Twitter still awash in fake news, study finds
Twitter is still rife with “fake news” a month out from the 2018 midterms, with more than 80 percent of the accounts that regularly spread misinformation in 2016 still active, a new study released Thursday found.
The report — sponsored by the nonprofit Knight Foundation and written by George Washington University professor Matthew Hindman and Vlad Barash, the science director at network analysis firm Graphika — found that those accounts together still publish upward of a million tweets a day.
“Fake news and disinformation are still very much relevant today,” Barash told POLITICO.
The Knight Foundation study examined more than 700,000 Twitter accounts, which linked to more than 600 sites spreading misinformation or conspiracy stories. The researchers compared accounts that were active in the month before the 2016 election to those tweeting from mid-March to mid-April 2017. They then rechecked the sample in September — after a summer crackdown by Twitter that included banning Infowars head Alex Jones from the platform — to verify that the accounts they were scrutinizing were still active and spreading misinformation.
“Twitter has absolutely taken some measures to take some sites down, including removing Alex Jones from the platform, but they have not taken the vast majority of what we looked at down,” Barash said.
In a statement, Del Harvey, Twitter’s global vice president of trust and safety, pushed back on the findings. She said that because the data were accessed through Twitter’s publicly accessible interface, the researchers could not see when Twitter had prevented other users from seeing certain “automated or spammy” content and accounts.
“We do this proactively and at scale, every single day,” Harvey said. “As a uniquely open service, Twitter is a vital source of real-time antidote to day-to-day falsehoods.”
She added that the company works “diligently to ensure we are showing people context and a diverse range of perspectives as they engage in civic debate on our service.”
The Knight Foundation study comes as news organizations are under heightened scrutiny over what they’re doing to counter fake news before next month’s midterms and about two weeks after another report from Stanford and New York University, which found that Facebook had been more effective than Twitter at curbing misinformation in the aftermath of the 2016 election.
Despite attention to the role of Russian social media accounts in spreading false information in 2016, the study found that non-Russian-linked accounts were likely more important in the proliferation of misinformation on Twitter.
“Most of the content in the space is American,” Barash said. “Russian assets may have taken advantage of existing fissures in our society, but they did not engineer this whole [problem].”
One of the Knight Foundation study’s principle findings is that much of the deliberately false information on Twitter comes from a concentrated group of sites and accounts. The study used a list of sites that researchers have deemed disinformation to classify what links were considered fake news. The authors found that just 10 sites were responsible for 65 percent of the fake news links that appeared on Twitter during the 2016 election campaign.
One of those, the conspiracy site The Real Strategy, which helped to spread the Pizzagate hoax, has since been blocked from Twitter. The authors wrote that in the pre-election period they studied, the site received more than 700,000 links from Twitter, the second most of any of the sites they looked at. After it was blocked, that number dropped to 1,534 links.
Most of the other top fake news sites remain active, Barash said, although the researchers did not identify them because they want to continue monitoring them for future study.
Barash said that solving the fake news problem isn’t as simple as shutting down Twitter accounts or blocking sites, since others would pop up to take their place. He said that whack-a-mole approach would help to some degree, but he’d also like to see Twitter adopt a “broader approach” to “educate or inform” the public about misinformation. He cited programs like that used by Facebook to flag stories marked false by independent fact-checkers.
He added that the study focused on Twitter because of the accessibility of its data. Other platforms, such as Facebook, do not make such data public.
The Knight Foundation study also found that Twitter accounts posting fake news tend to heavily follow each other, leading to an “ultra-dense core” through which links to misinformation echo and project outward.
Many of those accounts are bots, the study said, with 33 percent of the 100 most-followed accounts in the study showing evidence of being automated. More than 60 percent of a random sample of accounts were bots, the study found.