Twitter Looks The Other Way As Trump’s Tweets Amplified By Artificial Network

How millions of inauthentic tweets disseminate Trump’s falsehoods

Image for post
Image for post
Twitter’s CEO, Jack Dorsey, and Twitter’s Legal, Policy and Trust & Safety Lead, Vijaya Gadde, meet with President Trump (April 2019)

TLDR: Thousands of MAGA accounts are engaging in coordinated inauthentic activity, churning out hundreds of millions of retweets of @realDonaldTrump & other conservative accounts

The New York Times recently reported on Facebook, Twitter and YouTube allowing President Trump to disseminate “statements about unproven coronavirus treatments” — violating each platform’s respective Terms of Service (ToS) in the process.

Despite Big Tech making very aggressive public statements (re: mitigating against platform manipulation), they aren’t taking down misinformation that is being spread by Donald Trump.

Specific to Twitter, this post aims to illustrate that not only is Jack Dorsey allowing President Trump to disseminate falsehoods that violate their ToS, but that he also is permitting an easily identifiable artificial amplification network to distort the public debate (in favor of conservative ideology).

There is only one plausible explanation for such glaring levels of negligence from Twitter executives: Twitter Is Donald Trump’s Bitch (oh, and he’s also number 1 on Facebook!)

In November 2019, I wrote a post titled “Draining The (Twitter) Swamp: How being no-platformed led me to the heart of flag-bearing nationalist types.” Below I have included a snippet from that post, highlighting more than 2K hyperactive tweeting accounts which surfaced via Social Forensics’ analysis of the @LisaMar91564392 account:

Hyperactive tweeting accounts, in this case, were arbitrarily determined to be accounts that, on average, had tweeted at least 100 times per day since being created. Tweeting 100 times per day requires tweeting every 14.4 minutes.. for 24 hours straight! Often — but not always — this type of activity can lead to the discovery of nefarious intent.

This post will piggyback on the “Draining The (Twitter) Swamp” one.

More specifically, Social Forensics — in collaboration with Graphistry and Project Domino — have collected (and summarized) more than 5.6 million tweets originating from the 2,188 hyperactive tweeting accounts.

Methodology

Account statuses for the 2,188 hyperactive tweeting accounts were checked prior to collecting tweets.

Here’s a status overview of the 2,188 accounts:

Status details for each account may be found here.

Twitter’s API (via GET statuses/user_timeline) returns up to 3,200 of an account’s most recent tweets.

Here’s a preview of the report we generated, summarizing activity of the 1,792 accounts (click/tap to enlarge image!):

The full report may be found here.

More than 81% of collected tweets were retweets — which isn’t surprising considering these accounts form the core backbone of a pro-MAGA artificial amplification network:

Suspended Accounts

Of the 2,188 hyperactive tweeting accounts, 13% (277 accounts) were suspended by Twitter at some point between November 2019 and April 2020.

While commendable that Twitter is enforcing their rules, the reality is they are not fully scrubbing their platform of activity/accounts that violate their ToS.

In aggregate, the 277 accounts churned out (at least) 26.4 million tweets prior to being suspended (data on suspended accounts, including usernames, may be found here).

Translated: It took more than 26 million tweets for Twitter’s systems to determine these 277 accounts were in violation of Twitter Rules (meanwhile, Twitter continues to tout their “proactive” approach)

Nearly 50% of the accounts (134) reflected having at least 5 thousand Followers (average Followers: 11,042; average Following: 9,619). 34 accounts (12%) reflected having at least 20 thousand Followers. There were even a couple accounts — including @ChatByCC (covered in great detail in this post) — that reflected having more than 300 thousand Followers.

In other words, the accounts suspended by Twitter were incredibly active (average tweet count exceeded 95 thousand per account) and reflected having substantial followings.

Reviewing the creation year of the accounts suspended by Twitter, it’s clear that recency plays a significant role in Twitter’s (likely automated) decision making/workflow (76% were created in either 2018 or 2019):

Accounts That No Longer Exist

Of the 2,188 hyperactive tweeting accounts, 5% (108) of accounts that were active back in November 2019, no longer existed as of April 15th, 2020.

