YouTube has released a blog post on how it identifies policies that need to be developed and where it draws ‘the line’ on prohibited content.
The video and social media website has chosen an interesting time to put a post up like this. At the moment, Twitter (which was bought by Elon Musk two months ago) is going through a content moderation crisis, and the European Union has warned that the app could be suspended in member countries.
YouTube, meanwhile, wants to draw the line and make its moderation policies transparent to the public, and has given context for what guides its policy decisions and what would constitute content removal.
Of course, the platform noted that it works with NGOs and that it moved quickly to enforce rules against 5G conspiracies and anti-vaccine content.
But this post is about additional, internal content enforcement.
“Once we’ve identified an area where a policy update is needed, that’s where our trust & safety team comes in to develop a tailored solution,” global head of trust, safety and vice president of YouTube Matt Halprin and head of product management at YouTube Jennifer Flannery O’Connor said in the blog post.
“We start by assessing a few things. How commonly found is this specific type of harmful content on YouTube (and what’s its potential to grow)? And how is it managed under our current community guidelines?”
The blog post adds that it then comes down to watching between dozens and hundreds of videos, with the logic that a policy decision (such as a video takedown or the banning of a form of content) doesn’t just impact a single video — it impacts all videos.
So, policy decisions need to uphold two elements: they need to mitigate egregious real-world harm while balancing a desire for freedom of expression and allow for consistent enforcement by thousands of content moderators across the globe.
The team then mulls over different policy enforcement types (such as age restrictions) and the final send-off on the policy change is sent to the “highest levels of leadership”, including the CEO of the company, Susan Wojcicki.
But it all comes down to context, which is where the post mentions the strengths and limitations of machine learning.
AIs can assist YouTube moderators in flagging potentially harmful content, for example, but it all comes down to context and what the model would struggle to identify.
The post gave an example of a Hitler speech uploaded as a video, and when it would be necessary to take down.
“A speech by Hilter at the Nuremberg rallies with no additional context may violate our hate speech policy. But if the same speech was included in a documentary that decried the actions of the Nazis, it would likely be allowed under our EDSA guidelines. EDSA takes into account content where enough context is included for otherwise violative material, like an educational video or historical documentary,” the post reads.
“This distinction may be more difficult for a model to recognise, while a content moderator can more easily spot the added context.”
It’s good to see YouTube be proactive on what constitutes a takedown and what goes into its policy development, although keep in mind that the likes of Tim Pool and Stephen Crowder are still operating on the platform, while spruiking harmful ideas.
Not that this is news. Back in 2019, YouTube argued that keeping Crowder’s hateful content on the platform was a form of debate.
“We generally believe that open debate and free expression leads to better societal outcomes. But we’re careful to draw the line around content that may cause egregious harm to our users or to the platform,” the post also reads.
So, you know, good to know that YouTube is drawing the line somewhere.