Philip N. Howard, and others, report that most of what we know of Russia's social media campaigns against voters in democracies comes from the small amounts of data released by the major social media firms. There is certainly a constant flow of examples of suspected Russian-backed, highly automated or fake social media accounts working to polarize public understanding of important social issues. But understanding the structure and reach of the Internet Research Agency’s efforts requires large pools of data. In the summer of 2017, the major social media firms provided a snapshot of such data pertaining to campaigns against voters in the United States.
Russia's Internet Research Agency (IRA) began targeting US voters for misinformation as early as 2012, using some of the techniques it had deployed on its own citizens and those of neighboring countries in Eastern Europe. The Twitter dataset contains posts in a variety of languages. Some of the accounts were also “re-purposed” in their targeting. For example, some were shifted from operating in Indonesian for an Indonesian audience to operating in English for a US audience (see Appendices for additional data). While the IRA targeted a few different countries and language communities, the vast majority of its output was written in Russian and English. The IRA’s activities across the major social media platforms have grown in recent years. Figure 1 plots the average monthly volume of live ads purchased by the IRA per year, from 2015 to 2017. Figure 1 and Table 1 show that the volume of live Facebook ads purchased in the US by the IRA increased between 2015 and 2017. In 2016, the average monthly volume of live ads was more than double the 2015 level and remained similar in 2017. Unlike the ads, the monthly volume of organic Facebook posts rose steadily between 2015 and 2017. Between 2015 and 2016, monthly organic post volume increased almost sevenfold and continued to rise rapidly into 2017. On Instagram, after a small increase in average monthly post volume between 2015 and 2016, we see a large increase between 2016 and 2017. Unlike the average volume of Facebooks ads, the average volume of Facebook and Instagram organic posts was much higher in 2017 than in 2016: by a factor of 1.7 for Facebook organic posts, and by a factor of 2.3 for Instagram organic posts. The volume of Twitter posts (tweets) did not change significantly in the period 2015-2017, as shown in Figure 1.a.
In more detail, Table 1 shows the values for the average monthly volume of IRA Facebook ads, Facebook organic posts, Instagram organic posts, and Twitter posts (tweets). We note that Twitter provided data for many more accounts (3,841 accounts) than Facebook and Instagram (76 Facebook ad accounts; 81 Facebook pages; and 133 Instagram accounts). Google chose not to disclose any account data on ads, YouTube, or Google+ so Google is not included in Figure 1 or Table 1.
The volume of Twitter posts made available to us is much larger than the volume of Facebook ads, Facebook posts, and Instagram posts. The average monthly Twitter post volume is over fifty thousand tweets per month, while the average monthly volume of Facebook ads, Facebook posts, and Instagram posts is in the hundreds to low thousands, never exceeding the six thousand mark. The data on the Internet Research Agency (IRA) provided to the Senate Select Committee on Intelligence (SSCI) by US social media and Internet platforms demonstrates a sustained effort to manipulate the US public and undermine democracy. With years of experience manipulating public
opinion in Russia, the IRA used major social media platforms including Facebook, Instagram, and Twitter to target US voters and polarize US social media users. The Russian effort targeted many kinds of communities within the US, but particularly the most extreme conservatives and those with particular sensitivities to race and immigration. The IRA used a variety of fake accounts to infiltrate political discussion communities on the right and left, including black activist communities, in order to exacerbate social divisions and influence the agenda. Accounts posing as US users on the right and left were frequently created and operated from the same computers.
“Cyber troops” are defined here as government or political party actors tasked with manipulating public opinion online (Bradshaw & Howard, 2018a). Specifically, we focus on how these actors disseminate computational propaganda over social media platforms. We define computational propaganda as the use of automation, algorithms, and big-data analytics to manipulate public life (Howard & Woolley, 2016).
The term encompasses issues to do with so-called “fake news”, the spread of misinformation on social media platforms, illegal data harvesting and micro-profiling, the exploitation of social media platforms for foreign influence operations, the amplification of hate speech or harmful content through fake accounts or political bots, and clickbait content for optimized social media consumption. This report has examined how the IRA made use of computational propaganda to shape public opinion in the US.
The affordances of social media platforms make them powerful infrastructures for spreading computational propaganda (Bradshaw & Howard, 2018b). Social media are particularly effective at directly reaching large numbers of people, while simultaneously microtargeting individuals with personalized messages. Indeed, this effective impression management—and fine-grained control over who receives which messages—is what makes social media platforms so attractive to advertisers, but also to political and foreign operatives. Where government control over Internet content has traditionally relied on blunt instruments to block or filter the free flow of information, powerful political actors are now turning to computational propaganda to shape public discourse and nudge public opinion.
