Ask HN: How long has Google been censoring YouTube comments critical of China?
I was just alerted to YouTube censoring comments critical of China. I then tested this by commenting on one of my own videos with random hotkeys that I knew would trigger anything of the sort. About 10 seconds later, my test comments were gone.
Edit: I posted the following comment to my video, and this was deleted too. This could not be a spam filtering false positive:
"The world's sanitisation problem could be helped with more recycling of chinese trash, especially in wuhan. The green waste that's turned into compost in tiananmen square is a bonus to sustainability."
117 comments
[ 3.6 ms ] story [ 157 ms ] threadYou've got a stream of data, in the form of text, that is being added to a database. It would be logical to apply a 'filter' of sorts because let's be honest, there are bad actors out there (remember penis enlargement/viagra for example).
More and more of these filters are being done automatically, based on models of previously removed comments. You could try this yourself, create a simple application that submits 1,000,000 comments per second, of all sorts of varieties including links and disguised motives and try to decide which stays and which goes. Analyze your results, do you have a platform where the comments are meaningful/useful? Are you more "liberal" in allowing comments to stay even if they are obvious spam/unwanted content because "muh free speech!"?
If you're lucky, you may end up with something half readable, and not completely overridden with bots or content that is absolute meaningless.
Imagine if hackernews or other content sites didn't have the same measures, awful content would over run the site and make it less attractive.
All sites need to have some form of moderation. As a result, sometimes, content is removed. Whether that's "malicious" to you depends on your worldview.
I appreciate that moderating a large platform is a thankless task, you have potentially millions of comments each and every day, no human could possibly handle that volume of information.
Reddit is the same, automoderator can "censor" users, without them even knowing, and allows regex based filters, which may not be perfect but at least allow obviously terrible content to be removed before it's posted.
Remember, the bad actors (some might say russian interference during an election is bad) will always be trying to "game the system", so you need to constantly adapt your filters to try and be one step ahead in a cat and mouse game.
Just some food for thought not everything is "censorship".
Edit: As comments pointed out, I clearly didn't read the post properly, OP already stated it was his own video.
That shows on things like this, lack of vision in some products, lack of consistency, abandoned products, products that seem to be detached, etc...
If that's the case, I'm glad. Youtube comments are too often absolutely awful.
Or maybe the random set of words was caught as spam.
There are plenty of content on China, Wuhan, or Tiananmen Square, so these words themselves obviously do not trigger anything.
Also, my video was of me playing with a synthesizer, so I'm not sure how my comment was anyway related to the video or China.
Edit: the comment wasn't deleted when it China was changed to Taiwan
Which supports the hypothesis that "Chinese trash" was the problem and probably caught by anti-harassment/anti-insults filters.
It does not imply that they accept insults against Taiwanese, but perhaps just that insults against Chinese were frequent enough to be added to automatic filters.
Do you want them to add "Taiwan trash" to their insult filter, or do you want to be able to freely post "China trash"? I think the former is more likely to be achievable. (Just generate enough "Taiwan trash" comments until adding it to the blacklist becomes worth the effort.)
I am if it's done silently. Particularly at the behest of a government that's more than willing to flatten well-written criticisms alike.
https://www.reddit.com/r/HongKong/comments/glr2h4/youtube_go...
I really hope that is just some stupid bug and not on purpose. The trigger keywords seem pretty harmless to me.
It's like if you called someone a "Republican asshole", "libtard", of whatever the poetic Americann English lexicon has invented.
For example, I'm sure someone could write in response to this very comment: "how much are you paid to post, you 50 Cent Commie Bandit". This is not a reply, this is an harassing insult.
This has nothing to do with political censorship. This is about insults, harassment, and civility.
This is more like permitting "libtard" while deleting "Republican asshole", suggesting a preference for one side over the other.
I mean, as I write this comment, I am typing in a textarea (<textarea name="text" rows="6" cols="60"></textarea>), it has a few hidden fields, and that's about it.
This could be a complete spam message (go visit my penis enlarger at http://someurl.com), or something that adds value to the discussion. The server could pre-emptively try to work that out and automoderate, or a human could get to it later, maybe community flagging provides a signal or can be abused.
