When U.S. Sen. Lindsey Graham of South Carolina launched his bid for president in 2015, he was blasted by plenty of Twitter users.
“Lindsey Grahamnesty, Jeb (Bush), (Marco) Rubio, (John) McCain spend more time attacking conservatives than Dems,” tweeted @heyits_toby on July 12 of that year.
“@LindseyGrahamSC You SIR LINDSEY GRAHAM YOUR GOP HAS FAILED THIS COUNTRY! I SUPPORT TRUMP 2016,” was tweeted at the senator by both @Pati_Cooper and @KarenParker93 on July 13.
When Graham ended his campaign on Dec. 21, @Jeblary2016 posted, “#ChristmasWeek has started and we already have a present — @LindseyGrahamSC is out of the race.”
Long after the end of the 2016 election, those tweets and roughly 3 million others live on in a Clemson University database of tweets from Twitter accounts identified by U.S. officials as belonging to Russian trolls who attempted to influence U.S. politics.
Compiled by communications professor Darren Linvill and economics professor Patrick Warren, the trove of election-related tweets shines light on how the trolls — Russians with fake social media accounts who tried to stir up unrest — operated. The team’s research already has identified patterns in how the users of more than 3,000 fake accounts tried to influence U.S. voters, according to an academic paper now undergoing peer review.
“We know what and when” they tweeted, Warren said. “Now, we’re digging into the strategy. You think they’re going to do this, but then that’s not what they did.”
After months of studying the Twitter trove, the real work of analyzing the Russian strategy “is only getting started,” Linvill said.
The two professors created the database using Clemson University’s Social Media Listening Center, which uses powerful analytics software to monitor online activity across the web.
“It can scrape a huge amount of data off social media platforms — Facebook, Instagram and, lucky for us, Twitter,” Linvill said.
When the U.S. House intelligence committee last November released a list of Twitter accounts identified as being operated by Russia’s Internet Research Agency — a Russian government-run “troll factory” — Linvill and Warren realized Clemson’s media center could pull all of the accounts’ past activity, even after Twitter had deleted the accounts.
Their academic paper tracks Twitter activity from June 2015 to the end of 2017.
Making the database was a labor-intensive process. Clemson’s software only could download 50,000 tweets a day and the researchers had to go through them to exclude people they could identify, via research, as real Americans.
“Real people tweet like real people,” Linvill said. “They talk about their mom, something that happened in class, life events. Trolls won’t do this.”
Reviewing the tweets was time consuming. “I read a lot of tweets,” Linvill said, adding, for months, “our wives were troll widows.”
The researchers’ work is being noticed nationwide. The Washington Post made extensive use of Clemson’s findings in reporting on a burst of activity from Russian troll accounts about a month before the 2016 election. The same chart shows a second spike in August 2017.
“We don’t have a perfect story for that (August) spike, yet, but it is overwhelmingly right (wing) trolls,” Warren said in an email.