In Part Five of the series, we admitted that when studying a social virus, it is “difficult to place content (text, images, videos, news, identities) spreading on Twitter accurately into categories that are tied to the account history, such as human-vs-machine, manual-vs-automatic, real-vs-fake.” Difficult does not mean impossible and some errors along the way do not invalidate the research journey. So on we go.
Now in Part Six, we explore the largest 1-day spike of the most active day since we began collecting data for this investigation on September 7, 2019. In Part Seven, we introduce the most active hour in the most active day looking for more digital fingerprints of bots and trolls in the imminent Canadian election.
Ongoing Twitter Data Collection
We continue to use DiscoverText to collect Twitter data from the Search API using four key election terms. As of October 12, the archives contained more than 2.25 million Tweets.
We previously reported the largest 1-day spike in our collection occured on September 30. The data presented here and throughout the “Bots, Trolls and Elections” series use display time signatures set to St. Petersburg, Russia. As in an earlier post, we drop out the “trudeaumustgo” data and focus instead on three candidate Twitter handles.
On the day in question, we gathered 109,888 Tweets produced by 65,368 screen_names with 56,524 user_names and 51,299 user_descriptions. This is a non-trivial mismatch. If you look at just one variable and not the other, or the interplay, you can easily misinterpret the data. The “Top 25” most active “screen_names” and “in_reply_to” values are reported here.
Top 25 Most Active “screen_names” in the Largest 1-Day Spike
Top 25 “in_reply_to” Accounts in the Largest 1-Day Spike
Red Flags in the Data
Some of the most active screen_names in the 1-day activity spike are familiar “faces” at this stage of the research. We have seen canukcookie and JackNationalist active in other slices of the data.
In this instance we see they are Tweeting with persistent regularity to each accumulate 91 Tweets in the 24-hour period. You can almost see the rhythm of a human heart beat in these graphs, though it is the rhythm of a troll or an auto-retweet software application.
Finally, we present some of the 51,299 user_descriptions in the biggest one-day spike. Once again it bears mentioning that this apparent transnational social movement, whereby nationalist identities simultaneously slam globalism while paradoxically poking the Canadian electorate to be more nationalist, like U.S. President Donald J. Trump.
In Part Seven, we will dive deeper into the biggest 1-hour spike in the data during the 24 hours described above.