Monday, December 31, 2018

The Breaking News Network and a continued look at the bots of the 2016 election

We have heard all about Russian bots and alt-right bots who "interfered" in the 2016 election. The Russian ones get much attention despite being an amateurish operation. We're just starting to find out about the bots secretly commissioned by American billionaire Reid Hoffman while he was accusing Roy Moore of the same. But even now, you rarely hear the accusation of "bot warfare" leveled against the left. Are they guiltless in this respect?

The short answer is no. The more complete answer is that truly everyone was running bots throughout the 2016 election, and all the way up until Twitter had a crackdown mid-2017, but also keep in mind that not all bots are nefarious. Reddit CEO Ellen Pao recently tweeted (in reference to internet traffic metrics): "everything is fake." That succinct answer generally applies to social media as well.

One researcher estimated in March 2017 that upwards of 15% of twitter accounts are bots. This estimate was somewhat validated when Twitter did a mass purge of bot-like accounts in July 2017. During that purge, the @Twitter account lost 7.7M followers, which was 12% of their previous 62M followers. And that purge definitely did not remove all bots from Twitter, as you will soon see. So this is an extremely widespread phenomenon and not limited to a single country or ideology, no matter how the media and certain members of Congress spin it.

Determining what is acceptable or not on our social media platforms is a whole separate and extremely thorny issue. But taking a level headed approach to the problem is critical. That's why I thought the story of the Breaking News Network is so relevant and fascinating. It lies at the intersection of bots-on-social-media, the contentious 2016 election and the recent upheaval of the news industry as a whole.

How I stumbled onto this

I have been doing research on a guy named Felix Sater who is a part of the Trump-Russia story. Sater is mixed up with all kinds of shady characters and happens to consistently have twitter bots pushing his side of the story. If fact, at this very moment, if you tweet anything with the words "Sater" or "Bayrock" in it, you will trigger an auto reply from a twitter bot. You should try it.

I was trying to drill down into the exact moment when a certain story about Sater surfaced. The story ran in the Associated Press on Dec 4, 2015 and it was titled "Misconduct allegations follow Trump associate with mob past." I did a twitter search for that exact phrase and received a long list of tweets that looked exactly the same, from the same day. Trying to find some sort of pattern in the accounts, I noticed at least ten of them had the same profile picture:
And the account handles also shared a prefix. It was "POLS" then a city or metropolitan area. Usually bots have a random person as a profile picture and the handle is a name plus like eight numbers, so this was different. To be honest, I thought I had discovered a network of Russian bots at first, which led me to look deeper and deeper.

I soon realized these weren't Russian bots after all, but rather it was a Bay Area company running all these accounts, which is still a fascinating story. What follows is the sum of my research up to this point.

The Breaking News Network

The Breaking News Network is/was the brainchild of a Bay Area denizen named Pat Kitano.


Kitano launched the project in 2009. He noticed that traditional local news organizations were struggling mightily and how this left huge holes in coverage. The rapid rise in new consumption via social media led Kitano to question if the entire business model of newspapers and other news outlets was outdated. The margin for sustaining a strong local news staff on ad dollars (that were vacuumed up by Facebook anyway) simply wasn’t there.

Kitano’s solution was stripping the news delivery infrastructure down to absolute bare bones and sourcing the content from people who are not "journalists" in the traditional sense, but still have newsworthy updates to provide to the community, such as city politicians, civic groups, non-profits, and independent publishers. Essentially moving from a centralized newsroom to a decentralized network of publishers, promoted via an extremely low-friction social media hub provided by the Breaking News Network, with no ads at all.

This core idea led to the creation of @breakingsfnews, established in April 2009, which served as a testing platform and a "mold" for the creating rest of the network. The rest of the network followed in 2011/2012 with accounts such as @TulsaDailyNews@arlingtonvabuzz, @amarillojournal, @allaboutnapa, etc. There were 400+ local specific accounts created.

Late 2014 signaled a shift. Not only would the Breaking News Network provide a local, general hub for community news, it would micro-target audiences within those communities. Kitano laid out fourteen groups he wanted to provide "topical coverage" to:
  1. Arts and culture
  2. Civic sustainability and the sharing economy
  3. Education and personal development
  4. Environment, climate change and green tech
  5. Ethnic and cultural news
  6. Health and wellness
  7. Housing and real estate
  8. LGBTQ
  9. Local and artisan food movement
  10. Local politics
  11. Social impact
  12. Solving homelessness, poverty and hunger
  13. Startups, small business development and entrepreneurship
  14. Women and girls
In this vein, 600+ additional accounts were created. These had a prefix before the geographical identifier such as BLCK for a curated feed for the African American community, or CLMT for news related to climate change. Here are all the different prefixes:

  • ARTS 
  • BLCK 
  • CHNG 
  • CLMT 
  • FOOD 
  • IMMREF 
  • MSLM
  • POLS
  • RE
  • STRT
So you would have accounts such as @MSLMDenver or @ImmRefNYC. Not all 400+ cities had separate accounts for all these prefixes. In fact, very few of them had all strands. But the amount of  "prefixed" accounts ended up surpassing the number of general city accounts. You can get a better sense of the accounts created at this archived page.

By my count, the Breaking News Network ended up with 1,050+ Twitter bot accounts, which is not counting the accounts on other platforms such as Facebook.

What does "bot" even mean?

Kitano lays out pretty clearly how the automation works. Each account "listens" to a select group of content creators that are relevant to that account's stated mission. When the national content creators produce content, it is echoed nationally across the accounts. When local content creators produce content, it is echoed locally. This is an example of content curation for what would be a BLCK prefixed account:
source: Medium article
Of course, one man's "curation" is another man's "bias" but you see how the content flows from source to the various local feeds. Kitano emphasizes that his accounts echo good causes, not just partisan drivel.

