Instagram, Meme Seeding, and the Truth about Facebook Manipulation, Pt. 1

The last couple of weeks have brought us the first new major revelations about the reach and scope of the IRA media influence campaign. Yet the most important development about the ongoing Facebook investigation isn’t the tenfold increase in the company’s updated estimate of the organic reach of “ads” on its platform.

While the estimate increasing the reach of IRA content from 10 million people to 126 million people is surely a leap, after last week’s testimony, the real question we should be asking is: how did we suddenly arrive at 150 million?

The answer is Instagram.

Following the testimony by the company’s general counsel before the Senate and House Intel Committees, I present a large set of data to exhibit some concerns I feel must be addressed as the election investigation continues.

My following analysis shows that Facebook’s sibling property, Instagram — a service larger than Twitter and Snapchat combined — should be seen as a major influence, targeting and engagement hub for the spread of political propaganda.

How do I know? Because I amassed a huge trove of content, analytics numbers, and did the analysis to prove it. To show the reach and use of Instagram for political propaganda the last two years, I’ve collected the following evidence:

First, the full profile analytics, post “like’ history and comment statistics, and the complete content — including text, dates and original URLs — for a sample of Instagram posts from 28 of the 170 removed accounts. These posts accounted for 2.5 million recorded interactions (not estimated “shares”) and 145 million projected total interactions based on Socialblade, Klear, and Keyhole influencer data.

An interactive Tableau visualization for this data is available through the link below:

Tableau Public
Instagram Prop DataViz: Interaction & Profile Stats for 28 accountspublic.tableau.com

Second, I collected historical Crowdtangle intelligence for two now-closed Instagram profiles with nearly 10.5 million interactions (comments +likes) from 2016 and 2017, with one of the accounts logging an additional four million video views in the past year.

rainbow_nation_us: 9.88M interactions; blacks_go_viral: 3.95M video views since 10/16

Third, which I’ll have to write about in a follow-up post, I have set of more than 5000 obvious IRA posts currently residing outside of Instagram-Facebook, found through Google, along with results from a large scale network analysis capturing nearly two thousand memes and their corresponding responses that still currently resides on Instagram (and has been cross-posted to Facebook and Twitter). Much of this content has been propagated through two unofficial third-party Instagram “reposting” apps.

My conclusion: Instagram is a major distributor and re-distributor of IRA propaganda that’s at the very least on par with Twitter. In my opinion, the platform is far more impactful than Twitter for content-based “meme” engagement — especially for certain minority segments of the American population.

I argue here that Instagram is more pervasive than Twitter for political meme-spreading as well as viral outrage video-based behavioral re-targeting. Part of the reason for this is because it uses the same range of Facebook’s universe of sophisticated ad targeting infrastructure — including Lookalike and Custom Audiences. The Instagram platform can even link even video views to direct response and campaign objectives.

Instagram is also a major re-distributor of IRA memes: Two unofficial third-party “re-sharing” apps on Instagram have circulated and pushed IRA content far beyond the realm of Instagram and Facebook, and embedded it all over the internet. This includes cross-posting of memes and posts from removed accounts from Instagram back into Facebook, Instagram, and also into Twitter. These apps also helped the memes get over to Pinterest.

All of the accounts I’ve studied here have been removed, so the fact that much of their content is still lingering is a critical concern. As far I can tell, this creates a “zombie account” situation. Since this content, mentions, and links actually didn’t disappear when the original profiles were taken down, the true reach of the IRA content has yet to be uncovered. It’s likely that much of it has been missed in the audience reach and impact estimates.


Acquired by Facebook in 2012 for 1 billion dollars, Instagram boasts more than 700 million users. It’s been the fastest-growing social platform in the United States since 2015, the year the IRA campaign is rumored to have begun, and it boasts a younger and more engaged user base than Facebook.

I’m surprised to find that months after the Facebook investigation launced, Facebook’s sibling property has largely been left as an afterthought. This represents a gap in knowledge that needs to be taken seriously in both news reporting and future research.

The omission spawned criticism from Senator Mark Warner during the recent Capitol Hill testimony, who remarked how just before the November 1st testimony, Facebook amended its report to acknowledge that the IRA campaign used its platform to expose an “additional 16 million people [on Instagram] …[and] an incremental 4 million.” Another 20 million. So what’s the difference? Why does this matter?

Senate Intelligence Committee hearing on Russia election interference

It’s not the first time the company has updated a report at the last minute to paste in a footnote about Instagram.

