Note: The discussion was held under the Chatham House Rule. All quotes have been edited for clarity, brevity and anonymity.

“Reddit contains more pornography than a pornographic site at this point in time. So what we’re actually discussing is that there are sites which contain all kinds of information, where restricting access based solely on age may not be sufficient. Instead, there should be a consistent mechanism to ban content, not simply to ban the platform,” said one of the participants during MediaNama’s roundtable discussion on Age Verification and Restricting Social Media for Children on May 15 in Bengaluru, while discussing content classification and moderation as a possible alternate or alongside solution for protecting children from online harms.

“You can’t do content regulation, you can only do platform regulation,” said MediaNama’s editor Nikhil Pahwa. “Banning content seems impossible to me because there are billions of hours of content, billions of words of content that go up every hour in the world,” he further elaborated. He was responding to the comments on content-based classifications alongside the age-based restrictions.

One of the speaker emphasized on:

Community-based content moderation as an alternative to bans? “So self-declarations may not sufficiently work. We also need to understand that there are technical limitations even in content restriction. Unless there is a community-based approach that allows you to filter content, this may not be a sustainable mechanism. Age-based restrictions alone will not solve this problem,” the participant concluded.

“No, no. Again, you have moderators. So, the community can be gamed. It is gamed. You have downvoting on apps all the time. Community notes are also gamed. None of that’s a solution,” counted Pahwa. 

Can we create an automated YouTube-like content classification system for all content? “Regarding content classification, just try doing this: try uploading a video on YouTube. If you do anything involving pornography or harsh language, your video doesn’t get uploaded. So I’m really trying to understand what the content problem is here.” Pahwa further explained why it is difficult to implement an always-accurate content classification system at scale for all the content on the internet.

He said: 

“No, no. Content is the hardest problem to solve. You may classify content, but you may also classify it falsely. You’re only looking at extreme situations. Content—I work in content—is very difficult to identify. For example, is something hate speech or not, right? The internet finds mechanisms for bypassing filters,” said Pahwa. He then cited leet codes as a mechanism which people generally use to bypass language-based restrictions.

“It’s difficult, and especially when you’re looking at billions of pieces of content—and you’re only looking at YouTube. I’m saying there are people who gravitate towards platforms that maybe don’t have classification or restrictions because they want to watch a link that was sent to them by someone. So I don’t think content classification is feasible,” he continued.

Pointing out the content classification problems, one of the parent participants said the limitations of wrongful classifications and the lack of parental autonomy if the parent wants to override the classifications and restrictions decided by the platforms.  

The parent’s need to override YouTube’s content classifications: “Let’s say YouTube has a fantastic content detection algorithm, and it’s generally very good. Let’s say it works 80% of the time. Now, as a parent, that’s great for me because I have a generally safe pool of content to give my child. But it’s not 100% accurate. And what happens when it’s wrong and blocks my child from accessing content that I want my child to access? Where is my ability to say, “I disagree with this, and I want to override it? That power is what I’m asking for,” said a technologist parent during the discussion on parental agency over sole platform dependency.

Parents’ lack of knowledge on what content children access: “It’s not ignorance about the device. It’s ignorance about what websites they access and what social media content they consume,” said a doctor.

“Are you sure that all the [YouTube] videos which are meant for kids are safe? How many of you agree that Peppa Pig is for kids?” asked one of the participants at the discussion, and complementing this statement, another speaker pointed out Cocomelon cartoon videos for their overstimulating nature. “Cocomelon is designed at the angle of where the kid is looking from; the fast frame changes—all of that is meant for addiction by design,” Pahwa commented.

While the participants and speakers were discussing the issues with content classification, another speaker pointed out the platform classification itself as a problem.

He said that: 

WhatsApp is not just a messaging platform, it’s a content platform too: “I have platform-based [parental] control, but I have zero insight into what content she’s accessing. And platforms are not cleanly divided. WhatsApp, for example, is not just a communications platform. It’s also a content platform; that’s what status updates are. There’s nonstop content flowing through there. My daughter is learning facts about the world from random status updates, not from people messaging her. I can’t turn off individual features within an app. And because this is from Google, I can’t turn off ads at all,” said a technologist, who further said that he had to do network-level configurations to block ads because Google doesn’t give that modularity.

The need for consensus on defining the ‘harmful content’ before legislating: “Of course, there are some problems that are very easy to identify—pornography, violence, and so on. But there are other problems, like you said, such as hate speech. These are very nuanced issues. Try to find a principle that almost everybody agrees on, and then legislate accordingly. Again, adopt a harms-based or risk-based approach. Establish principles rather than laying down rigid rules about what people must do. And then, of course, use technology to identify gaps and plug them. I think this is the only way things are going to work, especially when it comes to protecting children,” said a female speaker. 

The content classification problem in the Central government’s proposed age-graded restrictions: “So if you have graded verification—let’s say, I mean, look, if you’re going to have [age] grades, I get that—there has to be a mechanism for identifying whether the child belongs to that grade, right? So how do we enforce that? Because a 10-year-old can view a 16-year-old’s content too. How do you enforce it? Otherwise, we’re facing an impossible task, I would argue. It’s impossible to enforce even a graded mechanism for access to content,” said Pahwa.

He was referring to the central government’s possible approach to address online harms instead of blanket bans. In March, media reports indicated that the union government may introduce three-tier social media restrictions.

The state doesn’t understand the nuance between good and bad content: “As a parent, I want to control what my child has access to, and that control should be in my hands. The moment the state decides what is good or bad, where is the state’s demonstration of nuance in understanding content? Because we seem to be banning platforms rather than banning specific content, or restricting platforms rather than restricting content,” said Pahwa.

If the harm is real, regardless of age, why are we only protecting one age group? One of the speakers argued that it’d be like giving a free hand to platforms if the government mandates them to keep online space safe only for children and not for adults. “If these [online] harms exist outside these specific age groups as well, are we then, with any form of age-based regulation, giving companies a free pass to not moderate content and not actually curb harmful content for anyone beyond the age of 16 or 18, or whatever the criteria are for being considered a child?”

The speaker further pointed out the enforcement problem: 

“With such regulations, you’re also designating some party as legally liable for enforcing them. Do we have the capacity to enforce it and hold companies liable, or could we slide down the path of holding parents legally responsible instead?” questioned the speaker.

Note: This discussion was organised with support from Meta and Snap. Community partners for the event included Pacta, Design Beku, and The Pranava Institute.

Read MediaNama’s whole coverage of this discussion here.

Also Read: 

• Karnataka CM Proposes Social Media Ban for Under-16s: What Challenges Lie Ahead?
• Digital Literacy: How parents and schools can build children’s agency against online harms
• The harms of social media: What is driving proposed age restrictions for children on the internet