A few days ago Donie O’Sullivan and Dylan Byers of CNN broke a story about how a specific Facebook and Twitter account (@blacktivists) had been operated by the Russian government to increase racial tensions in the lead up to the election.
https://twitter.com/donie/status/913571130975252480
The accounts have been suspended on both Twitter and Facebook, and can no longer be viewed. There don’t seem to be any public snapshots of the content either, apart from 2 tweets that Justin Littman, of the Social Feed Manager project, was able to find in an archive of election related tweets. A recent Guardian article links to an “archived” Blactivists page, but unfortunately the URL points to the Google Cache, which is now a 404 Not Found. It’s 2017, shouldn’t there already be practices for archiving web source material at newspapers like the Guardian?
Besides the shared handle, the actual linkage between the accounts was not divulged:
The Blacktivist Facebook account was among the 470 Russian-linked accounts identified by the social media network and disclosed to Congress earlier this month, the sources said. The matching Twitter account was among the roughly 200 accounts Twitter identified with links to those found by Facebook.
Facebook shared its findings about the accounts with Twitter, enabling Twitter to identify 22 matching accounts and an additional 179 accounts that linked back to those accounts, the sources said. This matching process went beyond public-facing similarities and included private information that could link the accounts.
As Luke Barnes of @ThinkProgress writes, these 470 accounts are likely to be only the tip of the iceberg. But it’s interesting that the only one that wasleaked out was the @blacktivists account. This issue is of such importance to our understanding of political discourse in the United States that I would hope the advertisements and their accounts will be made public. Let’s hope that Facebook, Twitter and Alphabet choose to attend the public hearing that has been called for, so that specific details can emerge, and we can collectively try to understand what happened and how to prevent it in the future.
As part of our work on the Documenting the Now project we have been creating archival collections of social media related to the 2014 protests in Ferguson as well as the larger #BlackLivesMatter movement. Specifically we have a dataset of 13,732,829 tweets that used the #blacklivesmatter hashtag between January 2016 and March 2017. Bergis Jules, our Community Lead, thought it could be interesting to take a look in this dataset to see if the @blacktivists user appears there. Sure enough, we found 90 tweets sent by @blacktivists that used the #blacklivesmatter hashtag.
Here are a few interesting things we noticed. Please keep in mind that we were only able to analyze the tweets sent by blacktivists that also happened to use the #blacklivesmatter hashtag.
As mentioned in the CNN article there was a fairly consistent use of the wrong apostrophe: “it`s” instead of “it’s.” There were 14 tweets that used the ` and 4 that used the ’. Perhaps this indicates multiple people authoring the tweets?
98% of the tweets included a photo, all of which have been flagged as unavailable in Twitter’s image cache (HTTP Status 403). For example: http://pbs.twimg.com/media/CrYoXmaWIAATfwm.jpg
18 of the tweets contained external links to Facebook. All of them were to Facebook posts, events, pictures or videos belonging to the Blacktivists account. Interestingly 11 of these links used a Google URL Shortener. So potentially there is a Google identity here that could be linked to the trail. Here is an example of one of these redirect paths:
https://t.co/S0dftVR0qL
→ https://goo.gl/Af5AO6
→ https://www.facebook.com/blacktivists/posts/293845137625256
Probably one of the more interesting statistics about these 90 tweets was the times of day that the tweets were sent:
90% of the tweets were sent between 17:00 and 03:00 UTC, or 8PM and 6AM Moscow time. The timing appears organic rather than bot driven. The user profile metadata in each tweet claims that the user is on Pacific Time. Of course this could be easily faked, but that would translate the time range into 9AM to 7PM PT, which seems a lot more reasonable for a person who likes to sleep during the night. This isn’t your run of the mill bot churning out retweets. In fact there are no retweets in this set of 90. If this is orchestrated from Russia it shows a high degree of sophistication.
Another interesting pattern can be found in the number of followers and friends that the account has.
We were collecting the data from Twitter’s streaming API. One of the nice side effects of that is that you get a snapshot of a users friend and follower counts at that point in time. You can see that the number of @blacktivists followers is growing over time, but the number of users they follow (their friends) is jumping around a lot. They start out following lots of users, and then quickly drop them down as they acquire followers. Then periodically they rapidly increase their friends only to drop them down again. Being able to add and remove a thousand followers rapidly like that seems like automated behavior. The hand crafted nature of these tweets with these fingerprints of automation point to a cyborg type of setup.
As we learned during the blocking of Politwoops by Twitter for a breach of its Terms of Service, Twitter doesn’t want deleted tweets to be published on the web. Honoring user intent is a key and important issue when it comes to the ethics of archiving social media. But in a wise move Twitter decided to reverse this decision and grant Politwoops the ability to perform this public service. These aren’t just any users — they are elected public figures. Their tweets (for better or worse) are part of the public record. The Documenting the Now project thinks that these blacktivists tweets, given the claim that is being made about them, also deserve to be made available. So we, like the Social Feed Manager project, have decided to release these tweets as a Google spreadsheet for others to analyze. Let us know what you find.
P.S. If you are a journalist and are interested in archival practices for web content please get in touch! [email protected].