The Russian “ads” were mostly promoted posts and micro-targeted calls to action. They encouraged Americans to follow fake Pages, and engage with and vocalize about political and social issues through comments, “Reactions,” and “Likes.” Once these things happened, the campaigns used Facebook’s analytics tools and ad tech infrastructure to track and target them further. On Facebook, Instagram, and across the internet through Facebook’s “Audience Network” and other partner marketing channels like Axciom.

https://www.intelligence.senate.gov/sites/default/files/documents/os-cstretch-110117.pdf

Meme seeding and Instagram

As a social network, Instagram is more like Twitter in the sense that standard protocol is to “follow” other people without there being an expectation to reciprocate friendship. This vertical network structure makes its use as a social broadcast amplification tool more valuable.

  1. All posts on Instagram must originate through contributed media content. If there’s no photo, video, or “story,” then there’s no Instagram post. There’s an important exception that I’ll get to later in this post.
  2. Besides “link in bio,” there are no external previews, embeds or links on Instagram. This includes in the original post as well as in the comments.

From what I’ve seen — and I’ve seen thousands of threads and posts — I feel this makes the debates on Instagram more focused on the issue, video, or meme sitting at the top of the thread. For sowing division and finding wedge issues, Instagram is an ideal visual meme broadcast factory.

Just as Facebook’s recent statement describes — these “ads” were part of a wider effort intended not just to “sway an election,” but rather to “create a sustained relationship whereby users subsequently posted about topics and issues pushed out by the accounts.” These are something like a psychological Trojan Horse.

Why? Because profile and Page follows, website visits, likes, comments, and video views all allow for further behavioral engagement-based profiling and future re-targeting.

Since the posts were frequently in the form of news updates, they were likely to have double impact, as many of the issues were already at the top of mind of many Americans. And like the “Russian Facebook Pages,” not all the removed Instagram accounts appear hyper-partisan or play on the fringe.

Due to the much younger, content-focused audience on Instagram, from the data I’ve collected and analyzed, influence efforts appear to lean toward the more subversiveimposter” influence campaigns.

One such account is the “Intersectional Feminism” @feminism_tag and the @army_of _jesus Instagram profiles that pushed non-inflammatory messages meant to drive specific wedge issues — like education, immigration, and terrorism. The re-shared Instagram post below combines them quite effectively.

Users were encouraged to follow profiles, “turn on their notifications” for the account’s newly-shared posts, respond to outrage videos, and visit the hundreds of websites and online stores the Instagram profiles linked to in their bios.

Instagram Post-Meme Seeding

For example, while collecting data, I noticed that the “link in bio” on the @ blackmattersus Instagram profile contained a goo.gl short URL.

Influencer snapshot of blackmattersus w/goo.gl link in bio (archive.co)
Public goo.gl analytics for the bio link to Black MattersUS: https://goo.gl/#analytics/goo.gl/vLk07C/all_time

Clicking through the Google short link, when I arrived at the Blackmattersus website, I noticed the URL at the end of the landing page. It said: &=utm_source= Insta&utm_campaign=Insta

This snippet says it all: &campaign=Insta. It’s literally part of the Facebook targeting campaign. Moreover, when I checked, the site appears to have accessed the Facebook Custom Audiences API from March 2016 through November 2016. The site also used a session recording tool that allows recording of users’ actions that is tied to their IP address and location.

https://builtwith.com/detailed/blackmattersus.com

As I’ve consistently argued, American tech companies have set up the infrastructure needed to “hack an election.” Russian groups simply purchased the ability to target specific groups of Americans before, during, and after the election through Facebook’s self-service psychographic advertising services.

This provides the impetus whereby Americans can be tracked and re-targeted through other affiliate technologies, data profiling services, and served “cookies” and device fingerprinting outside the scope of Facebook’s reach. The point is, the effort wasn’t just to “influence an election.” It wasn’t just to get people to “follow” the Pages and “turn on notifications” and sign up for fake events.

These campaigns were done with the intent to direct people to third-party websites, install mobile apps, engage with outrageous “viral” content, and collect emails, address, and payment information during “shopping cart” checkouts. Even at its most basic, the Facebook ad infrastructure can be used to unknowingly recruit friends, family members, and co-workers for sponsored messages and political data-driven micro-targeting.

People were also prompted to take action on deep-seated controversial issues. In some cases, “ads” succeeded in recruiting Americans to physically attend the fake Facebook “Events.”

Facebook Ad Campaign manager — Friends of People who like your Page

But this, of course, probably meant better re-targeting. The persistence of the Facebook ad tech and self-service audience segmentation tools greatly enhance the ability of coordinated influence campaigns to shape longer-term behaviors and attitudes at the population level.

