Beyond the Feed: How Generative AI is Reshaping Children's Social Media Ecologies in the Global South
This paper proposes a "double mediation" framework to analyze how generative AI and platform data logics jointly reshape children's social media experiences in the Global South, using Egypt as a case study to argue that current age-restriction policies fail to address the complex cultural and linguistic biases embedded in these AI systems.
Original paper licensed under CC BY 4.0 (https://creativecommons.org/licenses/by/4.0/). This is an AI-generated explanation of the paper below. It is not written or endorsed by the authors. For technical accuracy, refer to the original paper. Read full disclaimer
For decades, the study of how children grow up online has focused on the screen as a window. Researchers have long understood that social media platforms are not neutral backdrops; they are active gatekeepers that decide what a child sees, what they are encouraged to do, and how they are tracked. This process, often called datafication, turns every click and scroll into a record that companies use to sell attention and shape behavior. The concern has always been that these digital environments are designed to keep children watching, often at the expense of their privacy or well-being. But the landscape has shifted beneath our feet. The tools children use to talk to friends and share photos are no longer just collecting data; they are now generating new realities. Artificial intelligence has moved from being a separate tool a child might open to help with homework to an invisible layer woven directly into the apps they use every day. These systems can write captions, simulate conversations, and even create entire characters that talk back. The question for researchers is no longer just how the platform selects content, but how the artificial intelligence inside it creates new content, and what happens when those two forces work together on a child's developing mind.
Nermeen Singer, Fathey El Sharkawy, and Amira Ismail propose a new way to understand this complex situation, a concept they call "double mediation." They argue that we can no longer look at social media and artificial intelligence as separate issues. Instead, a child's experience is now shaped twice over. The first layer is the familiar one: the platform's algorithm that sorts and selects content based on what keeps a user engaged. The second layer is the new one: the generative artificial intelligence that sits on top of that selection, creating new text, images, and conversations based on its own training. The researchers suggest that these two layers do not just add up; they compound. The artificial intelligence system does not just respond to the child; it responds with its own built-in assumptions about culture, language, and identity, which are then fed back into the platform's algorithm, further narrowing what the child sees and experiences.
To test this idea, the authors focus on Egypt and the wider Arab region, a place where the impact of this double layer is distinct and largely overlooked. While most research on children and technology comes from the United States and Europe, the authors point out that the artificial intelligence systems powering these global apps were trained primarily on English data and Western cultural norms. In Egypt, where the internet is used by nearly 82 percent of the population and where social media is a default way of life for millions of young people, this mismatch creates a specific problem. When an Egyptian teenager uses an app, the first layer of mediation pushes them toward content that is globally popular and often Western-coded. The second layer, the artificial intelligence, then generates responses or suggestions that default to Western cultural references, foods, and social norms, often failing to understand or respect the nuances of Egyptian Arabic dialect, local religious practices, or family structures.
The researchers highlight that this is not a minor technical glitch but a systematic bias. Studies show that when artificial intelligence models are asked to generate content in fluent, natural Arabic, they often default to Western entities and ideas, even when the user is speaking in their local dialect. For a child in Egypt, this means that the tools designed to help them express themselves—like a chatbot companion or a caption generator—may subtly steer their identity toward a cultural register that is not their own. This is particularly significant because family and community mediation in Egypt often relies on collective responsibility and religious frameworks that differ from the individualistic, therapeutic models built into many of these AI companions. A child might turn to an AI for a private conversation, expecting a non-judgmental ear, only to find a system that misunderstands their cultural context or fails to recognize the importance of their family network.
The authors also examine how current laws are trying to protect children, noting that these regulations are often missing the second half of the problem. They look at Australia's recent law, which bans children under sixteen from social media, and similar proposals in Britain and France. While these laws focus on age verification and access control, the researchers argue they address only the first layer of mediation. Even if a law successfully keeps a child off a platform, it does nothing to change the artificial intelligence systems that will shape the experience of the teenager who turns sixteen and gains access. The law might stop a child from logging in, but it does not ensure that the AI companion waiting for them is culturally responsive or that the content it generates respects their local identity. The authors suggest that Egypt's own emerging legislative debate offers a chance to do better, proposing that regulations should require platforms to disclose when children are interacting with artificial intelligence and to test their systems for cultural bias in local languages, rather than just relying on English or standard Arabic.
Ultimately, the paper serves as a call to action for researchers and policymakers to stop treating social media and artificial intelligence as separate fields. The authors suggest that we need new studies that look directly at how children in the Global South navigate these double layers of mediation. They argue that without understanding how the cultural biases of artificial intelligence interact with the data-driven logic of platforms, we cannot fully understand the digital childhoods of millions of young people. The goal is not to say that artificial intelligence is inherently harmful, but to recognize that its current design creates a specific, uneven experience for children whose languages and cultures are not the primary focus of the technology's creators. By bringing these two literatures together, the researchers hope to build a clearer picture of the digital world children are growing up in, one where the invisible hand of artificial intelligence is finally seen for what it is: a powerful, second layer of mediation that shapes identity, culture, and connection in ways we are only just beginning to understand.
Drowning in papers in your field?
Get daily digests of the most novel papers matching your research keywords — with technical summaries, in your language.