If you followed Rebecca Hirschfeld’s @Beckster319 account on Twitter in the weeks leading up to the 2016 election, you would have seen that she’s an actress, an obsessed fan of David Bowie, not so much of Donald Trump, and will eat anything pumpkin flavored.
Around the same time, if you looked at Markiya Franklin’s @internalmemer account, you’d figure she supports Black Lives Matter and is a diehard K-Pop fan. Chris Osborne’s @skatewake1994 account was busy tweeting about then-candidate Trump and Osborne’s passion for surfing and snowboarding.
Hirschfeld lives in Illinois, Franklin in Florida, and Osborne in California, but their account handles were among 2,752 that Twitter identified as potentially connected to Russia’s Internet Research Agency, and submitted to Congress in November 2017. In February, special counsel Robert Mueller indicted the IRA for engaging in “fraud and deceit for the purpose of interfering with the US political and electoral processes, including the presidential election of 2016.” The IRA is the Russian propaganda arm considered a “troll factory,” distinct from the military-intelligence officers Mueller indicted last week for hacking into Democrats’ computers in 2016.
The three Americans are among more than 20 Twitter accounts that appeared on Twitter’s list of suspected Russian accounts yet show signs of being real people, according to an analysis by Clemson University professors Darren Linvill and Patrick Warren. Of those, WIRED independently verified that at least four accounts were created by people with no demonstrable ties to Russia. Their handles were published by a congressional committee, identifying them in some minds as Russian agents.
Hirschfeld, Franklin, and Osborne were suspended from Twitter without warning last year and included in Twitter’s list with no explanation. They lost access to Twitter accounts that they used to maintain social and career connections. They only learned why they lost their accounts when contacted by the Clemson professors or for this article.
“I know that whatever I post online anybody can see,” Osborne says. But “that they could just take some random tweets from my account that is obviously not a Russian bot and place it under that label is concerning to say the least.”
Twitter representatives declined to comment about specific accounts, including why they were suspended or included on the list it gave to Congress. Last fall, Twitter’s acting general counsel, Sean J. Edgett, told the Senate Select Committee on Intelligence how Twitter sought to identify accounts tied to the IRA. The company looked at accounts that generated election-related content and examined whether an account was created in Russia, whether the account was associated with a Russian email or phone number, whether the user ever logged in from a Russian IP address, or whether the display name contained Cyrillic characters. Twitter relied in part on an algorithm, but its review also included “internal, manual reviews conducted by Twitter employees,” Edgett said.
Twitter’s process initially flagged 36,746 accounts it believed may have been tied to Russia, Edgett said. It winnowed that list before turning it over to Congress.
Hirschfeld, Franklin, and Osborne were in the US around the time of the 2016 election. All say they used Twitter almost exclusively from their phones, through American telecommunications providers that should have shown US IP addresses. Similarly, their email addresses were registered to American providers.
In cases where Twitter wasn’t sure if a user was real, Edgett said, it would challenge the user to verify a phone number or complete a captcha to prove that they were, in fact, human. Hirschfeld, Franklin, and Osborne all say they do not recall being presented with such a challenge.
‘That they could just take some random tweets from my account that is obviously not a Russian bot and place it under that label is concerning to say the least.’
Their inclusion in Twitter’s list of suspected IRA accounts raises questions about Twitter’s investigation into Russian interference, and the extent of congressional oversight. One potential hint that Twitter appears to have overlooked is the age of the accounts in question. Both Hirschfeld and Osborne first logged in to their accounts in February 2011, well before the first known IRA activity in 2013. WIRED’s analysis found at least 63 accounts created between 2009 and 2013 included in the list provided to Congress.
Twitter has since updated its list of potentially IRA-linked accounts. In January, it said it added 1,062 accounts, for a total of 3,814. Neither Twitter nor Congress released that list at the time. In June, after WIRED contacted Twitter about the possible inclusion of Americans on its list, Democrats on the House Permanent Select Committee on Intelligence released a revised list of 3,841 account handles it said Twitter had flagged as possibly linked to the IRA. That list omitted seven accounts from the November list, including those of Franklin, Hirschfeld, and Osborne, yet their accounts remain suspended. In its correspondence with Congress, Twitter said it deleted the accounts “based on additional information, refinements to methodology, and enhanced understanding of IRA characteristics.” Twitter declined additional comment.
WIRED identified Franklin, Hirschfeld, and Osborne by examining a database of tweets released in February by NBC, which had obtained it from sources familiar with Twitter’s data systems after Twitter removed the tweets from public view. WIRED supplemented that database by identifying mentions and replies from the accounts Twitter flagged. The analysis showed the three Americans tweeting with friends and relatives about everyday life, including upcoming exams, performances at Coachella, and kale recipes. By contrast, Twitter accounts associated with the IRA rarely display any glimpse of a personal life. They rapidly tweet and retweet popular posts, or bring attention to current political events to gain followers, according to Linvill and Warren, the Clemson professors, who are researching the usage patterns and strategies of the IRA’s Twitter accounts.
