YouTube Warns That, Thanks To Covid-19, It's Handing Over More Content Moderation To The Machines And They Might Suck
from the be-forewarned dept
Content moderation at scale is impossible to do well in the best of times, but the solutions that seem to at least keep it from devolving into a total mess almost always use a combination of humans and technology working together. But what do you do when the humans are sick, self-isolating, quarantined, etc? While I imagine some may be able to work from home, it’s a difficult time to expect anyone to be at full productivity. So YouTube has made it clear that it’s turning over more content moderation decisions to the machines knowing full well that some of those decisions are going to be bad:
Our Community Guidelines enforcement today is based on a combination of people and technology: Machine learning helps detect potentially harmful content and then sends it to human reviewers for assessment. As a result of the new measures we?re taking, we will temporarily start relying more on technology to help with some of the work normally done by reviewers. This means automated systems will start removing some content without human review, so we can continue to act quickly to remove violative content and protect our ecosystem, while we have workplace protections in place.
As we do this, users and creators may see increased video removals, including some videos that may not violate policies. We won?t issue strikes on this content except in cases where we have high confidence that it?s violative. If creators think that their content was removed in error, they can appeal the decision and our teams will take a look. However, note that our workforce precautions will also result in delayed appeal reviews. We?ll also be more cautious about what content gets promoted, including livestreams. In some cases, unreviewed content may not be available via search, on the homepage, or in recommendations.
And, of course, this is absolutely the right choice to make — indeed, it’s the only choice to make given the circumstances. But it’s yet another reminder of how impossible and fragile the system is when people demand that humans review everything being posted to social media. Either way, don’t be surprised to hear many more stories of bad content moderation decisions not just on YouTube but elsewhere in the coming weeks. Of course, I still imagine people will scream and yell and take it personally, but at least recognize that some of the issue may be that the humans are all kinda preoccupied with more important things right now.
Filed Under: ai, content moderation, content moderation at scale, humans, youtube
Companies: google, youtube
Comments on “YouTube Warns That, Thanks To Covid-19, It's Handing Over More Content Moderation To The Machines And They Might Suck”
A reviewer can watch a video from literally anywhere there’s an internet connection. I’m surprised they haven’t instituted a work-at-home policy previously for most, if not all, of their reviewers given the current state of the bay area (as evidenced by the SnapCrap app).
Or, they could stop moderating and let people express their first amendment-protected views without being screened. I’d rather know that a YouTube producer is a racist or sexist than to have a video showing it being moderated out of view.
Re: Re:
A reviewer can watch a video from literally anywhere there’s an internet connection.
Not if you want to protect privacy.
Or, they could stop moderating and let people express their first amendment-protected views without being screened.
This is a nonsense position. Because you’re saying that a platform should allow spam, abuse, and nonsense. And that’s a bad idea.
Re: Re:
Oh, no, they’re not saying a platform should allow that content. They’re saying a platform should be forced into hosting that content by law.
Re: Re: Re: Re:
He actually said no such thing. He said that maybe they should ease up on the censorship altogether during the crisis instead of handing it over to machines.
Nowhere did he say anything about forcing them to do it by law.
But it was an A+ attempt on your part at a strawman. I’ll give you that.
Re: Re: Re:2 Re:
Speaking of strawmen…
First, to reiterate something that has been said many times, what YouTube is doing is not censorship (“you can’t say that anywhere”); it’s moderation (“you can’t say that here” or “we won’t host that here”).
Second, the AC didn’t say that YouTube should “ease up” on moderation but instead, “stop moderating and let people express their first amendment-protected views without being screened.” If they were to “ease up” on it, that would merely be a reduction of moderation, not a cessation (even if only temporarily) of it. Plus, the AC never said or even implied that the stop would be temporary, though that’s more of a quibble than a major issue.
Third, while I grant that inserting that “by law” bit isn’t necessarily entirely supported by what the AC did say (though I presume that that was inferred from the “first amendment-protected” bit), the “forced” bit is at least a reasonable interpretation, since the idea would be that they would remove no content at all (except possibly those that are DMCA claimed), conceivably because of the lack of manpower to make moderation work as effective as before (though, again, the AC’s arguments on this sound like they have nothing to do with the current pandemic).
Re: Re: Re:
Not to mention… doesn’t Youtube outsource a lot of its manual moderation to outside contractors? They wouldn’t have control over those contractors’ work conditions. A lot of contracting companies don’t like work from home because they can’t exercise the level of control they want like limited bathroom breaks and continuous activity monitoring.
Re: Re: Re: Re:
Even if it weren’t a privacy or security thing, I dare say that schools and workplaces being shut will increase activity on the site beyond what those workers could realistically cope with immediately – even if there were no transition problems moving to home only workers.
Re: Re: Re:
What’s the privacy angle here? For something like Facebook, sure, but isn’t the stuff on Youtube almost entirely public?
Re: Re: Re: Re:
I have the same question. Most companies that offer work-from-home as an option provide a VPN to use for said work. Those same employers typically provide the machines their employees will use. Apart from the risk of the employee working from Starbucks and leaving their unlocked laptop sitting at a table (that is no longer available) while they go to the bathroom, I’m not sure how privacy plays into this debate.
