Archive for November, 2015

A (Probably) Incomplete Taxonomy of Peñabots (And Their Friends)

Saturday, November 28th, 2015

In this post I catalog the different ways local, state, and federal governments in Mexico have used social media bots in different ways to their advantage. This is the other side of the excellent efforts made by projects like Botivist to link activists with the use of bots. However, what may be useful to activists may also be useful to state actors. I think a list of this kind is necessary because it is important to be able to recognize the ways in which such state actors (even well-funded private actors) can subvert speech online. In this list I mention cases where large groups of people rather than software are used. Partly because it is hard to discern between the two methods, and mostly because the outcome is essentially the same. The methods for spreading or silencing information can be achieved through either means (manual or automated). I believe this to be especially true since governments usually have plenty of resources at their disposal. At the end I pose some questions yet to be answered.

  • Fake Support
    1. Spread campaign messages by tweeting the same message or similar messages in a coordinated effort
    2. Retweeting supporters or online content (news, posts, photos) that is favorable to the campaign or political agenda.
    3. Coordinating large number of people with real and fake accounts to tweet the same hashtag to get it trending (used in conjunction to bumping off hashtags)
    4. Padding Twitter follower count.
  • Drowning Out Oppositional Voices
    1. Flood hashtags with spam so that discourse is impossible. This can occur when many bots tweet a trending hashtag so often that the feed becomes difficult to read because of Twitter’s auto-refresh feature. Filling the hashtag with spam could also hinder discussion since finding meaningful conversation among in a feed can be difficult. Another consequence of filling a hashtag with spam could be the removal of a hashtag from the TT list.
    2. Knock hashtags off TT list by replacing it with other favorable hashtags.
    3. Spread misinformation?
    4. Trolling
  • Defamation and Intimidation
    1. Intimidation
    2. Defamation
      • Lydia Cacho Ribeiro, Sept 2013, from QR gov.

Subverting Speech vs. PR Campaign
If a presidential campaign hires a PR company which utilizes a large amount of volunteers or paid workers or software bots to tweet in a coordinated effort to propagate your message, is this free speech? Does this subvert other’s speech? If this army of bots floods a discussion forum, does this act as censorship? What about when an activist organization does the same thing? I think there needs to be a discussion about what’s considered ethical when employing the use of bots.

Detecting Bots
It is difficult to know when we are dealing with software bots, with a PR campaign made up of volunteers, or merely just regular citizens tweeting in support of the government. Several people have tried to analyze Twitter data to “prove” the presence of bots, but I’m still left with a lot of questions.  What if there really was a lot of non-political Twitterers who were excited about the long weekend (#EnDiaDePuente)? What if the Twitter algorithm worked against the activist hashtag? I’ve been to tech conferences where the hashtag gets spammed, what if we’re dealing with regular Twitter spam and not state-sponsored spam? Are there any dangers in misclassifying spam and bots? Are there any negative effects in constantly blaming every piece of spam on the Peñabots?

Trending Topic Algorithm
The claims that TT such as #YaMeCanse were bumped by bots are hard to prove (despite the fancy graph videos). Similar claims of censorship occurred during the Occupy Wall Street protests where many activists claimed Twitter was censoring them by not adding #occupywallstreet to the TT list. In response, several bloggers, journalists and Twitter themselves, explained that the TT algorithms look for “trendiness” and “burstiness”, and that while OWS might have had a large volume of Tweets, it did not display the spikes in volume the algorithm is constantly looking for. Although I don’t doubt the presence of bots (even State bots), I am not convinced that anything other than the regular algorithmic forces were at play in knocking #YaMeCanse off the TT list. Furthermore, more research and discussions need to be had around the importance of trending topics to activist causes. Are they crucial? Are they overvalued?