Linvill and Warren have analyzed a database of roughly 3 million tweets from the accounts on Twitter’s list. They’re looking for signs that Twitter erroneously included the accounts of real people to exclude those accounts from their analysis. “We don’t want to use real people in trying to find out how Russian Twitter trolls are acting,” Warren says. “We would get the wrong answers.”
Working independently of WIRED, Linvill and Warren also identified Hirschfeld’s @Beckster319 account as potentially a real person, based on the tweets. Her fondness for pumpkin is a common theme. “I judge the change of seasons by the availability of Pumpkin Spiced lattes. #LaborDayWeekend,” she tweeted on Sep. 2, 2016. She also tweeted regularly about her family: “Just witnessed my niece dole out stickers from her sticker album at storytime hour like the Godfather of Kids. #kids,” she tweeted on Oct. 27, 2016, less than two weeks before the election.
Hirschfeld says she used the account almost every day after creating it in 2011. At the time, she worked in Los Angeles as an actress; she has since moved to Chicago, where she teaches elementary school, but still acts. She says she used Twitter to stay on top of current events in the media and entertainment industry.
Hirschfeld had assumed that tweeting about the #MeToo movement might have led to her account’s suspension. “I was posting some stuff that happened to me in Hollywood,” she said. “I figured I must’ve done something to piss somebody off and they reported me and took me down.”
Although Hirschfeld tried several times to reach Twitter after her account was suspended, Twitter Support only provided access to a chatbot. “You can’t talk to anybody,” Hirschfeld said. “You just get the voicemail on these things.” Eventually, Hirschfeld gave up and created a new account. Twitter declined to comment.
Hirschfeld says she invested time and effort into her Twitter presence before it was suspended, and losing her virtual network could hurt her acting career. Her new profile, @Bexster319, has only 36 followers, down from more than the 800 she had on her @Beckster319 profile.
‘That was my property. You can’t take that away from someone. I want my account back.’
Linvill and Warren, the professors, suggested that Franklin’s and Osborne’s accounts might also be real people after analyzing their tweet patterns. In Franklin’s case, Linvill says, “Interspersed, every few days, in all of that retweeting was the occasional comment about her mom. I saw one tweet where she was clearly asking a friend in class about a quiz.”
Franklin is preparing for her sophomore year studying civil engineering at the University of South Florida. Last year she took calculus and engineering classes during the day, but spent most of her evenings on Twitter, churning out, she said “like 40 tweets per day.” She started her first account, @ibeepepe in 2015, which focused largely on K-Pop. Shortly after, she started a separate @internalmemer account to connect with high school friends. There, she posted thoughts on the subjects that mattered to her, “like Black Lives Matter, anything about LGBTQ,” she said.
Osborne recently graduated from Point Loma Nazarene University in San Diego with a degree in political science and plans to join the Navy as a pilot. In 2017, he interned for US Representative Devin Nunes (R-California), the chair of the House intelligence committee. Democrats on that committee released Twitter’s list of suspected IRA accounts; they did so without consulting Republican members, according to a committee spokesperson.
Another American whose Twitter handle appears on the list is Robert Delaware (@RobbyDelaware), who lives in the former Soviet republic of Georgia. Following previous reporting of his account suspension by Vice in November, Twitter reactivated his account. He never received a response from Twitter about the issue, but he says, “I was just happy that my account was visible.” Representatives from Twitter refused to comment about any individual account, including Delaware’s.
In all, Linvill and Warren say they have identified more than 20 accounts on Twitter’s list that show patterns of being real people.
Twitter compiled the list, but Jameel Jaffer, executive director of the Knight First Amendment Institute, questions whether members of Congress “took the steps [they] ought to have taken before publishing information that [they] should have known would be deeply damaging to the people who were named.”
Jaffer says identifying the accounts as possibly linked to the IRA highlights the need for “digital due process,” and raises questions of how companies like Twitter support or stymie internet speech. “There is an important discussion to be had about whether social media platforms should be providing due process to individuals whose accounts are taken down,” he said.
Typically, Jaffer says, account suspensions don’t lead to lawsuits, since to be successful a user would have to demonstrate “some other harm that you have suffered because of your blacklisting.” In a prior case involving individuals placed on a screening list at airports, the plaintiffs couldn’t point to a tangible harm other than “stigma,” so the case was thrown out.
Affected users, however, see their rights as broader than those spelled out by Twitter’s Terms of Service, which specifies that the company has rights to “remove or refuse to distribute any Content on the Services, suspend or terminate users, and reclaim usernames.”
“That was my property,” Hirschfeld says of her account. “You can’t take that away from someone. I want my account back.”
For her part, Franklin is curious about her inclusion on the list. “I just want to know why. Did they just use an algorithm and flag people as trolls?” she says. “I don’t understand.”