A Thematic Network Approach to Russian Interference in the 2016 U.S. Elections

• In Event: Social Networks and Political Mobilization in the Trump Era

Abstract

Following the 2016 U.S. Presidential elections, evidence have been presented of Russian involvement in the U.S. political online discussion via a sophisticated network of artificial accounts (Benkler, et al., 2018; Jamieson, 2018). This online network of social media accounts was operated by the Internet Research Agency (IRA), a Russian company linked to the Russian Government. A main goal of IRA was sowing discord in the U.S. political system, to bolster the presidential candidacy of Donald Trump and hold back the Clinton campaign.

Twitter accounts operated by the IRA also engaged with other topics including popular culture (Bay, 2018), genetic engineering (Dorius, S. F. & Lawrence-Dill, C. J, 2018), fracking and vaccines (Broniatowski et al., 2018). Scholars have suggested that this non-political activity was aimed at creating a seemingly authentic network of accounts and to bolster their credibility (Jamieson, 2018; Linvill et al., 2019). However, understanding of the IRA activity, and specifically the ways personas were employed strategically remains largely based on anecdotal evidence (the authors, 2020), with a focus on individual topics (e.g., vaccines). In addition, little is known about the changes in activity over time, including how discourse have changed nearing, and after, election day.

For better understanding IRA personas activity, we developed a novel method for the operationalization of thematic personas. This allows us to offer the first large-scale systematic computational examination of the use of IRA “thematic personas”. Thematic personas are operationalized as social media accounts categorized based on their consistent in their use of specific topics and themes. These thematic masks were built, in part, by the consistent use of specific language patterns native to, and hence impersonating, a target group. For example, accounts were disguised as African Americans, Trump supporters, etc. (the authors, 2020).

Our computational approach combines unsupervised machine learning and network analysis. At first, we apply topic modeling to the full data set of 2,827,928 English language tweets from 2,689 accounts identified by Twitter as operated by the IRA. Using 10-fold cross-validation we chose to focus on an optimal model of 60 topics (α=0.01; Maier et al., 2018). For each account we estimated the use of language associated with each topic. We then calculated the similarity between accounts in terms of discussed topics, using cosine similarity over the topic-user matrix. We used similarity scores between accounts to create a “thematic network”, where each node is an IRA account, and edges are similarity in topical content (and NOT social relationship). Lastly, we used community detection (Blondel et al., 2008) to segment this network to distinct thematic communities. We identified nine communities, qualitatively labeled based on open-ended analysis: (1) Hard news, (2) Anti-Trump, (3) Pro-Trump, (4) Youth talk and celebrities, (5) African Americans, (6) Mixed international topics, (7) Ukraine, (8) Soft news, and (9) retweets and hashtag games.

Having identified the various communities and dividing IRA accounts into these distinct personas we are able to observe strategic changes in the use of persona types and topics over time, especially in the months leading to the elections.

References:
Bay, M. (2018). Weaponizing the haters: The Last Jedi and the strategic politicization of pop culture through social media manipulation. First Monday 23.
Benkler, Y., Faris, R. & Roberts, H. (2018). Network propaganda: manipulation, disinformation, and radicalization in American politics. Oxford University Press.
Blondel, V. D., Guillaume, J.-L., Lambiotte, R. & Lefebvre, E. (2008). Fast unfolding of communities in large networks. Journal of Statistical Mechanics, 10, 1-12.
Broniatowski, D. A., Jamison, A. M., Qi, S., AlKulaib, L., Chen, T., Benton, A., Quinn, S. C., & Dredze, M. (2018). Weaponized Health Communication: Twitter Bots and Russian Trolls Amplify the Vaccine Debate. American Journal of Public Health, 108(10), 1378–1384.
Dorius, S. F. & Lawrence-Dill, C. J. (2018). Sowing the seeds of skepticism: Russian state news and anti-GMO sentiment. GM Crops & Food 9, 53–58.
Jamieson, K. H. (2018). Cyberwar: How Russian Hackers and Trolls Helped Elect a President What We Don’t, Can’t, and Do Know. Oxford University Press.
Linvill, D. L., Boatwright, B. C., Grant, W. J. & Warren, P. L. (2019). “THE RUSSIANS ARE HACKING MY BRAIN!” investigating Russia’s internet research agency twitter tactics during the 2016 United States presidential campaign. Computers in Human Behavior 99, 292–300.
Maier, D., Waldherr, A., Miltner, P., Wiedemann, G., Niekler, A., Keinert, A., Pfetsch, B., Heyer, G., Reber, U., Häussler, T., Schmid-Petri, H., & Adam, S. (2018). Applying LDA Topic Modeling in Communication Research: Toward a Valid and Reliable Methodology. Communication Methods and Measures, 12(2–3), 93–118.

Authors

• Dror Walter, Georgia State University

• Yotam Ophir, University at Buffalo