Think of the last thing online that made your blood boil, the post you couldn’t scroll past without reacting. Now try to recall the last calm, carefully-qualified piece you read on the same subject. You probably passed the first one along and forgot the second, and so did almost everyone else. That isn’t because people are fools: ask them, and most will say they wish the furious stuff wouldn’t spread. It spreads anyway. This is about why. Once a network gets large enough, what reliably travels is whatever hits a raw nerve: fear, anger, us-against-them, the rush of belonging. I’ll call that the floor, not the only thing that spreads, but the lower limit the system keeps settling back down to.
Chapter 6 worked out where the distortion gets produced, in the bridge zone that staffs the News and Meme stages. The question here is what that zone settles on once the network it serves is big enough. Two ways the argument goes wrong, and I want to avoid both. Said badly, “raw emotion is the floor” is either obvious or cynicism in a lab coat; I want the structural version, an equilibrium of named mechanisms, not a verdict on human nature. And Hugo Mercier’s *Not Born Yesterday* argues that people are much harder to manipulate than this story seems to need. I take that seriously below, and I think the argument survives it, narrowly and in a specific place.
The floor, plainly
Once an information network is big enough and its gates have rewarded engagement long enough, the equilibrium of what propagates converges on what reliably evokes raw strong emotion in the receivers it has to cross: threat, in-group identity, out-group animosity, status anxiety, awe, anger. That set is the floor. Not the ceiling on what spreads, but the lower bound on what kind of content spreads once scale and time have driven the gates toward engagement-fitness. The list is a set, not a ranking; which payloads dominate depends on how the gates are tuned, and the current tuning favors the corrosive end (anger over awe).
The floor is about what wins at the gates, aggregated across enough receivers and enough time, not about any one receiver’s taste. A given reader may prefer nuance, and their preference doesn’t change what crosses the gate, because the gate runs on what aggregates to the highest engagement across a heterogeneous population, and that reliably converges on emotional payload. The floor is a property of the aggregation, not of the receivers individually. That is what makes it stable: no individual has to want it for it to win.
Three quick clarifications. Raw is doing work: the floor is low-precondition, immediately-triggered emotion (threat you needn’t argue for, outrage that skips the evidence), not the sophisticated kind. The floor is medium-conditional, not universal; different gate criteria yield a different floor, though the differences are smaller than one might hope, because the human attentional substrate is partly shared. And it is a lower bound, not a ceiling: plenty of above-floor content reaches scale (popular science, depth journalism that finds its audience), just against the gradient the floor sets.
Four lines of evidence
The Ukraine shift. *Psychology of Virality* (p.14) found that out-group animosity predicted virality in Ukraine before the 2022 invasion, and in-group solidarity surged as the strongest predictor after it. What spread changed with the context; what kind of content topped the ranking did not. The floor moved within the emotional set and never left it. That is its signature: different specific emotions win at different times, but emotional content of some kind reliably outranks calibrated non-emotional content.
The substrate. The same paper (p.8) names why: “Our attraction to high-arousal negative information likely has deep evolutionary origins… Social media algorithms appear to take advantage of our attraction to threat.” The gates exploit an attentional pull that predates any medium. The substrate makes the floor possible; the gates make it the equilibrium.
The paradox of virality. Also in that work (p.9): people across the spectrum say they do not want high-arousal negative content to go viral, and it goes viral anyway. A receiver-preference theory of spread cannot explain that; a gate-driven theory that exploits the attentional substrate can. Neil Postman saw the same structure in television (p.86): a medium “largely aimed at emotional gratification” makes emotional gratification the standard for what content looks like, the Huxley mode at a new scale.
Held at arm’s length. The floor’s winners are mostly reflective beliefs, in Dan Sperber’s sense (Sperber et al.): things you can state, defend, and repost but never act on, as against intuitive beliefs that silently drive what you do. You “believe” you should floss nightly and never do; you believe the floor will hold your weight and never say it aloud. Most of the alarming material we forward is the first kind, held the way you hold a bumper sticker, not the way you hold the belief that keeps you off a ledge. That is what the paradox of virality looks like from the inside: people share the furious thing as a coalition-marker without holding it in the way that would move their hands.
