Featured venbot

Published on May 25th, 2016 | by EJC

0

Manipulation Of Public Opinion In Venezuela Using Political Bots

By Rafael Melendez and Flor Aguilar

In the last couple of years, politicians have capitalized on the benefits of social media, setting up campaign accounts for the dissemination of talking points and propaganda. They hire consulting firms and PR experts to manage their online presence, hoping to sway voters their way or garner public support for their bills. In the current electoral cycle, we have seen how Trump tweets, Cruz’ tabloid news, and Hillary bots have all gone viral, affecting voters’ sentiment and opinion. Social media has thus come to play a large role in how the electorate becomes informed, or even ill-informed, about current political issues.

It is widely acknowledged that several regimes have actively taken advantage of social media to manipulate public opinion, employing both people and bots to engage in online political conversations. A bot is a program that runs automated tasks over the internet, often used to collect data or interact with users. Of particular interest are social bots, found in social media sites such as Twitter, which can post or message on their own and can be used by political figures and governments to affect the information flow on these sites. A team of researchers (Michelle Forelle, Phil Howard, Andrés Monroy-HernándezSaiph Savage ) from the University of Southern California (USC) University of  Washington, Microsoft Research, West Virginia University, and the National Autonomous University  of Mexico (UNAM) carried out research to probe the use of these social bots in Venezuela.

Venezuela has one of the highest Twitter adoption rates in Latin America, with roughly 14% of internet users in the country also using Twitter. Furthermore, recent studies have shown that Venezuelans are more likely to share political views on social media than any other people in Latin America. This makes sites such as Twitter an important venue to both monitor and influence public political opinion. In fact, a Freedom House report from 2013 acknowledged that the Venezuelan government had been using social media platforms to spread their point of view and drown out political opposition, and a Twiplomacy study found that Nicolas Maduro, current president of Venezuela, is the third most effective world leader on Twitter. Curiously, this last study also found that Maduro’s tweets were ten times more likely to be retweeted than liked, raising the suspicion that the accounts doing the retweeting could be bots. Other recent studies as well as reporting during elections or political crises have shown how aggressive campaigning and cyberwafare (ex account hacking) have been used by the government to attack the opposition.

Thus, the research team studied Twitter content pertaining to Venezuelan politics with the aim of gauging the impact of bots on online political discussions. Given Twitter’s role as a platform for posting and spreading information, a natural question to ask is whether the state uses bots to spread or retweet state generated posts to increase their visibility and impact on the public. In order to answer this question, the researchers identified six high profile political figures on Twitter, four from the ruling party as well as two opposition party leaders for comparison, and collected all their tweets from Jan 1 to May 31, 2015. For a sampled subset of these tweets, they then collected the lists of retweets along with information about the accounts responsible for the retweets and the platforms they used.

Determining the full extent of bot-generated behavior on Twitter is impossible, particularly given the data access limits imposed by Twitter. For example, for any given tweet, only information on the 100 most recent retweets is made available. Furthermore, detecting all retweets generated by bots is equally unfeasible. However, by using platform information, the researchers were able to identify retweets that are highly likely to come from bots.

Informally, platforms are the applications and/or devices used when posting content to Twitter. For example, the Twitter iPhone and Android apps, as well as the Twitter web site are all platforms. When an account makes a post, this information is stored along with the tweet. Though activity through certain platforms, such as smartphone apps, could be both executed by humans and bots, there are platforms, such as Botize and Masterfollow, specifically designed for bots. Botize, for example, is a service that allows you to use a Twitter account to create your own bot with a set of tasks. A key signature of these platforms is that bots using them retweet the same content at the same time.