Re: Re: Re:2 Re:
"Most companies that offer work-from-home as an option provide a VPN to use for said work."
Most such companies are currently having capacity stretched by workers who would never normally work from home now being required to do so.
Remember, the discussion here is not about business as normal, it’s about a very extreme and unusual circumstance that introduces a lot of problems that would never appear otherwise.
"Apart from the risk of the employee working from Starbucks"
I’m not sure about the current situation in the US, but where I am the fact they were sitting in a Starbucks in the first place would be a bigger issue than what they did there.
Re: Re: Re:3 Re:
So what’s the privacy issue? Are you saying some employees wouldn’t be able to get on the VPN, and so would continue to do their jobs without it?
Re: Re: Re: Re:
"isn’t the stuff on Youtube almost entirely public?"
Not all of it, and I’m sure they have the ability to restrict things to mod accounts while being checked.
Re:
The First Amendment stops the government from interfering with your speech. It says nothing about private platforms being obligated to host your speech — or about you being entitled to an audience.
Besides, we already have platforms that don’t ban/punish bigoted speech in general. We call them “4chan and its various knockoffs”.
Re: Re: Re:
Nor did the comment you replied to.
Re: Re: Re: Re:
Quoting that comment:
The implication here is that YouTube is somehow “censoring” speech via moderation — that YouTube would be a better platform if it were obligated by law to host all (legally protected) speech without moderating it. But the First Amendment lacks a “neutrality” mandate for private entities; YouTube has no legal, moral, or ethical obligation to host any kind of speech.
Re: Re: Re:2 Re:
And while the law may be silent, there’s fairly strong ethical and moral obligations to not promote abusive and Nazi content.
Re: Re: Re: Re:
Except for, you know, that part where it quite clearly did.
Re: Re: Re:
Hey, great strawman! But no one here said anything about the government forcing YouTube to to host speech.
The only suggestion was that maybe YouTube should choose to voluntarily ease up on the censorship during the crisis instead of making it worse by handing it over to machines that will censor everything, both the bad and the good.
Re: Re: Re: Re:
Look: this isn’t censorship, and the AC was advocated for not moderating at all, not just easing up on it. Furthermore, the point was that the AC said that the speech being moderated is protected by the First Amendment, which doesn’t actually matter regarding YouTube’s moderating decisions.
Re: Re: Re:2 Re:
" AC said that the speech being moderated is protected by the First Amendment, which doesn’t actually matter regarding YouTube’s moderating decisions"
Even more to the point – YouTube also has first amendment rights, so of which include the right to free association. YouTube are saying they don’t wish to associate with the people they are blocking. To argue they should not is to either argue that YouTube be forced to associate against their will – a first amendment violation – or to argue that YouTube is an agent of the government, which is not in evidence.
Re: Re:
"Or, they could stop moderating and let people express their first amendment-protected views without being screened"
They could, but they’d lose most of their advertisers.
Re: Re: Re:
Advertisers, and the large section of content creators and users who would rather not have to deal with the kind of deplorable scum that would quickly take advantage of that sort of hands-off policy.
Re: Re: Re: Re:
Exactly. The kind of neo nazi scum who personify platforms like Gab would move in, as would people showing ISIS recruiting videos and so on. The families and advertisers who make up most of the traffic and income would move out, then most content creators would follow suit since they lost all their income and followers.
YouTube make a lot of mistakes, some forced, some not, but to pretend everything will be hunky dory if they did nothing is hopelessly naive.
Re: Re:
Or, they could stop moderating and let people express their first amendment-protected views without being screened.
Yes, I’m sure bigots of all shades would absolutely love it if platforms stopped giving them the boot and showing them the door.
Once more for the slow I guess…
You have no first amendment rights on a private platform. Just because the government can’t punish you for saying something does not in any way mean anyone else is obligated to provide a platform for you to speak from.
Re: Re:
All the first amendment guaranteed is that the government will not stop you self publishing your speech at your own expense. Even if the only way for you to get your speech out is to buy tour own servers and services, and run your own blog, you first amendment rights are still being respected.
Getting what you asked for, but not what you wanted
While I imagine that this will take an already bad system and just make it worse, it strikes me that it will also nicely highlight how absurd and unreasonable the ‘you’re not doing enough to keep bad stuff down!’ crowd’s demands are in showing in a small way just how well that works in practice.
There have already been plenty of stories about how flawed the moderation they already have is, making it even more automated, which is basically the only way for a platform like youtube to step up moderation if pressed is only going to highlight just how flawed that is.
On the plus side…
We won’t issue strikes on this content except in cases where we have high confidence that it’s violative.
… The increased automation is a problem but hopefully a temporary one, but this is something that really should stick around. The strikes system is absurdly broken and easy to game, and it really should not be as easy as it is to kill off an entire channel, so any changes that makes that less easy are ones that should be here to stay and not likewise temporary.
Re: Getting what you asked for, but not what you wanted
An entire channel doesn’t mean much really as it can either be over a decade of content or one video. There could be a proportional judgement of strikes but that in itself would also be abusable – produce enough content to be judgement proof.
Wait a sec,
YouTube Warns That .. They Might Suck
lol