Ontario PC party leader Doug Ford’s Twitter messages from @fordnation are often amplified from a small army of hyper-partisan, largely anonymous accounts that publish with inhuman frequency.
The proliferation of mostly right-wing boosters for populist candidates and issues on Twitter and Facebook is suspected of shaping public opinion in Brexit, the 2016 British referendum in which citizens voted to leave the European Union, and in the election of Donald Trump as U.S. president later that year.
As Ontario nears the likely end of 15 years of Liberal Party rule, the role of social media in the June 7 provincial election campaign raises questions about what's real and what's not when it comes to political discourse on the Internet.
“Right now we’re in a very unsettling moment where it’s very hard to know what really is popular online and what could be the result of some form of astroturfing,” said Fenwick McKelvey, an assistant professor in communication studies at Concordia University in Montreal. He was referring to an artificial presentation of grassroots political mobilization.
While social media manipulation is not the sole preserve of those with right-wing ideologies, “the progressives don’t hold a candle” to the number or vigour of conservative amplifiers in Canada, according to John Gray, whose Mentionmapp Analytics tracks social network trends.
Amplification software likely used
Over five days in late April, Gray took nine snapshots of mentions and retweets of tweets on Ford and Liberal Premier Kathleen Wynne's Twitter accounts.
He found that one in five mentions or retweets of @fordnation were from accounts that posted more than 70 times a day on average. Gray said 70 is a conservative benchmark for output that could reasonably be assumed to have been aided by amplification software.
From each snapshot Gray grabbed the last 200 references to each account - 1800 in total for each. From the mentions of @fordnation he found 708 unique profiles.
Due to his own capacity constraints, Gray focused on the 254 of those which were not legitimate media or political party accounts and had tweeted at least twice in the April 21-25 period he studied. For Wynne he found 188 profiles.
He found a clear trend in which accounts posted both praise of Ford and criticism of Wynne, with many of the accounts also mentioning NDP leader Andrea Horwath (Ford’s nearest rival in more recent polls) in a negative way.
Wynne was "a magnet for the trolls," Gray said.
"It was almost proportionate, that for all the love Doug Ford got from his audience at the same time it was the number of profiles that were not favourable to Kathleen Wynne.”
Gray also found that the top 20 per cent of the most active accounts engaging with Doug Ford's account averaged about 156 tweets per day.
It is relatively easy for a motivated person to use software to automate their activity on social media, and many have legitimate reasons to do so. That makes it difficult to draw a line between so-called bots that do not represent a real person and amplification efforts of largely anonymous but otherwise real people that, if stifled, would constitute suppression of political speech.
No evidence of organized campaign tactic
Many of the accounts Gray identified were also amplifying messages from the likes of conservative media personality Ezra Levant, U.S. President Donald Trump, comedian Roseanne, and alt-right proponents Jack Posobiec and Mike Cernovic, his research showed.
“These profiles aren’t sharing anything else but politically charged content,” he said. “So they specifically exist either to support the base or to poke holes at (Canadian Prime Minister Justin) Trudeau, at Wynne, at the Democrats, at progressives, at the Liberals. It’s either praising Ford, praising (federal Conservative leader Andrew) Scheer, praising Trump, glorifying the right or denigrating the left.”
Gray recently went back to check on the profiles in his study and found that 10 appeared to have been deleted by their owners, six were suspended by Twitter and three were temporarily restricted. Twitter declined to comment on specific cases in which a profile had been disciplined, citing user privacy.
While there is no evidence linking these profiles to an organized campaign tactic, it is not the first time the Ford campaign has been linked to social media manipulation.
Concordia’s McKelvey and his colleague Elizabeth Dubois from the University of Ottawa studied a similar bot-like phenomenon in the Progressive Conservative leadership race in which “a few suspicious accounts” were found to have unsuccessfully sought to amplify the use of the #crookedchristine hashtag to support Ford and oppose Christine Elliott’s bid. McKelvey said one account was linked to a campaign staffer.
The "Crooked Christine" hashtag echoed the Trump campaign trail line about “Crooked Hillary,” which would often be followed by crowd chants of “Lock her up!”
The Ford campaign did not respond to a request for comment on whether it was involved in any amplification efforts in the provincial election campaign or the leadership race.
A recent report by U.S-based investigative news outlet ProPublica, in partnership with the CBC, also identified a number of groups running ads related to the campaign, without registering with Elections Ontario as third parties. These groups would only be required to register if they spent more than $500 in a campaign.
It wasn't clear whether any of these groups had spent more than $500, but several had promoted content that was critical of Ford, including one group running the www.notdoug.ca website.
Social media struggles
One of the most prolific Twitter accounts Gray found was @MousseauJim, whose profile picture shows a man smiling next to Ford. His biography says he is a ninth generation Canadian conservative and retired small business owner who supports the Ottawa Senators, opposes the imposition of Islamic law and believes that climate change is a scam.
“I have little doubt that MousseauJim is a real person, but he’s got a highly programmatic Twitter profile,” that averaged 350-400 hyper-partisan tweets daily over at least six months, Gray said. “Jim Mousseau is not sitting at his keyboard seven days a week churning out 400 tweets a day.”
But unless such accounts can be proven to have breached either Twitter’s terms of service or provincial electoral law, there is little to no justification for shutting them down, and neither Twitter nor Facebook have expressed interest in policing amplification conducted by otherwise legitimate accounts.
Twitter’s rules state that automating the reply and mention functions to reach many users on an unsolicited basis, for example by sending automatic replies based on keyword searches, is an abuse that is not permitted. But automation is allowed if recipients have clearly opted in to the conversation, are provided with an easy way out and the content is limited in volume. The rules on amplification were last updated in November 2017.
While Twitter’s open data policy makes it relatively easy for researchers to study the phenomenon, it is more difficult at Facebook, where there is more ambiguity between public and private online discussion.
“Twitter is an elite media,” said McKelvey. “I think it gets a lot of uptake because journalists are on Twitter and so they are aware that Twitter is important. I’m not sure that it is really important for the everyday voter.”
About 23 million Canadians are active on Facebook each month while 15 million use Twitter, according to the companies.
Facebook says it enforces strict community standards against bullying, harassment and hate speech and is also working hard to up its game on advertising transparency, both in the context of political campaigns and broadly.
Those efforts follow Facebook admitting that it sold thousands of politically divisive ads to Russian-backed accounts that were subsequently seen by some 126 million Americans during the 2016 campaign.
In Canada, it launched a feature late last year allowing users to view all the ads a specific group is running on Facebook, Instragram or its Messenger app, regardless of whether the viewer is part of the target audience, while in the United States it intends to begin clearly labeling political and issue ads.
It says it took down 837 million pieces of spam and disabled some 583 million fake accounts in the first three months of 2018.
But it is tricky terrain to navigate, made more difficult by the ease with which social media accounts can be created and reticence of the companies that run them to take decisive action that would cut down on a key metric investors use to value them.
“It’s in everybody’s interests to not think that social media analytics are fake,” said Concordia’s McKelvey. “It’s in the interests of the platforms to say the data is worthwhile. It’s in the interests of the politicians to look as though they are more attractive. It’s in the interests of the media to be able to have easy topics they can say are trending.”
Comments
Excellent article! More of this, please.