The University of Illinois Botometer is generally recognized as an effective tool to identify bot accounts. It is interesting the results you receive when you plug in various Breaking News Network accounts.


Most of the accounts are rated as having an over 50% likelihood of being a bot, but not all of them. The POLS accounts trigger the botometer especially. These discrepancies can be attributed both to the crudeness of the botometer and to the non-bot-like behavior of some of these accounts. But ultimately, all of these accounts were created with the same vision and are, by definition, bots.

So then, what kind of content are these accounts pushing out?

Anatomy of a single event

Remember that time during the primaries where Trump said that Hillary got “schlonged” by Obama? Well, at the time, a woman named Megan Carpentier (@megancarpentier--note the nice profile banner stating "F*** THE PATRIARCHY") wrote an op-ed about it which appeared in the Guardian. She tweeted it out from her personal twitter account and it was also posted on the main Guardian US account as well as the Guardian US Opinion account:

None of the “official” tweets of the opinion piece got much response. The Guardian US tweet got 7 retweets, the Megan Carpentier tweet got 5 retweets and the Guardian US Opinion tweet got zero response. I could find one blue-check account (an editor at The Nation) who tweeted it out, which received 32 retweets. But even the Guardian's own Ben Jacobs with his 181k follower count didn't tweet it out, even though he talked quite a bit about Trump's "schlong" remark quite a bit that day.

It did get picked up by a lot of other accounts though.


And to be clear, it was not just Breaking News Network accounts who seemed to be auto-tweeting this in a big way. It was posted by many many seemingly automated accounts all with the similar phrasing.

For example, here is a grouping of eight accounts, tweeting the exact same same thing within 2 minutes of each other and these were/are not affiliated with the Breaking News Network as far as I can tell:


And they all seem to have Indian sounding names and Indian looking profile pictures. Were these bots designed to appeal to Indian-American audiences?

Chronologically, these seem to be the first bots to tweet with that exact phrasing. So it's also possible these accounts were "trigger" accounts for other bot accounts, but I haven't been able to determine that either way. For reference, here is what Botometer thinks of these accounts

So this phenomena is not confined to the Breaking News Network at all. It is quite common in fact. But I hope you can see the concerns here.

This is an anti-Trump opinion piece which was not spread much "organically." Neither the author or publisher of the article was retweeted much, despite having substantial follower counts. And the article was not tweeted much from readers cutting and pasting the link or by pressing the tweet button from within the article. However, it was pushed substantially by bot accounts which might or might not have ulterior motives.

And the sheer quantity of accounts under the control of one entity means that that entity can create a twitter trend essentially out of thin air. They can attempt to inject talking points into the bloodstream of the public.

What content gets "pushed" would seem to depend on who is paying for the bots. During the 2016 election, lots of money was spent for partisan purposes (as in not just by the Russians) and likely a nice chunk of it was spent on bots. It's not clear who was underwriting the Breaking News Network during this time period or any of the other bot networks. Nor is it clear what, if any, impact it had. Just lots of questions and what ifs.

Outro

Twitter under-the-radar changed it's policies regarding automatic delivery of content in March 2017. Queries to the Twitter API were restricted. Each automated feed is a query to their API server so this move greatly hampered large bot networks. Bots relying on services like Hootsuite, Buffer, IFTTT, dlvr.it required re-tooling to comply with Twitter's new terms of service. Ultimately, Twitter released official changes to it's policies in June and November and did a big purge of accounts. Clearly Twitter felt the heat on this front

Due to these changes, most automated feeds in the Breaking News Network became unsustainable so the automation side of the accounts were shut down in March 2017. The accounts are still there, they are just dormant. @breakingsfnews still feeds today and has over 10k followers. The Breaking News Network Facebook pages are also still active.

The question of how we interact with and manage bots and automation in general on our social networks is a much larger issue which seems to be getting obscured by partisan talking points and reactionary purges.

For example, there was a short-lived service called thunderclap.it where Twitter users could authorize tweets to be posted on their behalf in support of a cause or campaign, creating a situation where many many accounts post the same text or hashtag in a timed rollout to maximize their viral effect. Should giving "autopilot" privileges to a service like that in furtherance of a given cause not be allowed? I don't know. In the mean time, thunderclap.it was shut down.

Transparency would go a long way but Twitter notoriously non-transparent. Even self-proclaimed "bot-trackers" like Hamilton68 refuse to publish the actual list of twitter accounts they are tracking. That gives independent researchers no way to "check their work" (that might be by design). But this issue is only going to become more relevant so more eyes and level heads are a good thing.

1 comment:

  1. Thanks for this thoughtful piece Nick; throwing out the baby with the bathwater tends to be bad strategy - understanding the bathwater tends to yield greater insight/s. It is complicated, thus in a soundbite, far too often under-researched world we're living in, this piece rings refreshingly deliberative and thoughtful. It amazes me we are not seeing a ground swell of independent analysis and research like the one you have done here. Have you seen colleagues or been reading others work on the subject? Cross sector or cyber security circles, sparking a more thoughtful dialogue about the world of social we are living in? 2020 will require this kind of thoughtfulness to ensure an election that reflects the will, and interest, of the people. Frontline's analysis of Facebook has revealed how social can be weaponized. But the opposite is also possible. Any thoughts on ways to optimize measures - incentive driven or certification of authenticity (beyond captcha)?

    ReplyDelete