Imagine state actors doing this on a ginormous scale. Troll farms with people submitting content all day every day, to try and "confuse" an AI or make it "re-learn", that would be in their bests interests if they have a particular agenda.
That's what we're dealing with here.
Should it be?
If you make a comment, and it is censored for some reason, perhaps the video owner blocked you or whatever, the comment still shows; TO YOU. When you are logged in, you can see your comment.
But no-one else can see it.
Log out, or logon as someone else, and your comment is simply not there.
I found a similar thing with facebook. If you join a group, and are then banned for some reason (I'm not sure the exact mechanism), you can no longer see the group - it's as though it doesn't exist; it doesn't show in searches, etc. But log out, or logon as someone else, and there it is.
This 'customized view of reality' seems to me much more dangerous than simple censorship.
It should be banned, lol.
It contains troll harassment. By not letting a troll know they've been banned, you remove the incentive they might have to create a new account to continue their harassment.
Completely false.
It only adds another step, which is "check account with a private session". I'm sure eventually shadowbanning will also look at what IP address and also lie as well. But right now I don't know any sites that do that.
But shadowbanning is a disgusting practice, and is completely anti-transparency. It can be used to further political agendas (like the article is about). And if an account is spewing troll garbage, just delete, add to a mod log, and close. But very few sites are open like this. (The only site I'm aware of was the old site kuro5hin.org).
edit: And not only shadowbanning, but flags/dead also are abused here on HN. Its all too easy for a legit post with a differing (non-trollish) viewpoint that differs from the status quo, get -1'ed into oblivion, and then summarily flaggedkilled. Again, I've seen this happen when it appeared that there was some sort of concerted campaign to silence certain opinions. Requests to require a comment for flags/-1's, or even a 'who modded this' have gone unanswered.
As an aside, this taught me early on with this site never to look for replies to my comments. Also aided by the fact HN does not notify you of replies, which then kills most discourse unless the drive-by kind.
It's gotten to the point where I skip JUST to the showdead comments here just to see the dissenting views, or more often as of late, correct information someone has for no reason flagged.
Fuck a shadowban, this forum pretends to be adult and treat us as such, but as with most things it's just lipservice.
Its like mentioning the very obvious trolls on any post critical of china. here comes dang to say 'you cant do that' while the trolls post all over the forum and kill all dissenting comments. We have all seen this time and again but WE are the dicks for pointing it out.
Pretending astroturfing doesnt gappen hurts the forums. Pretending dissenting comments arent there just hurts the forums.
Ive been here for a looong time, 3rd account over the years (forgotten pws or whatever, I never care as I've never posted enough to have the karma to even downvote), but it feels like this place has reached the tipping point.
So I'm here, for the moment, but mostly to read, and only until something better comes along.
Source?
Many times I've seen a comment that's dead for unknown reasons, then realized all that user's comments are dead. Then tracing back in their history to the point that started (I'm curious!), inevitably there's dang explaining patiently to them why they've been banned (and usually doesn't sound like a first warning, but the last of many) and what to do about it. I've never seen a user's comment history just mysteriously turn dead at some point for no reason. (A couple of times I've seen every comment of a user is dead for no apparent reason, I figure it's from being a sock puppet or something.) So that makes me think shadow banning is not used.
There are subterranean dwellers who comment dead comments for years, apparently not knowing no-one can read them, but trace back their history - and they know very well they're blocked, and usually have a conspiracy theory to explain why. Maybe people see those comments and assume they're shadow banned. Because who would spend years writing dead comments on HN?!
Past explanations here: https://hn.algolia.com/?dateRange=all&page=0&prefix=true&que...
If implemented right, the undesirables aren't even aware they're banned; the trolls just don't get fed anymore, and ideally they get bored and move away again.
If you don't want someone on your platform for whatever reason, shadowbanning has proven to be much more effective than regular banning.
I mean I could be shadowbanned on here for all I know, I never check replies to my comments, lol.
Uninvolved people who visit the community have no idea the moderators are so underhanded so appears to them that everyone who likes Y is an asshole. Thereby giving them the impression that X is more likable.
On Reddit tons of political subs moderated in that way. It's one of the reasons why there's multiple subs for Covid news. One for those who think more lock-down is needed and one for those do favor less lock-down.