In aggregate, the 108 accounts churned out (at least) 8.8 million tweets while their Twitter accounts existed (data on accounts that no longer exist, including usernames, may be found here).

Nearly 55% of the accounts (58) reflected having at least 5 thousand Followers (average Followers: 15,740; average Following: 13,577). 24 accounts (22%) reflected having at least 20 thousand Followers. There was even one account — @SoarOnTheWings — that reflected having more than 200 thousand Followers.

Here’s an old tweet from the account via the Wayback Machine:

Since we recorded the userid (4165642155) of the @SoarOnTheWings account, we can see the account has reemerged as “Constitutional Republic TEXT TRUMP 88022” — changing their username to @CRRJA5 (the account’s bio currently reads “RTed by the President”):

Changing usernames is a tactic commonly employed by nefarious actors attempting to cover their tracks.

Similar to the suspended accounts, the accounts that no longer exist were incredibly active (average tweet count exceeded 80 thousand per account) and reflected having substantial followings.

The same recency bias is present among the accounts that no longer exist:

Of the 108 no longer existing accounts, 52 (48%) have simply changed their usernames. Those may be found here (and below):

Here’s what the recency bias looks like when limited to the (56) accounts that no longer exist which did not simply change their usernames:

Active Accounts

As mentioned earlier, this public service research project involved collecting tweets from 1,792 accounts on April 15th, 2020. Said accounts were sourced in November 2019, as part of a hyperactive tweeting segment of accounts discovered by Social Forensics, in relation to our investigation into “Lisa the Nationalist for Canada” (@LisaMar91564392) — a cyborg account (partly automated, partly controlled by human) that continues to violate Twitter Rules.

Social Forensics later joined forces with with Graphistry and Project Domino to summarize the data being presented in this post.

Out of 5.6 million collected tweets, more than 81% (4.5 million) were retweets. The accounts being amplified by the inorganically connected (based on Following/Followers relationships), hyperactive tweeting segment of accounts, clearly point to inauthentic coordination, where the goal is to artificially amplify tweets that champion conservative viewpoints.

The most commonly retweeted account, not surprisingly, is President Trump (@realDonaldTrump) — who was retweeted 127,526 times.

Here’s a preview of the report we generated, summarizing the 50 most frequently retweeted accounts from our dataset (click/tap to enlarge image!):

A report highlighting accounts that were retweeted at least 20 times may be found here.

To be clear, we are not stating that Jack Posobiec (#18; 13,713 retweets), a correspondent and host for One America News Network (OAN) — or any of the other individuals on the list, for that matter — are personally behind the amplification of their tweets. We are simply pointing out that said accounts are benefiting from an easily identifiable artificial amplification network:

In addition to Posobiec, President Trump recently praised Ryan Fournier (founder of Students for Trump), another account making the 50 most frequently retweeted accounts (#21; 11,457 retweets):

It should be mentioned that Charlie Kirk (#4; 31,446 retweets) is the Chairman of Students for Trump. Kirk is also the founder of Turning Point USA (TPUSA), a “student movement for freedom, free markets, limited government.” In May 2019, Kirk created Turning Point Action (affiliated with TPUSA), a political action committee intended to target Democrats. Shortly thereafter, the newly formed organization purchased the assets of Students for Trump:

Among the 50 most frequently retweeted accounts, it’s worth calling attention to the fact that QAnon-focused accounts (green rows) appear the same number of times as U.S. government-affiliated accounts (orange rows). Effectively, this information operation seeks to co-mingle conspiracy theories with politics via leveraging a vast artificial amplification network.

Here’s a cycle through the QAnon-focused accounts:

For those unfamiliar with QAnon, this recent episode of The Open Mind is worth checking out.