The four compose. The attentional substrate is real; the gates exploit it; reflective preference doesn’t override the substrate-plus-gates equilibrium; and what the equilibrium produces is mostly belief held at arm’s length. The floor is the equilibrium of those pieces, and it is what the modern environment has settled at.
Engaging Mercier
Not Born Yesterday is the strongest objection to anything like the floor argument, so it deserves a direct hearing. Mercier reframes Sperber’s epistemic vigilance as open vigilance (p.31): humans are simultaneously open to useful information and on guard against unreliable sources, the two evolving together, and our errors run more toward not trusting when we should than the reverse. The historical record backs him: “the vast majority of mass persuasion efforts, from propaganda to political campaigns, from religious proselytizing to advertising, end in abject failure” (p.259), and even Nazi propaganda worked only where it “build[t] on existing consensus” and “failed abysmally” where it ran against public opinion (p.131). Strip the vocabulary and the point is plain, and a little flattering to the species: people are by default more suspicious than gullible, and the grand “the masses were brainwashed” cases mostly turn out to be people accepting what they already half-believed. An argument that the system floods us with emotional junk and the junk sticks has that to answer.
Four responses, and the argument survives on the fourth. First, vigilance protects against believing, not against attending. The floor wins by exploiting the substrate vigilance runs on, not by beating vigilance: attention is allocated to threat before vigilance ever engages, and what reaches the meme stage is set partly by that first pull. Second, vigilance is one criterion among many at the moment of action (identity, threat, novelty, time pressure), and when the others pull harder it does not categorically win, which is precisely the paradox of virality: reflective preference says one thing, the sharing finger does another. Third, and more speculatively, vigilance is trainable, and a medium that rewards low-friction, identity-confirming sharing plausibly trains it down over a lifetime; I hold this as a hypothesis, not a result. Fourth, Mercier and the book are operating at different timescales. His evidence is within-campaign and within-experiment: Cambridge Analytica’s tiny measured effects (p.139), belief-corrections that landed without shifting behavior (p.205). At that scale, mass persuasion does routinely fail, and that has to be respected. The book’s worry, here, is generational: decades of engagement-tuned media shaping what receivers come to want and discount. Mercier’s question is whether a campaign can install beliefs in a vigilant population in one cycle (usually no); the book’s is what decades of such media do to disposition, and his data does not reach that far.
Mercier has a deeper argument that cuts at the book’s causal arrow, and I think he is partly right. His reframe: people don’t do bad things because they hold false beliefs; they hold false beliefs to justify what they had already decided to do (p.202). The diagnostic case is that when Trump supporters accepted corrections to specific claims, their support didn’t move (p.205): the belief wasn’t load-bearing for the behavior. He calls the resulting dynamic the justification market, where behavior is set upstream by identity and interest and receivers then demand content that flatters the decision. The belief is the alibi, not the engine. But the arrows run both ways. Some outcomes trace through belief (a population that doesn’t understand vaccines decides differently); some through justification-after-the-fact (an identity-committed population shops for cover). They compose, with the weight depending on whether the behavior is information-driven or identity-driven, and the floor supplies both installable beliefs and on-demand justifications. Whether the justification market and Chapter 10’s captured equilibrium are two mechanisms or one seen from both ends, the gates shaping demand while demand is what the gates tune against, I can’t yet cleanly separate, and I flag it rather than resolve it.
One boundary on all of this. Mercier’s readings of the canonical cases (Asch, Milgram, the propaganda record) are his interpretations, not consensus ones, and his broader argumentative theory of reason has its own critics. So only the parts that sharpen the diagnosis get borrowed, the timescale distinction, the behavior-to-belief reframe, and the reflective/intuitive split, without adopting the framework wholesale.
