Russian Info Ops on FB and Twitter: Correlation with Violent Crime?

Russian Info Ops on FB and Twitter: Correlation with Violent Crime?

Background & Premise

Russian Information Operations targeting US citizens ramped up significantly in 2015 and even more so in 2016 leading up to the November presidential election. Further, these operations have a vastly broader scope than the presidential election alone—the operations are extensive and ongoing, they seem to be largely successful (at least as evidenced by the rhetoric of Russian sources who work or have worked for such efforts, and at least in comparison to similar Russian efforts in Finland, Germany, and France), and they target US domestic religious groups, social movements, state and local governments, educational institutions, and more.

These operations are born out of a robust info ops infrastructure with close ties to Russian oligarchs and government agencies. Most operations, and the Russian agents and agencies sponsoring and orchestrating them, were created with specific political goals in mind, such as:

·      Undermining the strength of US democracy and civil institutions, and eroding public faith in them,

·      Weakening US reputation and credibility abroad,

·      Further dividing the US population over contentious political issues,

·      Increasing violent and hate-based protests (marches, secessionist movements, etc.) in the US.

For the purposes of this SNA, data will focus on these groups’ efforts to incite, whether directly or indirectly, violence in the US. Russian info ops have been linked to white supremacist protests in Charlottesville and elsewhere, violence between Black Lives Matter and police officers (on both sides), armed secessionist marches in multiple states (most notably Texas), armed protests on behalf of religious groups (such as Westboro Baptist Church), and many other protests that were of a violent nature or that could quickly have escalated into violence.

One of the chief avenues of influence through which Russian agents influence US citizens is social media. US citizens frequently do not realize they are communicating with Russian agents or are part of Russian government funded and organized groups pretending to be US grassroots groups.

“Lit Review”/Other Work in this Field

Just one example of the research being done in this space is the Hamilton 68 project, which uses a variety of analytic tools (not SNA, but complementary analytics) to examine the extent of Russian Influence Operations on Twitter (RIOT), named for Alexander Hamilton’s 68^th essay of the Federalist Papers, which has to do with foreign interference in US elections. RAND has conducted extensive work on information operations, and Foreign Policy has for months been publishing multiple articles a week on Russian info ops in the US. SNA has certainly been used by both US and Russian organizations in examining this field, particularly in conjunction with SOF, though it is not open-source.

Research Question

The key research question is: are Russian information operations on social media significantly increasing the level of violent crime in the US?

Scope & Data Gathering Methodology

Citizens convicted of most violent crimes in the US are listed in public state databases. The SNA will be limited to FB and Twitter groups sponsored by Russian government organizations between 2015 and 2017, and Facebook and Twitter users who 1) were public members of these Facebook groups or were following these Twitter groups with public, personal Twitter accounts, and 2) who were subsequently convicted of violent crimes. This is publicly available data—we first look up those convicted of a violent crime in the last two years in a given state, and we subsequently run through the list of names to see whether each was a member of / was following certain FB/Twitter groups.

In addition to areas of potential insight from the SNA, elaborated on below, the fundamental answer to the research question will be answered primarily by comparing rates of violent crime among individuals who were followers of Russian influence operations with overall rates of violent crime in various states. If US citizens actively following several Russian agent-operated FB and Twitter groups are 2.4x more likely to be convicted of a violent crime as citizens not actively following such groups, for instance, then there is a clear correlation between these operations and the number of violent crimes in the US.

To allow for a more manageable scope, other social media platforms will not be included (though they are certainly relevant—see “Further Analysis” below).

Further allowing for a more manageable scope, at least at the advent of the project, the SNA will focus solely on groups that are publicly known to be run by Russian agents (notably by the Russian Internet Research Agency). The project will not examine all 3000+ FB ads that were Russian-sponsored—in addition to numerical problems, it is too difficult to determine who viewed particular ads, and FB does not disclose this data publicly.

There are a number of potential flaws with the methodology. First, the influence of info ops is not limited solely to violent crimes. Second, most of those involved in extremist and/or hate-based movements in the US are not subsequently convicted of violent crimes. Third, many of those influenced by Russian operations are not actively following such groups, but are influenced by intermediaries—friends, neighbors, coworkers. Fourth, though not final, there are doubtless many FB/Twitter groups run by Russian info operatives that we are not presently aware of. We are operating from a limited pool for the purposes of this SNA. However, the analysis will still yield important insights into the research question and perhaps into the larger field of information operations.

