On Monday morning, people on Twitter began posting photos of graffiti across Boston and attributing it to the internationally known street artist Banksy.
The hashtag #banksyinboston started trending locally as photos of Government Center, Trinity Church, and TD Garden appeared with some sort of decals on them.
However, people could not find any examples of the art in real life. Several pictures looked suspiciously Photoshopped. A handful of people who noticed the tweets investigated and promptly refuted the images spreading around.
The people tweeting out the Banksy photos were locals, according to their profiles, with with hundreds and even thousands of followers. One was a journalist, who asked if Boston Magazine or Boston.com was going to cover Banksy's newly discovered presence in Boston. Some of them listed Boston University in their profiles. Others were fans of Boston sports. But most of the accounts did not exist prior to September of this year.
They were all fake.
The Twitter accounts were run by bots and the people who created them were thousands of miles away.
A number of the fake accounts included a hashtag of #02805signup, which is not a zipcode in Rhode Island, as initially assumed by this reporter. The hashtag lead to a student's Twitter account, which mentioned a name. A Google search for that name and 02805 lead to a class course description.
Left: Meet Piotr Sapiezynski and Sune Lehmann, a teaching assistant and professor, respectively, at the Technical University of Denmark in Copenhagen. Right: Among the classes they teach is a graduate course on Social Graphs and Interactions, which focuses on the coding language Python and Twitter's API.
Lehmann teaches a class on network science and machine learning and this year decided to focus that on building bots on Twitter. Though a normal bot on Twitter has almost no followers, no activity, and exist to just inflate follow counts for people who pay for them, Lehmann's class wanted to create "intelligent" bots that would build and interact with Twitter followers and tell the difference between other bots and humans in their stream. And then influence them.
The class had 14 bots and wanted to see if they could make something trend on Twitter. But there wasn't enough people in Copenhagen who used the social media service.
"Since both Piotr and I lived in Boston, that would be closest to our hometown that had a significant twitter environment," said Lehmann. The students created bots that could fool people into thinking they were Boston locals. "What you're seeing in the profiles are just whatever [the students'] perception of Bostonians are," said Lehmann. "We were cringing at some of their choices but we had no influence there."
Around Thanksgiving, the class tried to spread #bostonthanks, hoping that hashtag would pick up and become the "official" local trend. But it didn't. Then the class tried to make #MeInThree trend to make people describe themselves, which also didn't pick up.
The third and last class project was inspired by the British street artist Banksy. A handful of photos were put together crudely using Google Street View.
"We didn't want to trick people. We wanted to spur the discussion," said Lehmann. "And we knew there was Banksy artwork already in Boston and were hoping people would rediscover those."
The class in Copenhagen did little to influence Twitter users in Boston with their #banksyinboston hashtag with their four week experiment, said Lehmann. But they did demonstrate how a few dedicated bots can make more of an impact than a single individual with the similar amount of followers.
"If you wanted to coordinate the launch of a movie or create buzz on the down low, this seems like a potentially really powerful thing that's really difficult to detect," he said. Anyone with the expertise could also use the network of "sleeper bots" to spread misinformation at any given moment in the news cycle.
"We are, to an increasing degree, influenced by robots getting us to do more all the time," said Lehmann. "We hope that this little experiment can be helpful in creating awareness of such subtle manipulations before they begin shaping our public conversations."