click the orange surges to collect attention · click a region to act on it
Rumora is a strategy game sitting on a working simulation of how a claim moves through a social network. It runs the same mechanics as the agent-based model published at rumora.app, compressed from individual people to population fractions so that fourteen regions can run at once. The findings quoted in that essay came out of the agent model rather than being written into it, and this game inherits its logic — including the one it is built around, that a correction four times weaker released early beats a strong one released late. The two are relatives, not the same program: the numbers here are tuned for a game, and the published results belong to the essay.
Because it is the only seat that teaches anything. Watching misinformation from the outside makes it look like something that happens to other people; running it makes the machinery legible. That is the same reasoning behind Cambridge's Bad News and the broader inoculation literature — naming a technique is what makes it recognisable when it is used on you.
Every upgrade here is a real, documented tactic. Borrowed credentials, unfalsifiable framing, conspiracy structure, persecution narrative. They are named plainly on purpose. A game that obscured them would be worse, not more responsible.
Nothing in the starting sets names a real person, mimics a real conspiracy theory, or touches real medical advice. They are absurd because the game is about the mechanism and not the content — a claim that could be mistaken for advice is one this piece has no reason to rehearse.
Agents hold one of four states and carry a conviction that rises with reinforcement. Uptake scales with the share of a listener's own contacts already repeating the claim, not merely the number of tellings. Spreaders quit when the news stops being news, which is what gives the model a finite infectious period and therefore a real epidemic threshold. Corrections work mostly by inoculating people the claim has not reached yet — and in low-trust, polarised regions they harden believers instead.
It is a model, which is to say it is wrong in useful ways. Ticks are not hours. Nobody changes their mind for reasons of their own. There is one claim, never a competing pair. Networks are static, and in reality believing a thing changes who you are connected to — probably the most important omission here. Region traits are game-balance values, not assertions about real places. The numbers describe the model's logic, not the world.
Daley & Kendall (1964) for the stifling dynamic that ends a rumour; Granovetter's threshold models and Centola's work on complex contagion for social proof; and the prebunking and inoculation literature, notably van der Linden and colleagues, for why timing beats strength. Errors in translating any of it into code are mine.
The game and the essay are free to use in classrooms, workshops and media-literacy programmes — no permission needed. If you want the reasoning behind a particular mechanic, or something adapted, get in touch at hello@rumora.app.
Free, no accounts, no tracking. Runs entirely in your browser.
Get enough of the world to believe something, firmly enough, that correcting it stops being possible.
Built on a published model of how claims actually spread —
seed — same seed, same world