Today we return to “TrudeauMustGo” as well as the three other candidate collections. We seek to update readers on the status of the election interference built on circulating this rhyming trope as well as the preponderance of purported MAGA accounts active in the Canadian election.
As of October 15, we had collected ~525,000 Tweets using the Twitter Search API and the search string TrudeauMustGo. These data include mentions of the phrase in the original text (roughly 50% of the data), but also in the retweets, replies, and user_descriptions as well. Over 103,000 Tweets in this collection (>19%) are replies.
MAGA, MCGA, KAG, Bot, Trump and Fellow Travellors in the Twitter Bios
Within this collection we isolated 4,171 Tweets with “trudeaumustgo” in the user_description. We found 103 distinct user-description data points and searched for “MAGA OR KAG OR Trump OR MCGA OR Bot” thereby revealing 25 out of 103 user_descriptions.
Why Such A Heavy Dose of Pro-Trump Accounts in the Broader Canadian Election Data?
The big question for us remains: how or why are so many virulently pro-Trump accounts on Twitter taking such an active interest in the Canadian election? Is it just a network effect? Or is it perhaps a transnational social movement by hyper-nationalists who misunderstand the logical implications of their own professed identity? Is this election engineering by foreign governments? Are the major actors coordinated or diffused?
Irrespective of the mutually reinforcing and multifaceted answers to each of those questions, it is the data itself that speaks louder than any all-encompassing theory of how the fragmented Internet works.
Let’s review more account descriptions; they are key primary data points with no established analytical framework for verification. What does seeing this data say to you?
Digital Footprints of Viral Tweets
The most viral Tweets are those which accumulate a combination of the most retweets and the longest list of replies. We address the preponderance of replies in this data later in a separate post, but for now, let’s begin with duplicates.
In Twitter data, we use a duplicate detection algorithm (similar to plagiarism detection) that was developed for our scholarly work supporting the review of public comments in regulatory rulemaking, particularly when a form letter campaign uses Internet affordances to generate tens or hundreds of thousands of duplicate and near-duplicate comments.
As it turns out, this particular algorithm is extremely useful for examining Twitter datasets and the creation of purposive, high-value training sets for machine learning.
This short video displays the top 100 most viral Tweets in our collection of TrudeauMustGo data.
Viral Tweets are interesting for a many reasons. Take group 13, for example, that cites “Twitter’s head of site integrity” to the effect that the TrudeauMustGo collection is “organic and not driven by bots.”
Let’s pause to think about invoking the public statements to the press by the head of Trust and Safety for Twitter, who was in fact quoted by CTV on precisely this issue, arguing that “the company’s initial investigations into the #TrudeauMustGo hashtag have found no evidence of ‘substantial’ bot activity amplifying the hashtag, despite multiple researchers continuing to voice concerns about potential hashtag manipulation.”
How does this public message impact the ability of Canadians to think critically about the nature of the content they consume, or the identity of the accounts they follow and which follow them? Is the press doing enough to counter digital disinformation, or accepting the platform position that there is no cause for alarm?
How About that Buffalo Chronicle?
Along the way, we were asked by a CBC reporter to look at a “news” outlet called the Buffalo Chronicle. We pulled more than 13,000 Tweets with a link to buffalochronicle.com and found mentions of “Trudeau” in more than 6,000.
To get a flavor of the Tweets, see below:
As in a Part Seven, to better understand the composition of the Buffalo Chronicle Tweet network, we loaded the data into a local Neo4J graph database, and analyzed them using GraphXR from Kineviz. Via that method, we uncovered that among 3,385 distinct users, there were 17 accounts with >=50 Tweets. Among the 17, one familiar user, SusanIverach, is connected to 442 others and the one user who mentions SusanIverach is PurpleGoAway, which we suspect is a bridge to a little bot and troll network visualized here.
What is the Take Away?
It is easy to spiral out of control into this data. It leads many directions, has numerous causes, each worm hole feels bottomless, and findings really cannot be reduced to a table, a visualization, an algorithm, or, apparently, eight blog posts.
I am deeply disappointed that after briefing a CBC reporter for one month, they chose to run a story about a poll, instead of exposing important acts of transnational information warfare on the eve of an election many scholars agree is under attack.
As a student of journalism history, and a fan of muckraking in general, I think this is alarming and insufficiently reported.