The floor in the book’s vocabulary
Walk it through the machinery. Transport throws off a cloud of variants (the neutral drift Information Evolution documents); selection at scale picks the winners on engagement-fitness, which converges on emotional payload. The medium sets those gate criteria, so the floor is specific to the media we built, not universal. Capture tunes it lower than an uncaptured medium would, because a captured medium runs on what holds attention longest and the most attention-holding emotions are also the most corrosive: the captured medium doesn’t just settle on a floor, it tunes the floor downward, toward anger over awe and out-group animosity over in-group bonding. Chapter 11’s LLM doesn’t escape the floor; a model tuned for user satisfaction reproduces it at industrial scale, and only the salvation case (faithful model, substrate custody) could produce above-floor content. And Chapter 6’s bridge zone is the floor in operation: its operators populated the floor, its audience selected for it, its product reflected it. Chapter 6 named the stage; this names the equilibrium the stage settles into.
What the floor predicts
The floor argument predicts: mass-network propagation dominated by emotional payload, with the specific emotion shifting by context (Ukraine); corrections systematically underperforming the originals they correct, because corrections carry less emotional payload; and engagement-tuned media producing a worse floor than media with different gate criteria (a comparative, partly testable claim).
It does not predict that every message is emotional (plenty of calibrated content spreads, just not at floor rates), that receivers are gullible (the floor wins despite vigilance, not by defeating it), that quality content can’t reach scale (Wikipedia and mature open source are higher-floor sub-networks inside the larger one), or that the floor is permanent (it is the equilibrium of substrate plus gates plus want-training plus capture state, and changing those changes it).
The design target
The value of the structural version is that it locates the breakage at the equilibrium, not at any operator, which tells a prescription what to design against: not the people, but the arrangement that produces the floor regardless of who runs it. Four structural moves are available. Build sub-networks with different gates that produce a higher floor (Chapter 12’s curation layer, smaller than the mass network but above its floor). Tune the mass network’s gates against the captured equilibrium (Chapter 10’s reform, hard and structurally resisted). Soften the per-claim cost so above-floor content fits a limited budget (Chapter 11’s decompression, conditional on custody). And cultivate open vigilance directly on the training side: if vigilance is trainable down, it is trainable up, which makes vigilance-cultivation a sibling of Chapter 8’s capacity-building rather than a separate project. Any prescription that engages none of the four is implicitly accepting the floor as the population’s information environment; Chapter 12 engages them together.
Where I land
The propagation equilibrium of mass networks with engagement-tuned gates converges, at decades-scale, on raw strong emotion as the floor of what reliably spreads, through supply-side gate-tuning, demand-side justification, attentional-substrate exploitation, and cumulative want-training of the receiver’s vigilance. The floor is medium-conditional, substrate-anchored, tuned lower by capture, and its most reliable product is belief held at arm’s length, with a smaller subset crossing into the kind that moves behavior. Mercier’s open vigilance is real and the argument survives engaging it, because vigilance acts after attention is allocated, is one criterion among many, is trainable, and his evidence is single-campaign while the worry is generational. The floor is what the environment has settled at, and the book’s prescription is what would have to change for the equilibrium to move.
Where I’m still uncertain
- The timescale distinction is load-bearing and empirically untested at the scale it needs. The strongest move under Mercier’s pressure is that his data is within-campaign and the worry is generational, but generational scale is where direct evidence is hardest. The cohort that grew up entirely inside engagement-tuned media is now mature enough to compare against earlier ones, and the book owes that comparison rather than asserting the effect.
- The cross-cultural claim is asserted, not argued. I’ve committed to the floor being similar across cultures with similar media exposure, on the grounds that the attentional substrate is shared. There is comparative work (and counter-work) on whether emotional-payload virality varies by culture, and a careful version would either cite it or bound the claim to Anglophone digital media.
- “Raw strong emotion” may flatten a distinction that matters. Anger and awe both reach floor-class virality but have opposite downstream effects: anger drives polarization, awe sometimes drives synthesis. Bundling them under one label may hide the difference a prescription would most want to exploit.
← Chapter 6: Where It All Gets Fucked Up · Interlude: Myths at Scale →