What's the lowdown on chinese tag-team accounts posting questions+answers?

I'm sure everyone has seen them. One account posts a question, the other account posts an answer within a minute. Each account might get used a handful of times before retiring.

I've largely been reluctant to touch them. The questions appeared legitimate the few times I bothered to translate, but they're usually not topics I'm familiar with, so I'm not the best judge of quality there. The times I've checked, I haven't been able to find that the material is copied from somewhere else, but searching for chunks of code rarely works, and the round-trip translation probably makes searching other text impractical.

I see some people marking them as spam, others as not spam. It certainly looks like it's automated, so maybe that alone is enough.

Any thoughts or observations? If you've drawn a red line, where did you draw it? Are there other distinguishing characteristics I'm missing?

[Rik]

I find the behavior odd. Given the short interval between question and response, some form of coordination is very likely, but I can't really see the point. I haven't done anything the times I've seen it (which is only a handful of times).

[DGM]

Given that they have a deep well of questions and answers, I have a feeling that it's reposted content from another forum -- possibly one of the .cn forums that the answers occasionally reference.

[Walter Roberson]

It is spammers copying from other sources.

I don't think I should go into more detail in public.

[Stephen23]

"The questions appeared legitimate the few times I bothered to translate"

They are copies from other forums. Until about one year ago these sock puppets copied entire threads (i.e. questions and answers) verbatim from other forums (which I found via internet searches, and reported to TMW), e.g.:

https://www.mathworks.com/matlabcentral/answers/1939044-matlab-partialdependence-plotpartialdependence

was lifted verbatim from

https://www.ilovematlab.cn/thread-612922-1-1.html

and

https://www.mathworks.com/matlabcentral/answers/1766965-ode45-for

are exact copies of this two-year old thread:

https://www.ilovematlab.cn/thread-601804-1-1.html

and ditto for this thread:

https://www.mathworks.com/matlabcentral/answers/1766950-ode45?s_tid=srchtitle

are copies of this two-year old thread:

https://www.ilovematlab.cn/forum.php?mod=viewthread&action=printable&tid=602726

However over the last year their modus seems to include some AI so that exact searches no longer find any matches.

It is clear that someone has created a lot of sock puppets to achieve this (question: if they just wanted to copy valid answers from another forum, why create so many different sock puppet accounts?).

My guess is that someone related to those forums is creating them, but for what gain... ? What is their motivation?

Keep track of the sockpuppet accounts: some of them are slowly building up reputation.... soon we will welcome new moderators :)

[Image Analyst]

@Stephen23 I also don't know their motivation. Even if they were able to build reputation and then thought they could post spam without being "caught" by the automated spam filters, we'd still catch their spam because we have actual human moderators reading every single post. So what's the point? It's never going to work.

[DGM]

However over the last year their modus seems to include some AI so that exact searches no longer find any matches.

Ah. That probably explains why I couldn't find anything over there. I didn't dig too deep, but a little obfuscation seems to go a long way.

Keep track of the sockpuppet accounts: some of them are slowly building up reputation.... soon we will welcome new moderators :)

Oh. I hadn't seen any of the accounts remain active long enough to build reputation. It a person wanted to do bot grooming, it would certainly help to have an entire forum's worth of material to feed your bots, even if you have no association with the forum. If there are favored accounts that are being allowed to build reputation, then that might indicate motivation -- though it seems like a lot of effort to fabricate a privileged account. Whether it's to abuse or sell, it can't be worth much.

Speaking of account status, maybe the forum needs a new status badge for known spammers. I know it would just encourage account abandonment, but at least they could earn something that's deserved.

[Walter Roberson]

I hadn't seen any of the accounts remain active long enough to build reputation.

It happens.

I am not going to discuss the spam detection processes in public.

[DGM]

If we're dealing with tag-team spam answers that show up in the quarantine, that's an indication that the spam filter missed the parent spam question. We should probably be checking the question to see if it needs deletion as well. I just cleared out a bunch of orphaned questions.

[Walter Roberson]

If you see an Answer show up in quarantine and it was created within a minute of the Question, then report both authors.

[Walter Roberson]

@John D'Errico Timezones -- Rik (European early riser) or I (North American night owl) tend to zap them before you see them.

[DGM]

Another thing I've noticed is that the answered_by_id: and asked_by_id: search queries don't actually work for backtracking. I don't know why that is. It doesn't seem to just be very new accounts that it fails on either.

For example, if you click on the username in this orphaned question, you'll see that they asked two questions. If you click on the provided link, you'll get a failed search, which might lead you to assume that all relevant items have already been deleted. You actually have to go to the profile page and check everything in their feed, many of which may have already been deleted. (I'll delete those tomorrow or so)

I've also noticed some accounts with a handful of reputation points (e.g. level 2), but no upvotes or history that indicates what they did. I thought even deleted content would show up in their feed, but maybe that's not always the case?

[Stephen23]

"maybe that's not always the case?"

It used to work imemdiately... but a few years ago, I also noticed the lag that you describe: I guess TMW made some fundamental change to the database they use. Given that the major internet search engines seem to crawl the forum within a day or so, I have given up on the inbuilt filters and rely on internet search engines (which also lets you filter by date, etc).

"I've also noticed some accounts with a handful of reputation points"

Yes. It is curious that TMW is happy to delete the posts, but is not concerned about this future army of editors.

[DGM]

Next time you see a brand-new answer show up on a brand new spam question that made it through the filter, go to the Browse tab and tell me if you can find the question in the wild. For some reason, it's like this spam content is just invisible. The only way I can find them is via the related links in the sidebar. Using the aforementioned search queries on specific users doesn't work. I have only once come up with a phrase that managed to select a single known question. Bear in mind, there are several hundred, if not thousands of these same reposted threads from last fall too.

I suspect that there's some noise suppression in place to keep the problem from being in everybody's face, even if it's not being deleted. I admit, I wouldn't have known about it if I used the forum normally.

[Walter Roberson]

@DGM please leave the posts for that asked_by example for the moment; I asked Mathworks to check out the indexing problem.

[DGM]

Sure thing.

Answer 1

[Image Analyst]

The Mathwork staff is working it behind the scenes.