Research: Swing States Biggest Target for ‘Fake News’ Tweets in 2016 Election

Oxford University researches have found over 780,000 tweets seeking to spread misinformation, coax groups, and persuade voters on ballot casting.

by Greg Gordon, Peter Stone, McClatchy Washington Bureau / September 28, 2017

(TNS) -- WASHINGTON — Voters in 11 swing states in last year’s presidential race, including Florida, Ohio and Pennsylvania, received more fake, junk and hyper-partisan information over Twitter than reliable, professionally produced news in the 10 days before the election, according to a British study of the social media platform’s potential impact.

The analysis by researchers at Oxford University of about 781,000 tweets provides fresh evidence that entities seeking to spread misinformation used social media platforms as a powerful tool not only to distribute phony or misleading information, but also to direct it to voters in key jurisdictions in attempts to coax some groups to cast ballots and dissuade others from doing so.

Nationwide, an average of 25 percent of election-related tweets contained material from established news organizations, the researchers at the Oxford Internet Institute wrote. However, they said, “a worryingly large proportion” of tweets about the election — as much as 57 percent in West Virginia — came from junk news sources, as well as Russia’s state-owned television network and the transparency group WikiLeaks, which published unverified stories.

In the election’s three decisive states, the researchers found that fake and junk news constituted 40 percent of the sampled election-related tweets that went to Pennsylvanians, 34 percent to Michigan voters and 30 percent to those in Wisconsin. In other swing states, the figure reached 42 percent in Missouri, 41 percent in Florida, 40 percent in North Carolina, 38 percent in Colorado and 35 percent in Ohio.

The research findings could be controversial given the wide range of content on the internet and the difficulty in tracking the sources of tweets, which are limited to 140 characters. But they appear to represent the most extensive independent analysis to date of the potential impact of polarizing social media content on the tumultuous campaign that ended with Republican Donald Trump’s stunning upset victory over Democrat Hillary Clinton.

“It turns out that junk news was concentrated in swing states in a measurable way,” said lead researcher Philip Howard, an Oxford professor of internet studies.

“The states that were safe, that were clearly going to go for Clinton or for Trump, didn’t get as much misinformation as where the races were really close.”

Release of the Oxford study coincides with a scheduled briefing Thursday at which investigators for the House and Senate Intelligence committees hope representatives of California-based Twitter will reveal what they know about Russia’s use of their platform as part of a broad cyberattack aimed at boosting Trump’s prospects and whether they took any steps to curb it.

Virginia Sen. Mark Warner, the ranking Democrat on the Senate Intelligence panel, has reviewed the Oxford study.

“We already know that Russia sought to hijack platforms like Facebook in order to disrupt our elections,” he said in a statement to McClatchy. “And the findings from this study raise further questions about the extent to which fake news and polarizing content may have been targeted towards our most competitive states. In the weeks ahead, I hope that Twitter will be fully forthcoming with our committee so that we can inform the American public about what happened in 2016, and make sure that it doesn’t happen again.”

California Rep. Adam Schiff, the ranking Democrat on the House Intelligence Committee, said the Oxford report confirms that social media platforms at the time were “rife with viral reports that were highly partisan, conspiratorial or simply false.” As a result, he said, millions of Americans were “easy prey on Facebook and Twitter for propagandists, peddlers of false news and hostile foreign powers such as Russia.”

In a June blog post Colin Crowell, Twitter’s vice president of public policy, government and philanthropy, wrote that “the firm’s “open and real-time nature is a powerful antidote to the spreading of all types of false information.”

“We cannot distinguish whether every single Tweet from every person is truthful or not. We, as a company, should not be the arbiter of truth.”

Crowell said Twitter strictly prohibits the use of robotic-like commands, known as “bots.” However, federal investigators believe Russians used bots to grab fake and harshly critical news about Clinton, some from far-right websites, and spread it across the internet via phony Twitter accounts.

Crowell also said Twitter bars “other networks of manipulation to undermine the core functionality of our service.”

The company has been “doubling down,” he said, in efforts to detect and suspend accounts engaged in suspicious or duplicative activity.

In a statement to McClatchy, Twitter said that outside research about the impact of bots and misinformation on Twitter through the limited data the firm makes public “is almost always inaccurate and methodologically flawed."

The company said the Oxford analysis did not appear to undergo peer review to ensure it was “authoritative and empirically sound.”

The researchers’ findings, however, lend credence to assertions that fake and hyper-partisan content on social media may have confused many voters, including extremist and false stories saying that Pope Francis had endorsed Trump, that Clinton had Parkinson’s disease and that she ran a pedophile ring from the basement of a Washington pizzeria.

Their analysis of tweets flooding the internet and cellular networks is by definition inexact. The researchers said they traced the locations of a statistically significant sample of 781,087 tweet recipients from the 22 million messages Twitter made public for the last 10 days of the campaign — still just 1 percent of its overall traffic. All tweets that were categorized used election-related hashtags.

Unlike Facebook, which enables clients to target ads to calibrated groups of people, tweets are less easily targeted. But Howard said users find crafty ways to do so, usually by using a geographic hashtag. For example, someone aiming for recipients in Ohio might have posted: “Hillary is corrupt” and used the hashtag #OhioMAGA (for Make America Great Again, Trump’s slogan).

Howard said the researchers traced about 3 percent of the tweets to Russia Today, the Kremlin-backed TV network, but that Russian involvement was much higher. Bots sent by Kremlin operatives contain no identifying information.

Some tweets carried misinformation from Facebook pages. Howard said others linked to YouTube videos that had been repackaged by Russia Today so that it is less obvious that it was produced by a foreign power.

“I feel quite comfortable saying this is about social media platforms broadly,” Howard said. “There’s a lot of cross-fertilization between Twitter and Facebook.”

YouTube is owned by internet giant Google Inc., which also is facing congressional scrutiny. Google has said it has seen no evidence that Russian operatives conducted a targeted ad campaign on its platforms. The videos by Russia Today, however, do not appear to have been paid ads.

Revelations about Russia’s use of social media networks during the 2016 campaign already have brought calls for tougher disclosure requirements so that recipients of political information will know where it came from. The Federal Election Commission voted this month to reopen the public comment period about possible new disclosure requirements covering the internet.

The intelligence committees, along with Justice Department Special Counsel Robert Mueller, are conducting parallel inquiries into the Kremlin’s massive cyberattacks aimed at boosting Trump’s election prospects and whether Trump’s campaign or other associates collaborated with the Russians.

McClatchy reported exclusively in July that the investigations are exploring whether the Trump campaign’s digital operation colluded with Russia’s disruptive cyber operations. That includes Russia’s use of bots to amplify fake and critical stories about Clinton from far-right news sites, such as Breitbart News and InfoWars.

Facebook has agreed to provide each of the investigators with copies of 3,000 digital ads that it traced to a Russian firm that has been tied to a “troll farm,” which has spread election-related propaganda to the United States and other western countries. Twitter says it is cooperating with the investigations, but has yet to provide specifics.

©2017 McClatchy Washington Bureau Distributed by Tribune Content Agency, LLC.