Two of the most frequently amplified accounts — namely, @Education4Libs and @Western2A — have since been suspended by Twitter (combined, the accounts were retweeted nearly 15 thousand times).

In the case of @Western2A, the account was amplified more than 7 thousand times (7,485) across just 49 days (meaning that @Western2A, on average, was retweeted 153 times per day!).

Another pair of accounts appearing among the most frequently amplified accounts — both inauthentic MAGA accounts (@superyayadize and @Lrihendry) — have bios highlighting they have been retweeted by President Trump:

The @ChatByCC account, mentioned earlier in this post (and connected to the same network of accounts) was an anonymous MAGA account that Twitter suspended shortly after it was retweeted by President Trump in December 2019:

According to The Daily Dot’s Zachary Petrizzo, “CC” was a “a pro-Trump Twitter account” supposedly operated by Jacob Wohl.

Contextualizing The Operation’s Scale

The average number of days across the most recent tweets for the 1,792 accounts is 36.

In other words, reflect for a moment on a group of 1,792 Twitter accounts that have retweeted President Trump 128 thousand times across 36 days.

To be clear, that’s an estimation — in reality, the group of accounts first retweeted @realDonaldTrump on November 2nd, 2016 and most recently retweeted the President on April 15th, 2020 (when tweet data was collected).

Point here being, when viewing the retweet counts of the most frequently amplified accounts, it’s important to keep in mind this 36 day average (i.e. the counts reflect “recent” activity).

While we collected 5.6 million recent tweets from the 1,792 accounts, they have tweeted nearly 325 million times since being created.

President Trump retweets comprise 2.3% of our dataset (5.6 million tweets). Extrapolating that percentage across 325 million lifetime tweets implies this artificial amplification network has retweeted @realDonaldTrump more than 7.3 million times.

It’s startling to frame the scale of the operation in that context.

Conversely, we can frame things in terms of daily tweet output. For example, the 1,792 accounts increased their aggregate tweet count by nearly 52 million tweets between November 14th, 2019 and April 19th, 2020.

Translated: This artificial amplification network churned out, on average, more than 330 thousand daily tweets over the past several months

OK, So What’s Your Point?

Early on in this post, it was declared that Twitter Is Donald Trump’s Bitch. And while that statement may seem extreme, it is absolutely warranted given the actions (or lack thereof) Twitter has taken towards mitigating against platform manipulation/information operations.

Take Heshmat Alavi, for example. The account was proven to be a persona controlled by several MEK operatives (an Iranian political-militant organization). Yet, as was surfaced by The Intercept’s investigation, the White House has actually cited Heshmat Alavi as a source to justify foreign policy decisions:

President Trump even amplified a tweet from “Heshmat” last month:

Within a few hours of The Intercept’s report being published (June 2019), Twitter suspended the Twitter account of Heshmat Alavi. A week later, however, the vocal MAGA-supporting account mysteriously resurfaced and has seamlessly transitioned back into pushing propaganda.

Twitter’s decision to reinstate the account resulted in my posing of the following question in a post titled “State-Sponsored Twitter Accounts Pushing For War With Iran” (October 2019):

Below you can compare how organizations with editorial standards vs. Twitter reacted in response to The Intercept’s investigation into the account:

Twitter’s reaction — namely that the Alavi account was deemed a “credible use of pseudonymity” — negates the fact that that account violates a number of other Twitter Rules (for more on that, here’s a great thread from Marc Owen Jones).

When you couple things like the reinstatement of @HeshmatAlavi with the easily identifiable artificial amplification network outlined in this post, it feels an awful lot like it’s President Trump — and not Jack Dorsey — who is actually calling the shots at Twitter HQ.

Twitter apparently has no issues being Donald Trump’s bitch.. so long as it’s good for their shareholders. Unfortunately, what’s good for Twitter shareholders in this case, results in the erosion of democracies across the world. Dorsey (and his executive team) realize how much Twitter relies on President Trump’s usage, hence they opt to look the other way as Trump’s tweets are amplified by an artificial network.

