Let’s get one thing straight. That toxic, polarized, outrage-addicted hellscape you scroll through every day? It’s not an accident. It’s not because of “that one algorithm change,” it’s not because Elon is or isn’t in charge, and it’s not just because people are terrible.
You’ve been told a convenient lie. A comforting bedtime story for the terminally online that blames a rogue algorithm or human nature for the digital sewer we’re all swimming in. The truth, according to a study that just crash-landed on the internet, is far more damning.
WTF Is Going On?
A couple of researchers from the University of Amsterdam decided to stop complaining and start experimenting. They built a digital terrarium for AI-powered “personas”—little simulated users with opinions, hobbies, and political leanings taken from real-world voter data. They dropped them into a bare-bones social media model with only three functions: post, repost, and follow. They didn’t add a rage-bait algorithm. They didn’t program the AIs to be trolls. They just wound it up and let it go, expecting to have to add layers of complexity to see the system break.
But they didn’t have to add a thing.
To their surprise, the basic, stripped-down model immediately and spontaneously devolved into the exact dysfunctional mess we see today. All the greatest hits were there:
- Partisan echo chambers where everyone agrees with each other.
- Insane attention inequality, where a tiny handful of elite users get all the influence.
- The non-stop amplification of the most extreme and divisive voices.
The core functions that define social media are the very things that guarantee its toxicity. The problem is structural, baked into the foundation like a faulty load-bearing wall in a skyscraper.
Why It Matters: The “Fixes” Are a Joke
For years, we’ve heard the same suggestions for how to “fix” the internet. So, the researchers tested them. They put their simulated social network through six different “intervention strategies” to see if they could nudge the system toward something healthier.
The results were a clinical trial-level failure.
- Switch to a chronological feed? Sure, that slightly reduced the influence of mega-accounts, but it came with a nasty side effect: it supercharged the amplification of extreme content. You traded a king for a mob of screaming lunatics.
- Boost viewpoint diversity? Forcing users to see opposing political views sounds nice. In the simulation, it had no significant impact at all. Zero.
- Use “bridging algorithms” to elevate unifying content? This modestly improved viewpoint diversity, but at the cost of increasing attention inequality—making influencers even more powerful.
- Hide the stats? They tried hiding follower counts and reposts to reduce social pressure. Nothing.
- Remove bios to limit identity signals? Nothing.
They found that even the most extreme interventions barely moved the needle, and when they did, it often made another problem worse. Why? Because you’re not tweaking a system, you’re fighting its prime directive. The researchers concluded that the “mechanism producing these problematic outcomes is really robust and hard to resolve.”
This isn’t just about platforms being “evil” or users wanting a toxic environment. It’s the unintended, emergent outcome of a system whose basic rules reward reactive, emotional engagement above all else. That angry retweet doesn’t just spread toxic content; it actively shapes the network, creating feedback loops that ensure you see more of what makes you mad tomorrow.
Deeper Dive: The Prism and the Power Law
If social media isn’t reality, what is it? Think of it as a
“social media prism,” a term coined by sociologist Chris Bail. It doesn’t reflect the world; it refracts it, bending the light of public opinion into something far more polarized and ugly than it actually is. Most people have fairly reasonable opinions, but the prism ensures the most unhinged voices get the most attention, creating a false perception that the “other side” is populated entirely by monsters.
This is driven by a brutal mathematical reality called a power law distribution. In the real world, many things fall on a normal distribution—a bell curve. Online, you get power laws. This means that 1% of users dominate the entire conversation. It’s an attention monarchy, not a democracy. The system is designed for it, because on social media, attention attracts more attention.
This structure fundamentally changes our culture. It dictates which politicians become powerful by mastering the art of outrage. It reshapes how newspapers write headlines, turning once-serious outlets into clickbait farms just to stay relevant in a game rigged for sensationalism. You can’t just log off and escape it; the logic of the prism now infects everything.
What Happens Next? Pouring AI on the Fire
If you think it’s bad now, wait. The researchers have a grim prediction. The current model of social media may not even survive the rise of Large Language Models (LLMs).
Think about it. We’ve just established that the network is structurally designed to reward the most engaging, polarizing, and emotionally manipulative content. Now, introduce AI that can mass-produce that exact type of content infinitely, for free. Actors who want to monetize platforms like X are already using AI to generate posts that maximize attention, which often means flooding the zone with misinformation and rage-bait.
As the AI gets better, the human-generated content will be drowned out. The system will be completely overwhelmed by optimized sludge, and we’re already seeing people retreat into closed, private communities like WhatsApp groups to escape the noise.
Mic Drop
Stop asking how to “fix” social media. It’s like asking how to make a casino less profitable for the house. The answer is: you can’t. The game is rigged. The fundamental dynamic of posting, following, and reposting seems to inevitably create the exact pathologies we all claim to hate.
The platform isn’t the product. You are. Your attention is the currency, and your outrage is the engine. Every time you log on, you are a data point in the largest, most chaotic social experiment ever conceived.
The only question left is the one you should have been asking all along: are you running the tests, or are you the test subject?