In Part One, I laid out the case of the intriguing stink of foreign campaign intervention. This was based on one or two slices of Twitter data taken from 400,000+ #TrudeauMustGo tweets.
This particular social media “campaign” is dominated both by hyperactive users, as well as users whose bios are pro-MAGA and pro-Trump and contain buzzwords of a Twitter-fueled democratic information apocalypse.
Every day, a deeper dive into this data suggests a clear and vexing threat to authentic political communication.
Data Collection for This Series of Posts
On September 9, 2019, we started fetching tweets using the Twitter Search API. For this purpose, we used the following keywords:
The first keyword is a suspicious hashtag, and the other three are the primary candidate Twitter handles.
Twitter fetches continue every 15 minutes at the time of this writing, though there were periodic interruptions for 429 error rate limiting.
The TimeTrack graph below shows the distribution of the “TrudeauMustGo” tweets over time.
Please note that the Twitter Search API returns matches when the text string appears anywhere in the body of the tweet or in the many fields of metadata. As a practical matter, we collect a tweet even if the text string match for TrudeauMustGo is not in the body of the tweet itself.
In fact, only 47.7% of the collection has the phrase “TrudeauMustGo” in the tweet. This means that the phrase match is in the metadata of more than half the collection.
More Findings from the Trudeaumustgo Data
We periodically check the corpus for duplicates, which provide a roadmap to the most viral tweets in a collection.
Based on the first 352,000 tweets, below are the 20 most viral tweets in a rank order list with their counts.
One notable feature of the table above is that for a collection of more than 350,000 tweets (at the time of de-duplication) covering a viral hashtag, none of the top 20 groups are that big.
Rather, we see a long list of semi-viral tweets instead. There are more than 100 groups with over 200 retweets, and the most retweeted username is an all out pro-Trump agitator. The username in 15 of the 100 largest groups is a “grassroots” campaign called “Canada Proud”, which does not give much information about its organization.
One simple way to visually inspect the (now) 27,500+ user descriptions in this collection is to export the metadata field and sort it alphabetically in a spreadsheet. Some patterns will jump out at you, such as:
- The use and nature of a quotation appearing first,
- The use of hashtags appearing first,
- The preponderance of Trump, MAGA, and America First rhetoric,
- As many #Benghazi first as #Bernier first (just 3).
The four bullet points above are from the review of 309 rows out of 27,500.
To make the point further, when you scan through the user descriptions, other patterns jump off the screen:
Further Investigation of Tweets on Canadian Elections
Whether this is the “normal” rate of participation by #MAGA movement faithful in a Canadian election remains to be explored. Perhaps this is the new normal.
Maybe the Internet and social media platforms destroy older notions of national identity even in a movement that is all about looking inward.
For now, I doubt it. There is a lot more happening here.
In Part Three, we will continue investigating the TrudeauMustGo campaign. We will also take a closer look at the role of Twitter accounts involved in the dissemination of Russian propaganda in English, and of information about the three major Canadian candidates.