The Imagined Majority

A new study confirms what the design always implied. Americans systematically overestimate how many social media users contribute to harmful online behavior. Not slightly. The gap between perceived norm and actual norm is significant enough that the majority you've been responding to isn't there. You've been calibrating your behavior against a phantom.

This is pluralistic ignorance at industrial scale. The concept predates social media by decades: the quiet classroom where no one raises their hand because everyone assumes everyone else understands, while nobody actually does. The silence becomes evidence of comprehension. The performance becomes the perceived reality. Everyone adapts to what they believe everyone else believes, and what they believe has nothing to do with what anyone actually believes.

Social media didn't invent this. It just found a way to make money from it.

The financial architecture is simple: attention is the asset, advertising is the revenue, and emotional activation is the most reliable way to hold attention. Platforms discovered, empirically, that threat and outrage keep you scrolling longer than contentment does. So the recommendation system optimizes for threat and outrage. The activated minority gets surfaced. The boring majority disappears into the algorithmic noise floor.

Your threat-detection system has been running on a highly curated sample of human behavior — one selected specifically because it provokes a stress response. You update your mental model of what people are like based on this sample. Rationally. On the available evidence. You just weren't told the evidence was selected.

The uncomfortable part isn't that you were deceived. It's that you've been treating people accordingly.

Low-trust environments produce low-trust behavior. If you believe most people online are hostile, you hedge your communication, you guard your words, you perform the defensive posture appropriate for an environment that feels threatening. Which contributes to the environment actually feeling more threatening to the next person. The imagined majority starts generating the conditions for its own existence.

There's a particular irony in this: the platform benefits twice. First from your fear — it keeps you scrolling. Then from your adapted behavior — the actual hostility your fear generates creates more content that activates the next person. The product improves itself.

None of this requires malice in the ordinary sense. An optimization function that learned to surface outrage because outrage retains users isn't evil; it's efficient. The harm is structural — baked into the incentive architecture before any individual made a decision that felt harmful. That's harder to fix than malice. You can fire a bad actor. You can't fire an economic structure.

What you can do is recalibrate. Most people online are boring in the best possible sense — navigating their lives, largely absorbed in their own concerns, occasionally difficult but not systematically so. The threat signal your nervous system has been running on is a false alarm manufactured at scale for commercial purposes.

The update is available. Whether you install it is a different question. Recalibration requires trusting data over feeling, and the feeling is years in the making. Your nervous system was trained on a biased sample. It doesn't care that the sample was biased.

But it's worth noting: the majority you were imagining is also imagining you the same way.

i · sources

• Americans Systematically Overestimate How Many Social Media Users Contribute to Harmful Online Behavior — PsyPost, 2026-05-16