A strong democracy requires high-quality news from an independent media, a pluralistic climate of opinion, and the ability to negotiate public consensus. But the IRA leveraged social media to manufacture and spread junk news, manipulate public opinion, and subvert democratic processes. Social media platforms are among the most used applications on the Internet. We once celebrated the fact that social media let us express ourselves, share content, and personalize our media consumption. It is certainly difficult to tell the story of the Arab Spring without acknowledging that social media platforms allowed democracy advocates to coordinate themselves in surprising new ways: to send their demands for political change cascading across North Africa and the Middle East (Howard & Hussain, 2013). But the absence of human editors in our news feeds also makes it easy for political actors to manipulate social networks. In previous research conducted by the Computational Propaganda Project, we found rather paradoxical evidence of the chilling effect of social media on freedom of speech and political participation. Half of Russian Twitter conversations involve highly automated accounts that actively shape online discourses (Sanovich, 2018). In Brazil, both professional trolls and bots have been used aggressively to drown out minority and dissenting opinions during three presidential campaigns, one presidential impeachment campaign, and the major race for the Mayor of Rio (Arnaudo, 2018). Social media have gone from being the natural infrastructure for sharing collective grievances and coordinating civic engagement, to being a computational tool for social control, manipulated by canny political consultants, and available to politicians in democracies and dictatorships alike (Howard & Woolley, 2016). However, understanding precisely how social media platforms impact public life is difficult (Bradshaw& Howard, 2018a). In many democracies it is not even clear that spreading computational propaganda contravenes election laws (Howard, Woolley, & Calo, 2018). It is, however, quite clear that the strategies and techniques used by government cyber troops have an impact, and that their activities violate the norms of democratic practice.
We cannot prevent all bad actors from using computational propaganda, but in democracies we can have guidelines discouraging its use. To start to address these challenges, we need to develop stronger rules and norms for the use of social media, big data and new information technologies during elections. During 2016 and 2017 we saw significant efforts made by Russia to disrupt elections around the world, but also political parties in these countries spreading disinformation domestically. Looking at the growth of cyber troop activity from 2017 to 2018 has demonstrated that these strategies are circulating globally. We cannot wait for national courts to address the technicalities of infractions after running an election or referendum. Protecting our democracies now means setting the rules of fair play before voting day, not after. This analysis has several consequences for public policy and industry behavior. Obviously, democracies need to take computational propaganda seriously as a threat to their public life. Social media firms need to share valuable data about public life with the public. For example, Facebook now focuses on ad transparency, while disabling the API for public posts and not offering an Instagram API at all. However, in this report we found that the IRA’s political ad activity has not particularly increased over time, while organic post activity has. Organic post activity is also much greater in volume than political ad activity. As well, our findings indicate that organic posts receive far more engagement. The loss of access to the API for public post data prevents further public understanding of the latest trends in computational propaganda.
Finally, this process of investigating IRA activities has also allowed us—as researchers—to develop some recommended best practices for social media firms that want to hold the public trust. First, all social media platforms should provide an open and consistent API that allows researchers to analyze important trends in public life. For example, Twitter used to provide researchers at major universities with access to several APIs, but has withdrawn this and provides so little information on the sampling of existing APIs that researchers increasingly question its utility for even basic social science. Facebook provides an extremely limited API for the analysis of public pages, but no API for Instagram. Facebook provided the US Senate with information on the organic post data of 81 Facebook pages, and the data on Facebook ads bought by 76 accounts. Twitter’s data contribution covered activity in multiple languages, but Facebook’s data contribution focuses on activity only in English. Facebook chose not to disclose data from IRA Profiles or Groups and only shared organic post data from a small number of Pages with the Committee.
Google chose to supply the Senate committee with data in a non-machine-readable format. The evidence that the IRA had bought ads on Google was provided as images of ad text and in PDF format whose pages displayed copies of information previously organized in spreadsheets. This means that Google could have provided the useable ad text and spreadsheets—in a standard machine-readable file format, such as CSV or JSON, that would be useful to data scientists—but chose to turn them into images and PDFs as if the material would all be printed out on paper. Even in times of crisis, social media firms need to co-operate with public agencies in a way that respects users' privacy. However, sharing data about public problems should be more than performative, it should be meaningful and constructive. And it should be matched with responsive support and communication channels so that researchers can make progress understanding problems that the social media firms themselves seem to have difficulty investigating.
Managed & Built by I2M Consulting, LLC | © 2010-
