What Happened When I Called Out 300 Russian Propaganda Trolls on Twitter

They make liberal use of cat photos and emojis, drape themselves in American flag iconography, and go by handles like “Saving America,” “4GodandCountry,” and “ROCK ON OHIO.”

And yesterday, scores of them started attacking me online.

Here’s a short story about it.

First, A Quick Bit Of Context On Russian Twitter Trolls

It’s no secret Twitter has a bot problem. I’m not talking about bots that automate your posts or compose funny song lyrics, I mean hostile propaganda bots (and the people who sit behind them, operate and leverage their automation).

The vast majority of these foreign disinformation trolls are known cyber-operatives within Russian “web brigades” (Веб-бригады), state-sponsored groups that spin up bots, trolls and propaganda distribution accounts across Facebook, Twitter and other social media. And their activities — and impact — have been widely reported by outlets like The New York Times, CNN, BuzzFeed and The Daily Beast.

One of the most famous troll factories, the blandly-named “Internet Research Agency” (IRA) in St. Petersburg, even created a specific team, the “Department of Provocations,” dedicated to spreading fake news and creating cultural division through social media. Sources say management at the IRA even required its 1,000+ employees and remote contractors to watch episodes of Neflix’s now-defunct show “House of Cards” to improve their English and learn political terminology.

Russia's Propaganda Troll Factory at the Internet Research Agency
The “Internet Research Agency”’s former home on Savushkina Street in St. Petersburg, Russia

The agency was reportedly “disbanded” in December 28, 2016, a month after the conclusion of the U.S. Presidential Election, though many experts agree their work still continues under different names and fronts.

And Russia’s trolling is particularly prolific on Twitter, due to the design of the platform’s newsfeed algorithms, coupled with the company’s own lax oversight.

Twitter’s leadership, for its part, continues to be — at least publicly — asleep at the wheel on the issue. Testifying on Capital Hill on October 31 in front of the House Intelligence Committee, Twitter said it identified 2,752 accounts controlled by Russian operatives and more than 36,000 “bots” that tweeted 1.4 million times during the election, well higher than its previously reported 201 accounts linked to Russia.

Facebook and Twitter Russian Election Meddling Testimony
Facebook’s general counsel, Colin Stretch (left), and Sean Edgett, acting general counsel at Twitter (center), testify alongside Google in front of the Senate Intelligence Committee.

But not only is the 2,752 estimate still wildly conservative, Twitter has taken few if any steps to address its troll scourge.

“I’m concerned that Twitter seems to be vastly underestimating the number of fake accounts and bots pushing disinformation,” noted Senator Mark Warner of Virginia in his opening statement. “Independent researchers have estimated that up to 15 percent of Twitter accounts – or potentially 48 million accounts – are fake or automated.”

And at least 400,000 bots posted political messages during the 2016 U.S. presidential election on Twitter, according to research by Emilio Ferrara, an assistant professor at the University of Southern California.

Based on my own personal analysis, there are well north of 100,000 coordinated bot and troll accounts active on Twitter today, a network I have reason to believe is linked to Russian intelligence and disinfo operations. @SparkleSoup45 alone has over 90,000 followers, the vast majority of whom are fellow bots and propaganda handles. Dozens of linked accounts have similarly questionable followings in the tens of thousands.

SparkleSoup 45 - Russian propaganda troll bot

The more I peer down the rabbit hole, the deeper it goes, and one of my next steps will be to dust off my programming skills to collect, build and graph a more comprehensive database of this ecosystem [12/8/17 update: I’m still in the process of working on this. Unfortunately, the massive amount of data that needs to be collected from the Twitter API necessary to do this, coupled with Twitter’s own rate limiting, has made this slow-going, and I don’t quite have the computing power (or expertise) to run a lot of machines in parallel].

There are fake websites, fake church groups, fake media brands, and even a few Twitter-verified pundits with highly suspicious account activity, all operating in these same networks.

Russian Fake News
Right Side News operates with the tagline “The Right News for Americans,” one of the more obvious tells it’s not being operated by Americans
Your Strategy is Showing
Funny coincidence how Right Side News follows so many other troll and bot accounts

The right question to me at this point is not “Is there a Russian bot/troll problem on Twitter?” but more “Why won’t Twitter police or even acknowledge the extent of its Russian bot/troll problem?”

Because it can’t? Or because it knowingly won’t?

Poking The Russian Propaganda Hornets Nest

What’s frustrating to me about Twitter’s Russian propaganda problem is it’s so transparently out in the open. Scroll down the responses to any tweet from Trump, Fox News or other major political or media handle, and you’ll see the trolling and disinformation instantly.

There are of course interesting questions about to what extent “trolling” should be protected as free speech, but my view is any trolling that spreads intentionally hostile or harmful information (which often contains racist, divisive or discriminatory sub-texts) deserves no censorship protection whatsoever.

After making repeated traditional attempts to get Twitter to take action  — including emailing Twitter, submitting online complain forms, and reporting accounts through the app — I put together a list of several hundred Russian propaganda accounts (a small sample size of the tens of thousands of active ones), tweeted it out, and mentioned a few journalists, including The Observer’s John Schindler and The Daily Beat’s Noah Shachtman,

Needless to say the response from St. Petersburg was swift.

