Pew: Bots are churning out the majority of Twitter links

Several times a day, a Twitter bot shares views of the earth from space. The botmaker, Jeremy Low, is based in West Philly, but the animated images that his @himawari8bot handle spits out come from Japanese satellites. Through swirling clouds, users can spot the southern tip of Taiwan or a swath of Australia.

A new Pew Research Center report charts the far-reaching spread of automated tweeting software such as Low's. The study, released Monday morning, observed that two thirds of Twitter links to popular sites came from bots. Analyzing data from the summer of 2017,  Pew researchers found that the 500 most prolific bot accounts during this period had tweeted more than 20 percent of news links on the microblogging platform.

While previous research has often focused on the role of Twitter bots in news dissemination, Pew researchers found that popularity of bots is truly cross-sector. Bot sharing of news links was merely average when looking at bot linking overall. But bots heavily tweeted links to porn and sports. Nine of out 10 Twitter links to prominent adult sites had been automated, researches found, while more than three in four sports links had been.

Aaron Smith, associate director of research at Pew, said that although Twitter bots might be viewed as "niche phenomenon," they shouldn't be.

"Whether good, bad or neutral, bots are an extremely popular way that links to material get disseminated," Smith said. "It makes it a very fertile ground for doing all kinds of automated accounts, whether they want to tweet links, or correct grammar, or [create] calculations for sporting events."

Twitter bot-building can be learned through online tutorials, which is how Low got started. Public accounts are the norm on Twitter. Researchers from Brown and Northeastern found that in early 2013, only 5 percent of accounts on the site were private. There are Twitter bots that tweet countless color swatches by html codes, generate made-up think-piece headlines, and synthesize Census data into storytelling about everyday people.

Low, who has made a series of bots, was attracted to the experimentation he saw on the site. "There's no real pressure to making them," he said. He doesn't think that bots are devoid of humanity; rather, they reflect the agenda of the person who designed them. He said, "You have an idea, and then you work something up, and then if it's funny, or poignant, or if somebody likes it, then you do it and it's out there in the world."

Twitter cracked down on bots in February, announcing that the company would be weeding out automated accounts that posted duplicates, unless the focus was on "weather, emergency, or other public service announcements." This followed a January company news release disclosing that in 2016, more than 50,000 bots tweeting about the election in the weeks before Trump's win had been "Russia-linked."

The purge reportedly impacted thousands of accounts. Twitter's general counsel told the U.S. Senate Judiciary Committee in January that hundreds of thousands had been removed in 2017.

Emilio Ferrara, a computer science professor at the University of Southern California, said that although the research corresponded to a particular "moment in time," the findings raise concerns about pervasiveness of "social media manipulation." Bots essentially allow older tactics for promotion and propaganda to be scaled up beyond traditional capacities, he noted. While a canvasser might walk one stretch of pavement at a time, Ferrara said, "machines do not have respect to these human limits."

Analyzing links to news sites, Pew researchers found that most popular bot-shared reporting didn't swing to the far-left or right. Links to publications that could be labeled moderate, such as the Chicago Sun-Times, were more popular among bots than links to the firmly conservative National Review. While more than half of links to Fox News appeared to come from automation, upward of three-quarters of links to Business Insider and Forbes, which the study classified as center-left, came from bots.

Researchers note that it's too early to define the cultural impact of so many tweets lacking a human touch.

Smith noted that Pew's study is only a start at understanding the societal impacts of so many inhuman tweets: "There's a lot more work to be done to start to imagine those types of questions."