The Theater Troupe That Gave Away the Trick
A small theater troupe is casting a play. One actor picks the mayor, another the schoolteacher, a third the local shopkeeper. Once you choose a role, you stick with it. The mayor can't suddenly quiz kids on homework. The shopkeeper can't hand out parking tickets. The costume and script lock you in, and the audience notices the second you slip.
Around 2016, thousands of fake accounts on a major social platform pretended to be ordinary people. Earlier efforts to sort them mostly asked: liberal or conservative? Fearful or cheerful? But that only described what they said, not who they were pretending to be. Nobody asked the casting question: are you playing a local newspaper, an activist, or just some person chatting about sports?
Someone finally asked that question. Four roles appeared. Some fakes dressed up as small-town news outlets, posting calm local stories. Others posed as movements, always on-mission. A third group played loud political voices that never switched sides. The rest acted like regular folks, chatting about weather and TV. Each role came with strict unwritten rules, just like actors in a troupe.
Here's the surprising part. More than half of all posts from these fakes weren't political at all. They were camouflage: complaints about Mondays, opinions on TV shows. A shopkeeper actor has to arrange shelves and greet customers before delivering the big line in Act Three. That filler isn't wasted. It's the cost of staying in character. And each role camouflages differently.
To test this, analysts sorted accounts into the four roles by reading their profiles. For accounts with hidden profiles, they compared hashtag patterns to the sorted ones and matched them to the closest role. Like recognizing which actor is which even with costumes half-hidden, just by watching their gestures on stage.
The approach correctly identified an account's role nearly nine times out of ten. When checked against a separate batch sorted by different analysts, accuracy stayed just as high. Even in a completely different language, the same patterns held. The shopkeeper acts like a shopkeeper whether the play is in English or Russian.
This doesn't spot fakes among real accounts. But once fakes are found, it maps how the whole network is organized: which roles draw the biggest audiences, which ones do the talking, which ones just amplify. Fake newspapers gathered huge followings but rarely replied. Regular-person accounts did most of the chatting. Reading that structure is like reading a cast list before the play starts.
And here's the irony. To fool people, each fake account had to follow the same social expectations a real person or outlet would. That obedience left footprints as distinct as stage blocking. The very effort to look authentic made the whole network readable. Because these patterns cross languages and borders, the same lens works wherever the next curtain rises.