Social media

The shape of social media is inseparable from how it makes money. The early platforms experimented with several models subscriptions, virtual goods, banner ads before settling on targeted advertising. Facebook's revenue grew from essentially zero in 2004 to over a hundred and thirty billion dollars by 2023, almost entirely from ads. YouTube followed a similar path. Twitter and Snap built ad-funded businesses, less profitably. TikTok, owned by ByteDance, runs the same model with the same logic. The product these platforms sell is access to user attention, segmented finely enough that advertisers will pay a premium for it.

The technical consequence of the advertising model is that platforms optimize aggressively for what they call engagement time spent, sessions per day, content consumed, ads viewed. This is stated plainly in earnings calls and product documentation. The objective function is reasonable from a business standpoint and difficult from almost every other one. Engagement is a proxy for value, and a noisy one. Content that holds attention is not always content that informs or improves. Outrage holds attention. Anxious comparison holds attention. Optimizing for engagement means amplifying whatever incidentally produces it.

The public conversation about social media's effects on mental health has run ahead of the evidence and then occasionally circled back to it. Jean Twenge's 2017 book iGen and Jonathan Haidt's 2024 book The Anxious Generation argued that the rise of smartphones and social media after 2010 caused a measurable deterioration in adolescent mental health, especially among girls. The empirical hooks were real: rates of self-reported depression, anxiety, self-harm, and suicide among teenagers rose substantially in the United States, the United Kingdom, and several other wealthy countries starting around 2012. The timing aligned with smartphone saturation and the algorithmic turn. The correlation is robust across multiple datasets.

The disagreement is about what the correlation means. Critics including Candice Odgers and Andrew Przybylski argued that effect sizes in individual-level studies are small, that meta-analyses show only weak associations between screen time and well-being, and that confounding variables — academic pressure, economic anxiety, the aftermath of the 2008 recession could account for much of the trend. Haidt's response was that individual-level studies miss collective-level effects: a phone is harmful not only because of what one teenager does on it but because of what an entire peer group does, which reshapes the environment all of them inhabit. Whether this is defensible methodology or post-hoc rescue is genuinely contested.

For most of social media's history, the platforms operated under a legal regime designed for a different kind of internet. In the United States, Section 230 of the Communications Decency Act, passed in 1996, established that online platforms were not liable for content posted by their users and could moderate that content without becoming legally responsible for it. The law was written for early bulletin boards. It became the foundation on which Facebook, YouTube, and Twitter built businesses orders of magnitude larger than its drafters imagined. The argument for keeping it is that without it, no platform could host user content at scale. The argument for revising it is that the platforms doing so today bear little resemblance to what the statute envisioned.

The European Union moved first toward substantial regulation. The Digital Services Act, which took full effect in 2024, imposes obligations on large platforms around content moderation transparency, risk assessment, advertising disclosure, and access for vetted researchers. Very large online platforms, defined as those with more than forty-five million users in the EU, face additional duties around systemic risk. The Digital Markets Act addresses competition concerns by restricting how platforms favor their own services. Whether this framework will work as intended is unclear. Whether it represents a meaningful break with the previous regulatory posture is not.

Social media is not one thing, and that is part of why it has been hard to talk about clearly. It is a set of design choices made under commercial pressure, refined by feedback loops the designers did not fully anticipate, deployed to a population larger than any prior medium had reached. The early platforms genuinely did connect people who had been out of touch. The current platforms genuinely do produce some of the harms their critics describe. Both can be true, and separating which effects come from which design decisions is the work the field is doing now.