Elite influence or Computational Propaganda: a case study of political discussion networks on Polish Twitter


Pawel Popiel and Emad Khazraee

In democratic theory, public opinion plays a key role in state governance, determining the problems and issues for the state to address within its policy agenda and providing feedback on existing policies. Yet, one of the enduring problems of public opinion is its vulnerability to manipulation by elites (Price, 1992), and other domestic and foreign actors. The development and diffusion of information and communication technologies (ICTs) add new layers of complexity to how such processes occur. Political voices that form public opinion are expressed not just in offline settings, but online on social media. These increasingly popular digital platforms provide additional, and potentially alternative channels for political communication. However, they also serve as sites for both state and non-state actors deploying new tools and strategies to spread propaganda and manipulate public opinion.

Instances of often highly-automated and coordinated efforts by political elites and foreign actors to disrupt and manipulate online political debate have been the subject of growing international attention. Against the backdrop of politically-charged elections and referenda in the West, accompanied by several prominent right-wing populist victories, news reports circulated about foreign attempts to influence electoral outcomes by manipulating online political debates during the Brexit referendum, the 2016 United States federal election, and the 2017 French presidential election, among others. Such “computational propaganda” (Woolley & Howard, 2016) includes the propagation of false news, amplification or disruption of political messages on Twitter by bots, and trolling political actors on social media. Yet, the effects of these activities on online political discourse remain unclear. Moreover, much of the research on the subject focuses on online publics in the United States and Western Europe. This raises the question of how the effectiveness of such efforts varies based on political and media contexts. To address this, we focus on Poland, where the only previous study found a prevalence of computational propaganda, including highly-active political bot accounts on Twitter engaged in spreading both left- and right-wing content (Gorwa, 2017). These observations raise questions about the content these actors circulate and its effects on online political debate.

Building on this, we ask the following research questions: 1) Who are the most influential actors in the political debate on Polish Twitter? 2) Are political bots and trolls among them? 3) What are the key features of the discourses these actors promote? 4) Do these discourses dominate the Twitter communities in which these actors operate? To address these questions, we examine the political information propagation processes via the analysis of retweet networks on Polish Twitter between September 1 and October 22, 2017. We apply a community detection algorithm to these networks to identify political divisions on Polish Twitter. We also conduct topic modeling to provide an account of the political debates that occur within them and the degree to which these debates are influenced by political elites, trolls, and other actors. We use a bot detection algorithm to examine the presence of political bots in these communities.

Our study suggests that political debate on Polish Twitter hews closely to a classic pattern of polarization, with little interaction between two opposing sides. Our findings reveal two cohesive retweet communities, divided largely between pro-government users, including government accounts, and the opposition to the ruling party. We find that the more influential users in the pro-government community are better connected and more influential within that community than influential users in the opposition’s community. They also produce significantly more tweets on major political events, with limited exceptions.

While we find some evidence of automation and bots, whatever effect they have on the political debate remains overshadowed by impact of public political and media figures, who wield agenda-setting influence in online and offline worlds. Thus, our findings temper previous accounts of computational propaganda in Poland, and draw attention to the very real political asymmetries between the state and its opponents that can be obscured by focusing on the political effects of automated bots.

This blog post provides an introduction to research conducted through the IPO’s Joint Digital Rights and Internet Freedom Research/Advocacy Projects for the Filter Bubbles and the Political Debate on Twitter in Poland Project. This publication is currently under review, but you can read the policy brief on the findings from the Panoptykon Foundation, here.

Price, V. (1992). Public opinion. Newbury Park, CA: Sage.

Gorwa, R. (2017). Computational Propaganda in Poland: False Amplifiers and the Digital Public Sphere. Retrieved from Oxford, UK: comprop.oii.ox.ac.uk

Woolley, S. C., & Howard, P. N. (2016). Political communication, computational propaganda, and autonomous agents. International journal of communication (Online), 10(2016), 4882-4890.