Analyzing Twitter Data
Twitter is VERY noisy. Even very specific words that seem to only talk about one specific event like “balacera” (shootout) are hard to analyze. Especially when the word becomes part of quotidian culture. For example, in Mexico people tend to tweet when there is a shootout in their city, usually to warn others to stay out of the area. In trying to analyze ‘balacera’ tweets, I’ve found that people also talk about past shootouts, potential shootouts, and even tell jokes about shootouts. So one needs to be careful when analyzing tweets. Some of the best insights on Twitter use I’ve seen have come from qualitative research methods like interviews.

Bots Are People Too

Friday, November 27th, 2015

I came to love bots because of @TwoHeadlines, @CongressEdits and NYTimes Haiku. Not only did they bring me great entertainment, but as is the case with @CongressEdits, I thought they were providing a great service. However, not all bots are cool. Lurking among us are bots sponsored by governments. Bots whose mission is to give a false impression of support or to drown out opposing views. These types of bots have been used in recent years by the governments of Mexico, Venezuela, Russia and Syria. However the methods by which these governments use bots are not as clear and transparent as they are with political bots like @CongressEdits. In this post, I’ll use the government of Mexico’s use of political bots as an example of how the traditional definition of political bots might not be enough cover what’s happening in Mexico.

I’ll start by using Wikipedia’s definition of bots and is the one widely used to describe bots. Bots are “software application that runs automated tasks over the Internet. Typically, bots perform tasks that are both simple and structurally repetitive, at a much higher rate than would be possible for a human alone.” I like this definition because it leads me to ask the question: What if you had access to enough resources and low-skilled labor such that you could perform the simple and repetitive tasks with real people? This is how our story of the Peñabots begins.

Shortly after the start of the 2012 Mexican presidential election, then candidate Enrique Peña Nieto rose to Twitter dominance. It seemed the candidate was enjoying an unusual amount of support from social media users. To some savvy netizens, this seemed suspicious. Sure enough, Twitterverse eventually discovered that the EPN campaign was making use of large armies of volunteers to show support for the candidate and to drown out any opposition that could appear online. The name Peñabot appeared shortly after this discovery in November 2011. It is unclear whether the word “bot” in “Peñabot” was originally used to fit the traditional definition of a “software that performs a repetitive task” or if it was originally used to refer to actual volunteers who under direction from campaign managers would mindlessly perform repetitive tasks such as tweeting the same text and retweeting other supporters. In case of the latter, the word Peñabot is similar to Limbots and Obamabots and is often used in parallel with Pejezombies, or the followers of competing candidate Lopez Obrador.

penabot2

Although people seem to be pretty confident that the campaign (and now actual government) used software bots, it has been confirmed that EPN did in fact use armies of volunteers to take over the conversation on Twitter. One former EPN social media manager confessed that at one point the campaign was coordinating 20,000 volunteers to show support for the candidate and to target and drown out any opposing hashtag that could surface during the campaign trail.

The ability to coordinate such a large army of volunteer may be effective in bumping hashtags off the trending topic list since Twitter’s algorithms measure “burstiness” of tweets. One example where volunteers overran a hashtag was with #MarchaAntiEPN (March Against EPN) in the state of Tabasco where volunteers bumped the hashtag off the list with #TodoTabascoConEPN (All of Tabasco With EPN). People discovered the presence of Peñabots because the majority of #TodoTabascoConEPN tweets came from a single location in Mexico City (apparently tweets were geotagged?).

There is no reason why EPN’s government couldn’t carry out their campaign using actual software bots (RT’ing pro-government tweets, tweeting from a list of pre-written messages, crowding hashtags and making your own hashtag trend). However, since they relied so heavily on “manual” tweeting to achieve something that a bot might be able to do, I wonder if it makes sense to expand the definition of bots to include such large-scale tweeting campaigns?  Does it matter to citizens whether a pro-government message comes from a person or a bot if they can’t tell which one produced the message? Does it matter to citizens if they knew it was a bot or a person acting as part of a larger campaign? 

Human or Bot?