Automation in Venezuelan Twitter Networks
Something unusual is happening in Venezuelan twitter networks.
I found repeater tweets under hashtag #TeamHDP and decided to run further analysis in gephi to take a closer look. There is an enormous amount of activity boosting this hashtag and it’s coming from a disproportionately small number of communities.
The three main influencers in this hashtag are DolarToday (2.9 million verified followers), HDPY0 (142K followers) and YoSoyJustin (130K followers) — are all notably large accounts and would naturally have a considerable reach. These influencers are being boosted by several accounts using automation to tweet and are also being amplified by shared networks which may be groups of bots.
DolarToday is a US website based in Miami that, according to wikipedia, “is more known for being an exchange rate reference to the Venezuelan bolivar” and “monitoring the Venezuelan economy.”
Currently, with no other reliable source other than the black market exchange rates, these rates are used by Reuters, CNBC, and several media news agencies and networks. The Economist states that the rates calculated by DolarToday are “erratic”, but that they are “more realistic than the three official rates” released by the Venezuelan government. The website maintains that the rates are not manipulated in order to undercut the Venezuelan government.
The DolarToday website has been a target of Venezuelan government censorship as well as a lawsuit by the Central Bank of Venezuela for allegedly falsifying the country’s exchange rates. In 2013, President Maduro accused the website of “fueling an economic war against his government and manipulating the exchange rate.”
DolarToday is also promoting opposition protests in Venezuela. Its twitter account is the most influential account in the hashtag #TeamHDP. Its tweets are being boosted by automated accounts that exhibit repetitive, bot-like characteristics and are using a social media management tool called IFTTT (If This Then That) to automate their tweets.
What immediately caught my attention in the #TeamHDP hashtag data are the shared networks between the influencers. A new white paper titled The Fake News Machine: How Propagandists Abuse the Internet and Manipulate the Public (PDF) by Lion Gu, Vladimir Kropotov, and Fyodor Yarochkin, describes a guru-follower structure of Twitter users in relation to propaganda campaigns. The paper documents groups of bots that follow and retweet the same “guru” (influencer) accounts who are usually real people. Some groups of bots actively follow and repost tweets from two or more gurus at a time as shown below in Figure 59 (page 46).
#TeamHDP tweets from June 15 show a large volume of repeater tweets boosting an opposition sit-in. One example of the repeated text translates to:
“ACTIVE RESISTANCE! Neighbors in Carabobo activated the sit-in”
This phrase along with several others has been tweeted multiple times via DolarToday and other accounts in the #TeamHDP network along with a shortlink promoting the sit-in. This campaign rode 2 hashtags: both #TeamHDP as well as #14Jun.
Viewing the data directly from the Twitter API reveals the scale of automation and repetition that is taking place in this hashtag. The following are screenshots from the excel archive I downloaded via Tweet Archivist. These are several of the accounts tweeting repeater tweets and using IFTTT to automate the campaign. There are more that I edited out for brevity, which account for the jumps in row numbers. Towards the end of this dataset, an application called “SWAT Comunicacional” is also shown as a source of tweets — this is an app made by DolarToday and available on their website that Twitter users can log into and authorize DolarToday to tweet to their own accounts via OAuth. (screenshots below)
The source of DolarToday’s tweets towards the end show as both SWAT Comunicacional and the website itself. The following screenshot shows the signup for SWAT Comunicacional on DolarToday’s website and the OAuth login screen to grant DolarToday access to individual twitter accounts.
The Spanish description of the SWAT Comunicacional app translates as:
A Message, 10 million soldiers and the backing of the most influential portal in all of Venezuela
Initially I only processed 11,756 tweets which consisted of 6529 nodes, 7773 edges and surprisingly only 34 communities. Networks with this many nodes typically contain hundreds of communities, the modularity of this network is odd. Stats shown in Tweet Archivist at just over 12,000 tweets show a large impact. The top users shown below, besides DolarToday, are accounts using IFTTT to automate tweets.
I was confused how only 34 communities could have such a tremendous impact so I repeated the process with more tweets to confirm my calculations were correct.
The following network graphs contain 18,334 tweets, 8611 nodes, 10,499 edges and only 37 communities. Which again, it is very unusual for only 37 communities to make this many tweets. Normally hundreds of communities would be represented in this size dataset. Compare the numbers of communities in my study of #Nieto122 to see the vast difference in the two networks.
Turning off the labels and zooming in to the network, the groups of shared accounts are very obvious.
Another screen capture from Tweet Archivist at 18,334 tweets shows the extraordinary reach of this network; over 1 billion impressions in just 24 hours. And again, same as the previous dataset, the top users shown in Tweet Archivist with the exception of DolarToday are all using IFTTT to automate their tweets.
The following list of accounts are examples of the top users supporting #TeamHDP and using IFTTT to tweet. All are tweeting more than humanly possible. Info via Social Bearing:
User Handle — Tweets per day
- DolarToday — 204
- RivaldoVZLA — 105
- LlaneroDigitalV — 266
- Zuliano73–194
- noticiasmiami — 394
- notiwebon — 271
- CalaveraVzla — 117
- solorzanobd — 343
- CopiarPegar — 1092
All gephi graphs above were made using the Fruchterman Reingold layout algorithm. Below are some additional graphs that I that created using Force Atlas 2 which show the connections between different elements in this network and the shared groups of accounts between them.
Bot activity has been recorded previously in Venezuelan networks. Mexican researchers from LoQueSigue documented bots in #PrayForVenezuela, a 2014 hashtag which was a worldwide trend at the time denouncing violence, repression and alleged censorship in 2014 protests. In addition, NoBotsPolitico from Spain documented networks of fake accounts that supported the La Salida protests in Venezuela until June 2014, went silent for 8 months then started up again — tweeting anti-podemos propaganda in hashtags relating to the 2015 elections in Spain.
It’s not possible to determine who is controlling the automated accounts supporting #TeamHDP however the networks are not organic. This appears to be a well-organized, deliberate operation boosting opposition protests.