Within minutes, I started getting dozens of responses, many in broken English, ranging from dodgy denials:

to some less subtle retorts:

Mostly though, the Russian trolls seemed to simply enjoy their trolling, as if to say “yeah, you got us, but there’s nothing you or Twitter can do to stop it.” Several of the accounts that joined the conversation also appear to be active meme-creators and participants in the 4chan ecosystem, signaling the murky intersection between Kremlin propaganda and Alt-Right discourse.

Russian Troll Responds in Russian
Unexpectedly solid Russian for someone whose profile claims they’re from Oregon. One of several responses in Russian I received on the thread. Some of the replies in Russian were later deleted.

24 hours later, the responses are still coming in. My original tweet’s racked up 100+ replies from Russian trolls, I’ve been placed on several harassment lists, seen shady emails popping up in my inbox, and observed at least one attempt to hack and spam my website.

Some of this activity is coming from automated bots, but a lot of it is being carried out by actual people on the other side of the screen.

I’m not pointing any of these out for sympathy. I did this expecting consequences and possibly backlash.

What I do want to point out is how serious this operation is. In the grand scheme of internet influence, I’m a relative nobody. Nonetheless, the trolling response was swift and sizable as soon as I started to peel away at the truth. The G.R.U., F.S.B. and S.V.R. do not like when one of them is caught with its hand in the active measures cookie jar.

And if Jack or any other Twitter executive thinks this is trivial or just an inconvenient consequence of using Twitter in 2017 — and I say this respectfully, as a lover of the core product — they’ve really lost touch with the reality of what they’ve brought to life.

What’s interesting as well is just how long this problem has festered under the surface. Just in the half hour I spent researching I came across propaganda accounts who were registered on Twitter as far back as 2009. This isn’t a short-term bug in the system; many of these accounts are here to stay (until they’re exposed, at least).

Russian propaganda troll bot registered in 2009

Similarly, as existing accounts are uncovered, deleted or banned, new ones are created to replace them. Many of the larger ones even have dedicated backup accounts that are already set up.

Twitter List of Russian Trolls and Bots

Will Twitter Join The Fight Against Propaganda?

Finding the right balance between creating an open platform for expression and battling malicious communications within it no doubt poses challenges. But the mechanisms exist for Twitter to solve — or at least, significantly curtail — its troll and bot problems if it’s willing to take a long hard look in the mirror and commit to do so.

For one, methods of bot detection are growingly increasingly sophisticated, and can be done based on factors like time series analysis, semantic analysis, natural language processing (NLP), account interaction patterns, IP address, and other attributes.

It’s also likely that the best overall solution is a person-program pairing, much like centaur or Ivanov chess where a human player is paired with software. Facebook has recently taken steps in this direction, moving to hire 1,000 editors to help it fight ad fraud and fake news. Recurring queries could be run at regular intervals, then flagged accounts could be manually reviewed and researched by human security editors.

Yes, mistakes will happen, but the overall outcome will be far better for the integrity and quality of Twitter’s user experience than the disinformation buffet that exists today.

And of course there may be reasons why Twitter isn’t stepping up to the task, one in particular being Wall Street. Although Twitter’s stock is up 22% YTD, investors have penalized the company for lack of active user growth.

If bots and trolls do in fact make up 9-15% of Twitter’s active user base, as University of Southern California and Indiana University researchers suggest, Twitter’s leadership may have little interest in cleaning house.

If that’s the case, it’s sad and short-sighted.

At the end of the day, active user growth is the vanity metric, whereas Twitter’s lasting measurement will be its reputation and legacy.

I’d love to see Jack and the rest of the Twitter team take responsibility for its role in the Russia problem. If they don’t take it seriously, history may not be kind.

How to Make a Facebook Messenger Bot with Zero Code: A Step-by-Step Tutorial

As a so-so programmer, I recently looked up how to create a Facebook messenger bot that could automatically share new posts from my blog (as an alternative to someone subscribing via email or RSS). Normally, if I can find a suitable Github library or clear enough tutorial, and the app isn’t too complicated, I can hack it and get it working. What I was surprised to find is the social media bot ecosystem has already evolved so quickly that you can build your own Facebook messenger bot in a few minutes without writing a single line of code. Even better, you can do it completely free. So I thought I’d share a quick tutorial on how to make your own marketing or customer service bot (or just something to impress your friends).

Want to see my messenger bot in action? Click the button below to chat with the (simple AI version) of me [note: this may launch the messenger conversation on your phone if you have the app open there but not in your browser].

test the facebook messenger chatbot

As you can see, it’s a pretty simple, but there are a lot of different possibilities if you take the time to invest in your bot’s communication logic. So let’s get into the tutorial.

Continue reading “How to Make a Facebook Messenger Bot with Zero Code: A Step-by-Step Tutorial”

Why I Think Zuckerberg is Trying to Kill Influencer Marketing

I have a new theory — call it a prediction — about the future of influencer marketing.