In fact, Dorsey even follows many of the accounts that are frequently retweeted by the artificial amplification network — several examples are included below:

2020 U.S. Presidential Elections

Facebook’s mantra/strategy (deny, deny, deny) prior to begrudgingly acknowledging their role in the manipulation of the 2016 U.S. presidential elections (literally, they were forced to concede that Cambridge Analytica was a thing), is a playbook Twitter knows quite well.

Specific to recent Canadian and U.K. general elections, Social Forensics has documented platform manipulation/information operations that Twitter has publicly refuted, instead stating “significant” manipulation and/or “bot” activity was not detected on their end.

In most cases, these comments come from an anonymous source at Twitter.. sometimes a named Communications employee (who is simply parroting uninformed messaging; propaganda, if you will). In rare instances, these type of comments will come directly from Yoel Roth, Twitter’s Head of Site Integrity, and/or Vijaya Gadde, Twitter’s Legal, Policy and Trust & Safety Lead.

Point here being, active measures are taken to disassociate accountability from any individual/group of individuals employed at Twitter (same dynamic exists at Facebook, Google, Amazon, Microsoft, et al.).

Here’s the red pill: Far too much power — and far too little accountability — is currently centralized among a handful of tech billionaires (and other extremely wealthy Big Tech executives; Twitter’s Gadde, on paper, earned a cool $28 million when Twitter went public, for example)

The self-governance process of Big Tech is incredibly opaque; moreover, allowing them to grade their own own homework simply doesn’t work.

Eroding democracies across the world from the perspective of Facebook (and other Big Tech companies), historically, has been nothing more than one of the costs of doing business. While earnings are privatized, society continues to bear the burden of business models that are misaligned with societal well-being.

In the case of Twitter, while their governance process is opaque and inconsistent, they should at least be commended for taking a much more open approach to data — the dataset summarized in this post, for example, can be sourced by anyone via Twitter’s API combined with OSINT tools.

We are six months away from the U.S. presidential election, and it’s alarming that Twitter is allowing such easily identifiable amplification networks to run wild, polluting the public discourse and potentially shaping perceptions, which is exactly what political advertising is designed to do.

Unlike Facebook, where political advertising is allowed, Twitter “made the decision to stop all political advertising on Twitter globally” in late 2019:

Twitter’s CEO, Jack Dorsey, has largely been praised for Twitter’s stance that “political message reach should be earned, not bought.”

In reality, Twitter seems to have no problem with “political message reach” being bought, so long as it is purchased via platform manipulation services rather than Twitter advertising products.

Effectively, Twitter executives have granted the Trump Administration and Trump supporters use of their platform to run a massive, free political advertising campaign — said differently, Twitter Is Donald Trump’s Bitch.

It is worth calling attention to the fact that platform manipulation is by no means limited to one side of the political spectrum (i.e. Twitter similarly looks the other way while inauthentic Resistance accounts game their platform!):

Moreover, politically focused platform manipulation represents a small fraction of the issues faced by those tasked with ensuring Twitter provides an accurate reflection of public sentiment, globally.

Bottom line: any individual or organization — and with relative ease/limited funding required — has the ability to morph Twitter into a conduit for propaganda

Geoff Golberg is an NYC-based entrepreneur who is fascinated by graph visualization/network analysis — more specifically, when applied to social networks and blockchain activity. His experience spans structured finance, ad tech, and digital marketing/customer acquisition, both at startups and public companies.

Geoff is the Founder/CEO/Janitor of Social Forensics, where he spends his time developing techniques and building tools to identify social media manipulation.

Social Forensics collaborated with Graphistry and Project Domino on this research.

Written by

CEO & Founder, Social Forensics | Previously: Co-Founder, Elementus | Featured in BBC, CNN, BuzzFeed, and Quartz, among others | SocialForensics.com

Get the Medium app

A button that says 'Download on the App Store', and if clicked it will lead you to the iOS App store
A button that says 'Get it on, Google Play', and if clicked it will lead you to the Google Play store