Some of YouTube’s algorithms have been broken for years.
Some years ago my daughter who was around 4 at the time was watching some YouTube videos that were clearly for kids, part educational part entertainment. It might have been before they had YouTube Kids.
An ad came on for I think “Walking Dead”, showing really scary and gruesome Zombies. She was covering her eyes waiting for it to end.
I couldn’t believe it, “The Algorithm” decided an adult Horror series ad was appropriate to show in a kids video.
Sometimes, because some features are disabled on "for kids" videos, and you don't want people making videos targetted at kids and then marking them not for kids so that the features that YouTube doesn't feel are safe for kids are still enabled.
Overwhelming majority of kids videos are not marked as such. In particular, non English kids videos are super rarely marked as for kids.
That is the reason why youtube kids is something we gave up on fairly quickly. It is useless as majority of kiddy educational content we wanted them to watch was not available. And majority of what they themselves wanted to watch was unavailable too.
How does this even make sense though? In what world does it ever make sense to have an AI flag random videos as “for kids”? It should never happen in that direction, only the other. This is a basic conditional statement.
It’s obviously not the design spec to have to hand out tags randomly.
People monetize children's videos by saying they are not for children. Google have been in hot water for serving ads to children and the way around it is to remove ads (or classes of ads) from childrens videos.
Creators wants that extra money so they try to post children videos and calls them “for adult”.
Yes, it is their fault. But it is not like they decided “let us push adult content to children”. Nor is it “let’s randomly decide between adult and childrens content as that might be a great idea”
Obviously a mistake was made. But let’s be real: it was made in an attempt to right wrongs, not in an attempt to wrong rights.
Google got in trouble for doing tracking of kids for ad purposes on YouTube. Not tracking kids views is important to their regulatory compliance. Once of the easiest ways to do that is to have "kids" videos where the viewers are presumed to be children.
These videos will make less money (at least, I've been repeated told less tracking is less ad money), so Google has to worry about people mislabeling their content to get better ROI.
I'm not really an uploader so I haven't seen the feature. Is it a "three way switch": "for kids", "not for kids" or unspecified, or just "for kids" or unspecified?
It's either/or, and not very clear as to what the real difference is. I upload songs with occasional swearing in them, so I always tick not for kids on my uploads.
A year or so ago, every second image ad on YouTube was literal porn.
From a technical standpoint it was pretty obvious what happened. The image was overlaid with a grid of white lines. As the images were likely downsampled to reduce the required computing power, when the image was cropped right, the ML algorithm only saw a plain white image.
The only way to appeal this was a Google Support Forum where people posted screenshots and some "Gold Member Community Moderators" or whatever they were called, dismissed all complaints. They claimed that every ad was checked, porn getting through was impossible, and the people should factory-reset their phones as the only explanation was that viruses on their phone were injecting these ads...
Google's product forums are just users supporting other users. "Community moderators" and "product experts" are just users who have farmed enough upvotes to get a badge. Not sure what else is in it for them, but Internet points are hard to come by these days...
Yes, but the buck still stops at Google, especially if the forum is the only point of contact, and the community moderators act like they speak in behalf of the company.
Even Reddit, which is notoriously bad with its moderator policies, at least has a process how to raise disputes with moderator behaviour, and escalate them to actual company employees. Google does not even care.
I don’t trust “the algorithm” to pick ok videos for my kid, so I have a playlist of videos I’ve screened for when I think he’ll be old enough to watch (he’s not quite two). So far, tamer nature videos from BBC and similar.
Your ad anecdote makes me worry that even my approach isn’t careful enough.
I’m coming around to that as I now deeply understand my mother-in-law’s response to why she got rid of their TV when my husband was a toddler, and didn’t get one again until they’d retired:
“I didn’t have enough spare time to watch every program with them.”
And she was worried about what 1970s/80s West German broadcast TV (all publicly-funded) might expose their little developing minds to! She has a strong distaste for many of the post-deregulation commercial broadcasters in Germany, and an absolute horror of the Internet.
She never limited what they could read once they were old enough to go to the library alone, though, and as they were born in the 70s, the Internet wasn’t a factor until they were well into college and grad school.
I removed Youtube from the TV so that my children only now have access to age appropriate material on BBC iPlayer, Disney plus and Netflix.
There is some quality for children's material on Youtube, but it's by far in the minority and I just don't have time to monitor everything they watch, nor should I need to.
24 comments
[ 2.0 ms ] story [ 64.5 ms ] threadSome years ago my daughter who was around 4 at the time was watching some YouTube videos that were clearly for kids, part educational part entertainment. It might have been before they had YouTube Kids.
An ad came on for I think “Walking Dead”, showing really scary and gruesome Zombies. She was covering her eyes waiting for it to end.
I couldn’t believe it, “The Algorithm” decided an adult Horror series ad was appropriate to show in a kids video.
That is the reason why youtube kids is something we gave up on fairly quickly. It is useless as majority of kiddy educational content we wanted them to watch was not available. And majority of what they themselves wanted to watch was unavailable too.
https://www.youtube.com/watch?v=P28CKVT8TIU
Kids are the ones to whom videos need to be restricted.
Adults should be able to watch both kids and non kids videos.
So the only thing needed is for creators to mark a video as “for kids” and the algorithm to confirm that it actually is for kids.
Nothing else should even be available for kids.
If they want, they can have an additional option for creators which is “definitely not for kids”, much like you’d have an “explicit” option for music.
These wouldn’t require an algorithmic check and simply not be available to those under 18.
People monetize children's videos by saying they are not for children. Google have been in hot water for serving ads to children and the way around it is to remove ads (or classes of ads) from childrens videos.
Creators wants that extra money so they try to post children videos and calls them “for adult”.
Obviously a mistake was made. But let’s be real: it was made in an attempt to right wrongs, not in an attempt to wrong rights.
These videos will make less money (at least, I've been repeated told less tracking is less ad money), so Google has to worry about people mislabeling their content to get better ROI.
From a technical standpoint it was pretty obvious what happened. The image was overlaid with a grid of white lines. As the images were likely downsampled to reduce the required computing power, when the image was cropped right, the ML algorithm only saw a plain white image.
The only way to appeal this was a Google Support Forum where people posted screenshots and some "Gold Member Community Moderators" or whatever they were called, dismissed all complaints. They claimed that every ad was checked, porn getting through was impossible, and the people should factory-reset their phones as the only explanation was that viruses on their phone were injecting these ads...
Even Reddit, which is notoriously bad with its moderator policies, at least has a process how to raise disputes with moderator behaviour, and escalate them to actual company employees. Google does not even care.
Your ad anecdote makes me worry that even my approach isn’t careful enough.
“I didn’t have enough spare time to watch every program with them.”
And she was worried about what 1970s/80s West German broadcast TV (all publicly-funded) might expose their little developing minds to! She has a strong distaste for many of the post-deregulation commercial broadcasters in Germany, and an absolute horror of the Internet.
She never limited what they could read once they were old enough to go to the library alone, though, and as they were born in the 70s, the Internet wasn’t a factor until they were well into college and grad school.
There is some quality for children's material on Youtube, but it's by far in the minority and I just don't have time to monitor everything they watch, nor should I need to.
https://youtu.be/ZPdgWYB9fdw