Shared from Mashable
There is perhaps no more worthlessly modern pastime than yelling at a politicized bot on Twitter, so it’s fortunate that a pair of students at the University of California, Berkeley, have built a tool to help us figure out which propaganda-spewing accounts aren’t real.
Rohan Phadte and Ash Bhat, founders of Robhat Labs (solid name choice), built a site called botcheck.me. First reported by Wired, their creation lets users type in a Twitter handle, and from there it’ll tell you whether that account does a lot of things that indicate it’s probably a bot.
The duo initially went out in search of a way to combat fake news on Facebook, but Bhat said a lot of their digging into fake accounts there led them to fake accounts on Twitter. From there, they started analyzing traits of what they call “high-confidence bot accounts.” They were in search of politicized bots, not other types, and found common traits among these propaganda accounts, such as:
“tweeting every few minutes in a full day, endorsing polarizing political propaganda (including fake news), obtaining a large follower account in a relatively small time span, and constant retweeting/promoting other high-confidence bot accounts…”
Phadte and Bhat fed the algorithm behind botcheck.me with a bunch of those accounts and accounts of regular users, so the algorithm would learn to tell the difference. Now, they say botcheck.me can identify a “high-confidence bot account” with 93.5 percent accuracy.
Their work has shown them how bots have evolved in the past year or so. At first, Bhat said a lot of these accounts just tweeted images, but they’ve grown into accounts that tweet “articles” with text generated by the headline of whatever they’re linking to, so Twitter users won’t be tipped off by bot-mangled language.
They’ve also seen how real accounts can transform into bots. A lot of us probably follow the account of someone we know who never used Twitter that much, and eventually stopped altogether. Those accounts sometimes get hacked and sold to someone who then turns them into a bot account, meaning Twitter users can follow bots without ever having meant to.
The pair have also found some weird shit: One time, scrolling through a conversation between two probable bots, they saw one account imply it had bought a car on Monday and then decided to buy a different one on Tuesday.
“If you go far back enough, the conversations that they’re having between themselves you’ll notice are just gibberish,” Bhat said.
Twitter, Facebook and other huge tech companies are dealing with something of a PR shitstorm of late as Congress tries to figure out just how the Russian government influenced the American public by manipulating social media ahead of the 2016 presidential election. Russian actors set up bot networks designed to tweet inflammatory propaganda aimed at users in the U.S. and elsewhere.
Twitter says it’s found 36,000 bot accounts with ties to Moscow, though Virginia Sen. Mark Warner said he’s “concerned that Twitter seems to be vastly underestimating the number of fake accounts and bots pushing disinformation.”
In addition to Botcheck.me, Phadte and Bhat built a Chrome extension that adds a blue bot-check button to every Twitter account a user sees, so users can verify the authenticity of an account right there.
Such a tool is undeniably useful to users, though as Bhat alluded to, it’d be more useful if Twitter would do something similar. Twitter, he said, should be verifying the accounts on its platform, not users. In doing so, the company would build authenticity into its product, perhaps making the Twitter experience a little less like diving into a screamfest.
Original Article and Images from Mashable