Further Research

This project leaves substantial room for further analysis that would likely provide compounding insights. In terms of initial analysis, this SNA project would be best approached 1) by first looking at FB groups, and second looking at Twitter operations; and 2) by working state-by-state through databases listing individuals convicted of violent crimes. This is publicly available information, though it varies state by state—some states, such as Texas, list all individuals convicted of “Class B Misdemeanor” crimes and above, while some state databases include individuals convicted of felonies only, etc.

For further analysis, other social media platforms could be included. The same day that FB shut down the FB group of the largest secessionist movement in Texas, for instance, which was run directly by Russian agents, the group moved to Instagram.

Additionally, the further analysis that might yield the greatest insights would be to delve deeper into the connections between the Russian influence operations and the extent to which US individuals were effectively influenced. This would address a more qualitative, equally essential part of the research question; SNA is just one important technique to get to the heart of the problem.

SNA would be especially valuable in the beginning as a means of demonstrating the magnitude of the problem both to policymakers and to the targeted, influenced citizens themselves, particularly in an increasingly “bubbled” society. Ultimately, further analysis might yield insights into the methods and processes through which the Russian info operations infrastructure works, and ways in which the US citizenship might be made more resilient to such influence—particularly where such influence leads to greater violence.

Network Analysis

This project will create a two-mode SNA using FB and Twitter groups sponsored by Russian government agents/agencies as one mode (“node type 1,” distinguishable by color) and specific individuals who were convicted of violent crimes and who were members of or followers of these FB and Twitter groups as the second mode (“node type 2”).

In the analysis, I will examine different levels of dichotimization of how many Russian-sponsored FB/Twitter groups of which each individual was a member. If available data is sufficient, and if macros provide an effective tool to gather the data, I’d also like to code for and examine different levels of dichotimization of how active each individual was in each group (i.e. number of posts in a Facebook group, number of retweets of a Russian agent’s post, etc.). This would create an additional, valued matrix for my analysis.

Centrality measures and tie strength might prove useful for analyzing which FB/Twitter groups were most influential, and the extent of their reach. Some US citizens active on social media might, for instance, be shown to have significantly extended the reach of such groups if they show a high Betweenness between group creators and violent criminals—event organizers, for instance. There might also be a correlation between those convicted of violent crimes acting out of strong feeling towards a particular political issue, and the number of FB/Twitter groups they were influenced by (Eigenvector and Egonets are two ways of further examining this)—from this, one might argue there is evidence of “stacking” influence. Centrality measures might also provide insight into which contentious issues were most effectively manipulated into violence—whether, for instance, followers of race-focused groups were more likely to be subsequently convicted of violent crimes than followers of religion- or secession-focused groups. Many other possible insights from tie strength and various centrality measures are possible with further analysis.

Further insights would likely be gleaned from faction analysis, particularly when analyzing the overall network by region. It might be possible, for instance, that certain regions of the US were more prone to influence escalating into violence than others. With sufficient data, faction analysis might also show which Russian groups were most effective at inciting violence among US social groups.

Attributes

Attributes that could be linked to specific nodes include:

Node Type 1:

·      The specific government or government-contracted agency/corporation sponsoring political information operations targeting the US public.

o   If publicly available data is sufficient, research might allow for two distinct attributes- one for sponsorship, in terms of funding, political endorsement/authorization, and logistical support, and another for orchestration, referring to the agents actually authoring and posting in such social media groups, pretending to be US citizens with divisive opinions, etc. (operating a level below the individual/institutional sponsors).

o   Again, if publicly available information proves sufficient, these attributes might be further subdivided by the ties or types of ties that info op sponsorship institutions have to high level Russian politicians, oligarchs, and other power brokers (President Putin, members of the Politburo, billionaires). Further SNA analysis is required before these ties and types can be effectively categorized.

·      Attributes corresponding with a particular US region or particular, divisive US political issue.

·      For further analysis, broadening the scope: Social media groups and activity limited specifically to influencing US opinion; to influencing Russian opinion; to influencing opinion (and available information, and upsetting the veracity of available information) in other countries towards the US and/or towards Russia; and any overlap/interrelation.

Node Type 2:

·      Type of violent crime of which the individual was convicted, and crucially—requiring further research and analysis—whether the crime was motivated by hatred towards specific groups of a diametrically opposed political stance. E.g. violent protests by white supremacists influenced by Russian propaganda designed specifically to promote hatred towards blacks, or towards immigrants; lone wolf attacks against police officers by citizens who were members of Russian-operated groups that promoted violence against law enforcement; etc.

·      Whether the individual followed Twitter groups or Facebook groups, or both.