data shows.

So to be clear, I don't think we're solving an offensive content problem, namely because ot human review processes are cleaning it up pretty quickly. So if there's a problem, it's a bulge in overall human effort. A computer might save us all a lot of time.

  • Did you have a particular one in mind? Link? – JTP - Apologise to Monica Oct 13 at 17:13
  • Yes. Sorry. I see. No need to bring it here. Can this met be closed as a dupe of that one? Prob, not? – JTP - Apologise to Monica Oct 13 at 17:19
  • You yourself suggested going to SE staff with this. It's my understanding that meta.skeptics.SE is definitely not the right way to reach them, so OP opened a question on this meta. It might not be the best way to reach them, but now complaining that OP tried this seems odd. Also, my experience is that the Holocaust questions make up at least your suggested 20% of spam, likely more. Getting actual stats on this from SE staff would definitely be interesting. – tim Oct 13 at 17:59
  • And if some research into this shows that it is indeed the problem we think it is, something like "holds for moderation" seems like a reasonable idea. I'm sure those could also be bypassed, but if it would catch at least some percentage of these questions, that would already be something. – tim Oct 13 at 18:01
  • @tim These 'new ideas' may be interesting, if the problem is shown to be bigger than the examples. Bigger=posts with staying power, despite AS in them. I second fredsbend in that quickly deleted posts cannot be the problem you meant. The problem I see are only those that remain online despite having AS content to varying degrees. Those that get perhaps 1–2 deserved flags and then a mod decides 'nah' & moves on. On H:SE the resident denialist tries his shtick once a week and disappears from view before a feed is updated. How do we tackle thin or thickly veiled AS that is online? – LаngLаngС Oct 13 at 18:09
4

Posting this as an answer, so that it will not be in the edit history of the question, because people were casting doubt on what the questions were like.

This was the most recent question about the treatment of Jews by Nazi Germany.

Screenshot of the recent question

You can see that it is extremely similar to the second most recent one I encountered, and from a while ago. If data from this had been fed to the smoke detector, it would have detected the most recent one, but it did not get detected.

Screenshot of an earlier one

These questions were redundant, because there’s already a well answered question Did Hitler support zionism?

This post was around for 8 hours, 20 minutes.

A post left for long

This post was around for 6 hours, 28 minutes.

Another post left for long

If any moderator wishes to delete answer, go ahead. I do not mind. I hope this is sufficient for people to see that there are bad faith, offensive, questions occurring.

  • 2
    If someone defines Zionism as the repopulation of Israel with Jews, (a definition that may have been used in the 1930s) then it becomes a reasonable question to ask. – Ian Ringrose Oct 16 at 13:44
  • It's quite possible to ask a good-faith question about Nazis and Zionism. The Holocaust was not the only proposed solution for what the Nazis termed "the Jewish Question"; a number of options for expelling Jews from Greater Germany were also proposed. The best-known of these was the Madagascar Plan, but I wouldn't be surprised if "send them back to Israel" was also floated. – Mark Oct 16 at 20:09
  • Being from Germany, I hesitate to write this but ... if the "duplicate" question was problematic, then the dupe-target should also be, and a link to wikipedia shouldn't be a problem either. I guess one problem is the suggestive way how they (mainly the last ones) are phrased, right? Otherwise, I think these are just questions that are ...*ehrm* let's say ... calling them "very low quality" would give them undue credit. And they should be flushed down the toilet - but only for exactly this reason. – Marco13 Oct 16 at 22:28
  • 1
    @Marco13 If you follow both, dupe and target, then you should realise that 'target' does answer the dupe (if taken at face value = assuming good faith) and that the "dupe" is really a primitive pushy and soap boxy piece. And No. They are not really just low quality. The short WP link thing is just raising awareness for a tangential fact (let's say: superseded by actual history? Just reading the Wp page given shows the absurdity of the goal of the question) that's used out of context for conspiracy theories. Most probably a 'constructed trap', poor as it is. Ample reasons for deletion. – LаngLаngС Oct 16 at 23:09
  • @Marco13 The “dupe” maybe phrased in bad faith, but the target for the dupe is ideally a legitimate, good faith question. IF one finds a possible dupe target that is also offensive, then both ought to go. The answers on this page show that abusive posts of the type discussed are being actively curated and moderated, however, so I don’t think you need be overly concerned about abusive targets. – Jon Harper Oct 16 at 23:33
  • My main point was that the questions obviously are as bad as they can get, but not more than that (even though it's clear what the questions are aiming at ... maybe I'm just blunted by the more obvious stuff). However, I'm neither a historian nor a moderator, and certainly don't want to argue about that. You're gonna sort that out, I guess. – Marco13 Oct 16 at 23:51
  • For whatever it's worth, I don't see anything here that would convince me these were posted with any kind of antisemitic motive. I can easily imagine a naive teenager stumbling across claims on the internet about the Holocaust that they find shocking, not knowing how to fact-check them, and ending up at Skeptics. It's not obvious to me that I'm seeing anything here but run-of-the-mill bad questions, of the sort that some sites in the network (like Stack Overflow) have flowing in by the minute. Nor is it obvious to me that any remedy is needed beyond those applied to other bad questions. – Mark Amery Oct 25 at 11:45
  • @Mark not even the “more witnesses for aliens than the holocaust” looks bad-faith to you? – Andrew Grimm Oct 25 at 22:15
  • @AndrewGrimm it looks bad faith. But I'm still a long way from being sure. People genuinely believe stupider things. – Mark Amery Oct 26 at 15:58
3

I think the best way to deal with bigotry in general is to take everything said at face value and potentially answer it if it is interesting enough or downvote it if not, or if a claim is trivially falsifiable. (Note that I find it especially valuable to sincerely answer even bad faith Holocaust denials. I have personally learned a lot by taking seemingly stupid or provocative questions seriously and answering them best I could. It never hurts to have a reality check.1 Since the claims on Nazis and the Holocaust are probably factually resembling each other the answers will be readily available anyway.)

One of the reasons for my opinion is that "bad faith" is difficult to prove and will often result in false positives (see the recent mod ban).

It appears to me that the rules applying to all posts are sufficient to deal with provocative troll posts, including this particular flavor: Bad posts are downvoted, offensive posts are flagged and removed. Does that create an unsustainable burden on the moderators?


1 Have you ever tried to explain to some skeptic in detail the earth's thermodynamic equilibrium mechanism and how within this framework the greenhouse effect works (CO2 vs. H2O absorption bands, temperature gradient in the atmosphere, sun activity and earth orbit effects etc.)?

2

Individual questions seem to be resolved as quickly as much be expected. The only way to speed things up should be paid moderators working across the network.

However, I do think that some of the concern might be that these questions are not made in good faith (i.e., a serious historical question about an unpleasant topic). Bad behavior like that should be grounds for a moderator warning (no idea how you would phrase it). A pattern of such behavior should be grounds for a ban (user or IP) under the CoC as well.

You must log in to answer this question.

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