After an initial survey, the researchers identified 11 different platforms. Overall, 2% of all sampled retweets were bot generated. Particularly striking was the distribution of bot generated content by political parties. Though bots were used by all parties, bot generated retweets accounted for at most 2% of any ruling party figure, and less than 1% for the main opposition. In the case of the third party, they accounted for almost 5% of retweets.

image1

Image 1: Platforms for Retweeting Venezuelan Politicians, Disaggregated by Politician, 2015

Delving deeper, researchers looked at the 10% most active retweeters who used a bot platform at least once. What they discovered was that bot accounts could be categorized as either political bots or citizen bots, with the former comprising a majority. The third party (VP) stands out as the most active, using bot accounts that represent party branches of cities and states. Most of these used Botize and had the party initials as part of their usernames (ex. @VPBejuma, @VP_Carabobo). Local accounts, such as those from municipalities, tended to exclusively generate retweets from broader, state-level or national VP accounts (ex. @VPBejuma). This suggests that this network of accounts is being strategically used to disseminate party propaganda down to the local communities. Bot citizens, in contrast, had unique usernames that did not imitate party or candidate names, and used a different set of primary platforms: TweetDeck (43% of bot citizens), Masterfollow (26%), and the main Twitter Web Client (25%). All their content was exclusively about Venezuelan politics, and many retweeted the same messages at the same time. Image 1 shows an overview of the platforms these accounts used for retweeting. Image 2 shows example of bot accounts detected.

bot

Image 2: Screenshot of Bot Accounts @salud_dia1 and @elchinito_pon Retweet Incident where both show the same content.

By examining the retweet content from bots, the researchers conclude that during the sample period, bots were used mostly for impression management. Retweets highlighted the performance of politicians at public events or projected them as world statesmen negotiating with foreign leaders. These retweets, constituting only about 2% of activity, did not corroborate claims from earlier studies concerning the aggressive tactics by the ruling party to influence opinion and choke off opposition, though the researchers state that this is surely due to the lack of elections or a political crisis during the sample period. They therefore recommend the study be repeated during a time of political activity in order to truly capture how the government uses social media to further its goals. To read more, checkout the research team’s research paper.

About the Authors and Researchers:

rafael
Rafael Melendez-Rios is a PhD student in the computer science department of the University of California, Santa Barbara. He is interested in tracking dynamic characterizations of certain demographic groups in media and identify patterns signaling shifts in public opinion of these groups (ex. minorities, immigrants, LGBTQ community).
flor
Flor Aguilar is a computer engineer from the National Autonomous University of Mexico (UNAM). At UNAM Flor has done research on crowdsourcing, especially designing platforms that can crowdsource volunteers for political causes using online bots. Flor works on developing large scale systems that can coordinate people to execute any task. Her research lead her to graduate with honors from UNAM, and is currently a software engineer at the Bank of Mexico, where she continues to work with big data analytics.
michelle
Michelle C Forelle received her M.A. in Media, Culture and Communication at NYU and her B.S. in Film at Boston University. In between her degrees, Michelle worked as a multimedia production coordinator in Cambridge, MA, and taught HTML and CSS at a local community center. Her initial research focus was the intersection of music videos, new online video formats, and fan communities, a topic she continues to pursue.
image
Andrés Monroy-Hernández is a researcher in FUSE Labs at Microsoft Research and anaffiliate faculty at the University of Washington. His work focuses on the study and design of social computing systems. Research topics include crowdsourcing, peer production, remixing, civic tech, urban computing, and online learning.
saiph
Dr. Saiph Savage is an Assistant Professor at West Virginia University.  She creates computational models and platforms to mobilize the citizen crowd. She designs interactive tools that empower communities to coordinate their citizens to produce collective action and create real world change. Her research involves the areas of: Civic Media, Social Computing, Crowdsourcing, Data Visualizations, Context-Aware Computing, Human-Computer Interaction.
phihow
Philip N. Howard is a professor and writer. He holds faculty appointments at Central European University in Budapest, the University of Washington in Seattle, and is a fellow at Columbia University’s Tow Center for Digital Journalism. Currently, he works as the Director of the Center for Media, Data and Society and the founding Professor at the newSchool of Public Policy at Central European University. His website is philhoward.org, and he tweets from @pnhoward

Tags: , , , , , , , , , , , , , , , , , , , , , , , , , ,


About the Author



Comments are closed.

Back to Top ↑