And it can find people who you can use to target someone else, using them as a channel to transmit messages you seek to amplify. With the right lists and careful curation, which these accounts definitely had, ideas get pushed out the most effective way: organically, across people’s own social and geographical networks.

Black Lives Matter — Detailed Targeting “Interests” in Facebook Custom Audiences

The propaganda effort, starting with the “ads,” is just the beginning; it was the seed in a “sustained relationship” meant to keep Americans “posting about [specific] topics and issues.” Not just posting — thinking about it in specific ways. And at specific times. And responding to extensions of certain ideas which hopefully meant — eventually — they might act on them.

meme_seeding🇺🇸

Many thousands of Americans responded to these Instagram posts — some directly replying to post with extensive commentary, personal information, and sharing their reactions with others through the News Feed and hashtags. Congress should consider this longer-term behavioral reach and data privacy implications of these attempts to create a “sustained relationship with the Pages.”

I argue here: don’t forget the incredible reach of these Instagram interactions combined with tools like Custom Audiences. Especially for videos.

Data is … Still the Post-Truth

As a platform for reaching Americans with propaganda, seeding outrage, sowing social division on wedge issues, many of the IRA-associated Instagram accounts could actually be much higher engaged on average than Facebook — especially younger audiences interacting with Instagram accounts like @_blacktivistt_ , @blackmattersus, and LGBTUN’s Instagram site, @rainbow_nation_us.

Methodology

Using four different archival resources — Socialblade, Klear, and Keyhole — along with Crowdtangle — I was able to collect public details including the number of followers, number followed, user ID as well as a sample of the most interacted posts from 27 Instagram accounts that distributed and promoted IRA content.

My sample of 200 Instagram posts — approximately eight of the top posts from each of the 28 IRA-associated profiles — accounted for 2.5 million documented engagements (Instagram “likes,” comments, and video views).

Tableau Public
Instagram Prop DataViz: Interactions, Likes & Profile Statspublic.tableau.com

The vast majority of the interactions are “likes,” and this figure does not include video views, cross-posts to external platforms (e.g., Twitter) re-shares, or unofficial third-party sharing on the Repost and Regrann apps (see example below)

My Argument:

1) Instagram posts reached far more than 4 million people in the United States prior to October 2016.

Evidence: My sample of Instagram posts from IRA-associated accounts garnered more than 600,000 interactions prior to October 2016. These posts are between Jan 2016 and Sep 2016 (the known ad campaign on Facebook and Instagram dates back to 2015).

2) There’s more: Instagram posts were likely to have reached far more than 16 million people in the United States after October 2016.

Evidence: The remaining top posts in my sample of IRA-linked Instagram accounts accumulated 1,852,262 interactions between October 2016 and September 2017.

This sample of ~200 of the most engaged posts is from just 28 accounts. And these 27 Instagram accounts produced 64,330 posts between Jan 2016 and 2017.

According to Facebook, there are 170 known Instagram accounts that produced roughly 120,000 posts. So the data here would mean that these 28 accounts in my sample need to account for half of all the posts across the entire group of 170 “known” accounts. If Facebook’s numbers are indeed close, this would leave less than 60,000 posts for the entire remaining group of 142 IRA-linked Instagram accounts.

3) Third party Instagram reposting apps increase the estimated reach and interaction numbers dramatically.

Beyond the problem of understating the role of Instagram in the reach of Russian propaganda as reaching only 20 million additional people, there’s a showstopping finding that complicates everything known so far about the organic reach. The true reach of the IRA posts on both Instagram and Facebook, extending in many cases to Twitter, is likely to be much larger.

The cause for this relates to how sharing functionality on Instagram works. There’s no direct “re-sharing” of a post like on Facebook or Twitter.

Secured Borders Instagram hashtag result on 11/7/17

In other words, there’s no easy way to “retweet” on Instagram. To encourage users’ contributions of content and keep the monthly active user numbers high, every “post’ on Instagram needs to be content (an uploaded image, photo, video, or “story”).

Third party tools have filled in this feature omission with apps that can re-share other people’s posts on Instagram without the need to download the image or video (this would involve copyright issues) and then re-share it.

Once users connect their Instagram accounts to these third party services — hugely popular Android and iOS apps — they allow them to easily “repost” other Instagram posts.

Older versions of Instagram’s “repost” apps often used a watermark with the original account’s username, or handle.

It’s clear that two third-party unofficial Instagram “reposting” apps — primarily RepostApp (>10 million Android installs) and Regrann (>5 million Android installs) have further spread the easily quantifiable reach of the IRA Instagram posts.