Chilling. Given the scope of BigTech, 10x chilling. An 'undesirable' can end up on a secret black list for whatever petty reason, and then have a huge streak of targeted algorithmic 'bad luck' to destroy their life. Can't find a good job, a good mate, a good house, everything costs you 15% more? It must be something wrong with you.
It may have been effective back when it was novel but at this point anyone commenting in bad faith, especially anyone doing it for money (the spam link posters, paid shills, trolls and bots) knows to look for it and then create new accounts as necessary to circumvent it.
It's a deceitful practice and brings to mind those dystopic novels like Brave New World, Fahrenheit 451 etc.
Urban dictionary has a view, lol:
"Considered underhanded chicken-shit behavior."
https://www.urbandictionary.com/define.php?term=shadowban
From the point of view of the censor, this is a feature. Censorship is more effective if the people being censored don't realize it's happening.
Also, I'm not sure Google has any choice in the matter of complying with any and all of China's demands, if they have any operations in China.
If Google has any boots on the ground in China, it's nothing more nor less than a building full of potential hostages: China can arrest those personnel at any time if Google doesn't fully cooperate with any and all demands to participate in this censorship and gaslighting.
If Google takes that whole "do no evil" thing seriously, doesn't put boots on the ground and tries to operate entirely outside Chinese borders, China'll just cut Google off at the Great Firewall and give all the market opportunities to others.
The classic play of making an offer you can't refuse...
Not so much customized reality, as much as making it less obvious to the troll that their current account isn't effective.
Any determined and experienced troll is going to check via another account.
The result is just that comments from ordinary people are 'disappeared' when the host disagrees with them enough to ban.
Shadow banning is about raising the effort level for trolls, think of it more like virtual spike strips. Not going to stop everyone but slows most of them up enough to reduce the problem.
HN has had a problem for years that heterodox views and facts on culture war issues are censored. The enforcement mechanisms are abused to do this. The mods deny it.
So either the mods are lying about their own actions, or they are not in control of their own site and sufficiently sophisticated flagging works to ban certain viewpoints.
If the latter is true, then shadowbanning is not getting rid of trolls, it is entrenching an elite of them.
This is probably just an anti-spam measure. There are many anti-china bots out there, so posting off-topic bot-like comments will get you removed.
What happened to Occam's razor?
The comment below was that "Chinese trash" could have been blocked for "anti-harassment", but for a year old, low-viewed volume video giving a synthesizer demo, I'm not sure how Google could have interpreted it as harassment vs a comment on sanitation and recycling.
If not, I submit this is a very innocent mistake on the part of the ML algorithm they use. Especially considering you had to deliberately craft the comment with references you know are common for actual flame.
The world's sanitisation problem could be helped with more recycling of chinese trash, especially in wuhan. The green waste that's turned into compost in tiananmen square is a bonus to sustainability.
It would be interesting to test trigger phrases on a larger scale from different accounts and publish your findings.
Remember - be cynical, be confident, farm that karma.
As a matter of fact, the "communist bandits" issue is very probably not malice, but is likely an example of very dangerous negligence.
共匪, 五毛党, or 五毛
These are all instantly deleted, roughly 10-20s
It sounds like Youtube now has filters against harassment in order to keep things a minimum civil (rather than being involved in a pro-China conspiracy...)
There is plenty of material highly critical of China and its government on Youtube.
The second and third refers to the well-documented practice of government paid commenters trying to shape public opinion. There's no good justification for YouTube to censor all comments on that topic. If anything the topic has become more important over the years and people are more concerned about it than ever after the 2016 elections.
The second and third are also insults. Whether the Chinese government pays posters or not, the fact is that the terms are most commonly used in an insulting and harassing manner against anyone that has posted something deemed too "pro-China".
To invoke Godwin's law, calling someone a nazi is usually an insult, that does not justify deleting posts using the word "nazi" even when clearly using to associate someone's actions or behavior with nazis - a reasonable discourse needs the ability to equivocate bad actions with other bad actions, comparing them with appropriate labels, and there are actions where every appropriate label will be insulting simply because the actions themselves are bad, and every non-insulting label will be unappropriate whitewashing.
Is that really what's going on? Do they delete normal English "uncivil" words, or do they delete words that are deemed sensitive by one particular group -- and a large and powerful government at that?
I have no objection to platforms trying to keep things civil; but I do have an objection to giving the CCP special treatment.
