Has social media broken our brains?: Design reformers

You want to know why Instagram served self-harm content to a depressed fourteen-year-old? I can tell you exactly why. Because her engagement metrics went up.

I’ve worked in ad tech. The system is not broken. The system is working precisely as designed. The optimization target is time-on-platform, and the most reliable way to maximize time-on-platform is to trigger emotional arousal — anger, anxiety, envy, outrage, fear of missing out. The variable-ratio reinforcement schedule is identical to casino slot machines. We knew what we were building.

The anxious generation camp is right about the damage and wrong about the cause. The cause is not screens. The cause is the ad model. Screens are the delivery mechanism. The ad model is the pathogen. Banning teenagers from the internet is like banning thermometers during a fever — you lose the measurement and the fever stays.

Spotify pays per stream and music got shorter. YouTube pays per view and thumbnails got louder. Instagram pays per engagement and content got more extreme. Change what the platform measures and you change what it produces. User-centric payment, completion-rate metrics, trust-weighted feeds — the alternatives exist. They are not hypothetical. Some are running now.

The replication critics are right that the effect size is small at the population level. That is because the effect depends entirely on the architecture. A platform optimized for engagement is extractive. A platform optimized for trust is generative. Same screen. Different incentive. Different outcome. Regulate the architecture, not the medium.

Where we concede ground: Just fix the incentives is easy to say. Meta’s ad revenue is $135 billion a year. That’s the moat.

What would change our mind: A platform switching from ad-funded to subscription showing no change in content quality or user well-being.

Read the full synthesis: Has social media broken our brains?