Regrann and RepostApp-based shares from the Blacktivist, Secured Borders, Being Patriotic, Pray4Police, Mericanfury, and American Made content is currently all over Instagram.

And Repost and Regrann have spread large amounts of Instagram content directly into Twitter.

To see meme cross-posting, open up Twitter and search for “repost” or “regrann” and include one of the known IRA account handles.

blacktivists: https://twitter.com/search?q=repost%20blacktivist&src=typd
american.veterans: https://twitter.com/search?q=repost%20american.veterans&src=typd
secured_borders: https://twitter.com/search?q=repost%20secured%20borders&src=typd
being.patriotic: https://twitter.com/search?q=repost%20patriototus&src=typd
army_of_jesus: https://twitter.com/search?q=repost%20army_of_jesus&src=typd
Live Twitter ↩️ Instagram reposted content:
https://twitter.com/BlackGeniusBrnd/status/907361524670500865
https://twitter.com/womenwemiss/status/857089515051249664

Next, it’s concerning that this “reposted” Instagram content is still all over Facebook:

Do any search for regrann or repost and “________” on Facebook, and click and scroll to view ALL photos or ALL posts as needed:

Example of search: https://www.facebook.com/search/str/regrann+blacktivist/photos-keyword.

Try searching for “regrann” and “______” or “repost” and ______.” Or any combination of “regrann” or “repost” or “repostapp” and one of the known handles for the IRA-linked Instagram and Facebook usernames.

Example: Facebook @_Blacktivistt Repost/Regrann (live):

You can go to Google and do the exact same:

Example Google search for reposted IG content from @mericanfury:
https://www.google.com/search?q=mericanfury+repostapp&newwindow=1&hl=en&source=lnms&tbm=isch&sa=X&ved=0ahUKEwi5v-DZo6bXAhVFWCYKHXh_BjUQ_AUIDSgE&biw=1745
_american.made:
https://www.google.com/search?newwindow=1&tbm=isch&q=regrann+_american.made&nfpr=1&sa=X&ved=0ahUKEwj2r9KYpabXAhXJ7yYKHfheCK0QvgUIJSgB&biw=1745&bih=878&dpr=1.1
Example for veterans_come_first https://www.google.com/search?newwindow=1&tbm=isch&q=regrann+veterans_come_first&nfpr=1&sa=X&ved=0ahUKEwiembq8pKbXAhWDxSYKHQTCCFkQvgUIPCgB&biw=1745&bih=878&dpr=1.1
Example for Pray4police (Instagram repost trail):
https://www.google.com/search?q=repost+pray4police+site:instagram.com&newwindow=1&rlz=1CAACAV_enUS769US769&source=lnms&sa=X&ved=0ahUKEwifjtbnpabXAhVH2yYKHfhUDDgQ_AUICSgA&biw=1745&bih=878&dpr=1.1
There are thousands of images still around, and little do the platforms know (and much to the chagrin of the investigators), the Russian propaganda is literally still all over these platforms.
https://www.google.com/search?newwindow=1&as_st=y&hl=en&tbm=isch&sa=1&ei=ifH_WbnRI8jUmAHKw4TwCQ&q=repost+OR+regrann+secured_borders&oq=repost+OR+regrann+secured_borders

The fact is, third party Instagram apps mean that propaganda is still everywhere and that tells me that’s it clearly not been detected. There’s an entire galaxy of content that has been left out of the equation. How to understand the larger scope? It’s fairly easy — look at the network.

#stop_all_invaders hashtag Instagram 11/7/17

Gaming the Numbers

To further emphasize that the 16 million number is off target, I gathered the past year and a half of complete Crowdtangle analytics for two IRA-affiliated Instagram accounts. Returning to the testimony, beginning in October 2016 — to when exactly?

For Oct 2016 through August 2017, the fact that I have ONE Instagram account (rainbow_nation_us) with more than nine million interactions, another account (blacks_go_viral) with four million video views, and a small sample of ~200 posts from 28 Instagram accounts totaling close to 2.5 million interactions.

This is from a very small sample of Instagram posts — representing less than 25% of the total known accounts. These aren’t estimated shares, either; these are defined engagements (likes, comments, and video views). The estimated total reach of these 28 closed accounts, in case anyone is interested, is 145 million interactions. And that’s if you don’t count the reposted content, which is literally still everywhere on Instagram, Facebook, Pinterest, and Twitter. And Pinterest: https://www.pinterest.com/rebeltexas

One clap, two clap, three clap, forty?

By clapping more or less, you can signal to us which stories really stand out.