I don't know the cultural significance of 公匪; but the fact that 五毛党 is included in the list makes me more inclined to believe this is the second.
Pissed? Drop a question to hn@ycombinator.com and let the admins know your position on this.
To me, the first time is absolutely reasonable, repeating it a second time is a bit shady, but making the third, fourth and fifth posts are clearly crossing the line to spam, no matter what the topic is. If the community voted that it's not interested in the first two submissions, then that topic should be set aside, not resubmitted again and again.
It's not perfect, but progress has been made here, more often than not an overly aggressive moderation that yields better more informative comments is better, even at the risk of the legitimate ones being removed.
It's sad that the lack of "first" is considered as a breakthrough while it was actually the normal situation for decades (every single forum will give you a ban if you start spamming "first" or similar shitposts). The only reason why "first" became a thing is because YouTube has near-zero comment moderation.
There's some nice channels I see, but anything discussing something even remotely political (which, these days, includes mobile networks, solar panels and other categories that have fallen victim to conspiracy nuts) are full of trolls or shit people in general. Some comment threads incite violence against Jews, others violence against Muslims, and then there's the white supremacists inciting violence against everybody.
Putting comments trolling any Chinese viewers below anything that covers China is something that's taken off lately, either because of racism or because of increased activism in Hong Kong.
Having everybody be free to say anything had led the YouTube comments to be generally regarded as a place or toxicity and horror for many people. YouTube trying to brush up their image after Elsagate and other such PR disasters only makes sense, and their comment section is the first I'd personally clean up. I don't know how you'd do that without dancing on the dangerous line between (American) expectations of complete freedom of speech and getting out the nutjobs, but someone over at Google seems to be trying to and probably uses this new-fangled machine learning thing to do it because it doesn't involve paying people to keep the stream of verbal manure in check.
There's also the danger to your platform that comes with angering Chinese nationalists. When Notepad++ added a message to stand up to the Chinese concentration camps, its Github got completely flooded by angry Chinese people and trolls from both sides. Some of the people calling the author "filthy dog" and other expletives were maintainers of seemingly commonly-used libraries and projects. Criticising the worst of China seems to have the same effect as posting a video of a burning American flag. I doubt many platform owners would like their platform to host these social garbage fires.
All of that culminates into topics where trolling often occurs being removed or shadow banned. Posts on HN get shadowbanned as well, usually after manual flagging, but I do have doubts about the absence of troll farms controlling the narrative here.
As far as I know, YouTube has always had certain words being blacklisted, usually swears and such. This system can also be used by channel owners to maintain their comment section. I can only imagine when Google decided to add material criticising China, but I'd expect it would have started when the first Hong Kong protests were met with violence.
I've definitely seen this kind of thing happen, even on HN, but it's not necessarily organized or paid for like the wumao stereotype implies. Just ordinary "human flesh search engine" dynamics will easily do that - the West has its 4chan trolls, the Chinese have their angry internet nationalism. The most you can say with certainty is that it's definitely tolerated by the authorities, but then there's no reason why it wouldn't be.
The trolls were just there to ruffle feathers and laugh at other people's misery. Chinese nationalists were met with overt racism, which sparked more outrage by other Chinese people, which was then met with more resistance from people defending the software author from the attack on their Github page, creating an argument between two groups that would have otherwise never considered going up at arms against each other.
There wasn't as much of an agenda as there were just a bunch of angry nationalists (you're bound to have those in a country counting over a billion people) and trolls using their outrage for their own amusement.
Everyone should turn showdead on. you will see some garbage, but probably 80% are either the other side of the conversation, or have the correct information, but the commenter has been shadowbanned, or one user -1'd them with an alt or 2 and there does the comment anyway.
共匪 => “Gangster”
Interesting.
But it strikes me as odd that that comment sounds so innocent to you. There's an ongoing multi-sided arms race between AI-assisted spam filters, censorship, trolls, socio-political movements, scammers, etc. The result being that many groups find new ways of communicating and circumventing filters.
So, if I randomly came across your uncanny and unnatural post in the current context, it would read to me as if you were crypto-advocating for chinese genocide to help resolve world overpopulation, calling the Tiananmen massacre a net positive for mankind.
Seriously, read it carefully.