Recently, Instagram has been meeting with some of its latest advertisers and media partners to council them on the social platform’s latest algorithm update — a transition that will rank newsfeed content by relevance, rather than chronology, similar to its parent Facebook’s own newsfeed algorithm. Instagram’s advice to marketers, succinctly summarized, boils down to this: “Make better content to keep up with the aesthetic expectations of users, and get ready to advertise [even more] to distribute it, because Instagram organic reach will ultimately follow the downward trend as Facebook’s.”

While any marketer caught off guard by this shift couldn’t have been paying attention to the last few years of social media history, what’s interesting to me is how the organic reach race-to-the-bottom keeps reaffirming the same, cyclical social advertising sequence: Continue reading “Why I Think Zuckerberg is Trying to Kill Influencer Marketing”

The Year Social Media Moves Beyond Social

Social is entering a new era in the history of its communications potential. In doing so, ‘social media’ companies like Facebook and LinkedIn are briskly redefining their identities, business models and the boundaries they are able to connect people — or brands to people — within. All told, 2015 looks more and more like the year social will formally move beyond social, and the time when advertisers and technologists stop talking about a company, marketing channel, event or job title as ’social,’ and, instead, simply describe it as something that is.

After all, what is or isn’t social anymore? Facebook, YouTube and Twitter are now closely interwoven throughout all modern media — from live event and TV experiences to journalism to federal government policy awareness — and thanks to mobile are now first screen centers of attention.

How do you define a social company? Today, Facebook generates more annual advertising revenue than Fox News, CNN or MSNBC, with a much faster underlying growth rate fueled by mobile device adoption and budget reallocation to digital.

image01

As we’ve talked about in the past, Facebook is also distancing itself from its own company pages and contest tabs, becoming a modern media company that connects people and serves ads across a network that extends well beyond Facebook.com. And if the definition of a social company is as open-ended as one that creates or facilitates interactive communities, brands as diverse as Amazon, eBay, Uber, Github, Kickstarter, Venmo, Medium, Pandora, Spotify and a litany of other companies are also intrinsically social businesses. ‘Social’ is where people spend time on the internet, it’s what people intrinsically want to do in their lives and with their phones, and it’s been a central element of human behavior for thousands of years.

Continue reading “The Year Social Media Moves Beyond Social”

The Speed of Social Media in Tragedy

Today’s ugly, tragic Boston Marathon bombing felt as raw and visceral as the grizzly Newtown, CT school shooting or any other recent rip through our collective sense of safety, priority, and moral good.

As I sit here now however, trying to reconcile today’s losses, thanking my lucky stars that my friends and family-members nearby escaped unscathed, and searching for more information about what could have motivated such a senseless attack beyond a broad, jaundiced hatred of American freedom, culture and national pride, I can’t help but notice the 2013 Boston Marathon Bombing conversation is dominated – and our country as united as ever – by the speed and reach of social media. Particularly Twitter.

A family member’s text message right at 3pm first alerted me to a trickle of local Tweets starting to swell from Copley Square about “explosions at the finish line.” Without a pause I reacted the first way that made sense: I grabbed my phone to try to reach as many people as quickly as possible, blasting out Twitter, Facebook and Google+ posts identifying the blast address and urging anyone who might see it to steer clear. I don’t know if anyone actually read it, I don’t know if Facebook’s organic reach-killing newsfeed algorithm cut circulation there to 10% of my social circle – but I took to social first to communicate something genuinely important, timely and urgent, making it discoverable to at least 1,500 people in my immediate network. In that way, maybe, just maybe, that helped someone connect with a loved one faster, get out of the blast area before the third bomb was set off in a controlled detonation by police, or respond to urgent need. Digital did all that before I could even start dialing my phone, and I wasn’t alone.

Twitter conversation size Boston Marathon bombing
The size and speed of the Boston Marathon bombing response on Twitter

Throughout the rest of the afternoon, as I re-connected with friends and family both near and far, Twitter remained my go-to source of information, with incredible coverage from people like @BostonDotcomsports producer Steve Silva (@stevesilva) and the Globe’s social media team. And while I rarely spend more than 5 minutes a day on Facebook unless I’m pushing out an update on my startup through Hootsuite, for the past few hours I’ve been glued to the steady, focused stream of articles, encouragement, thanks and prayers that have pushed aside the usual scattering of topically-fragmented, light-hearted status updates. The rapid interconnectedness of social media has made it a global outlet for grief, and I’ve been incredibly touched by the amazing support, kindness and community on display in the face of overwhelming tragedy during such a historic, positive event.

As Kathrine Switzer, the first female marathon runner once said, “If you are losing faith in human nature, go out and watch a marathon,” and I think also, watch how everyone’s responded to this particular marathon to check in with loved ones, keep one another safe and care for those hurt this afternoon. And, ironically, social media, a platform many typically deride as frivilous or de-humanizing, has bolstered our sense of connectivity and ability to reflect, reconcile and comment on tragedy like never before.

Nothing can undo today’s loss, so the best we can do is use it as a reminder that life is random, finite and precious, and use that as a catalyst to be better to each other, better at what we do and keep the conversations that matter going.