Trolling for Terrorism: Russian Influence and Political Reactions to Attacks

• In Event: Global Disinformation & Power

Abstract

National security events can trigger significant responses from politically attentive audiences online. Terrorism, in particular, is designed to produce a “strategic effect” shaping decisions or the broader informational and political environment in which decisions are made (West 2023, 3). Influence operations likewise seek to shape attitudes and behaviors of decision makers – a segment of citizens, authorities in office, those in non-elected roles, political opposition, and so forth. This research examines the effects of the Russian Internet Research Agency’s (IRA) influence operations in shaping attention and reactions to terrorist events during the 2016 election.

Terrorism is a politically salient topic because it can induce a visceral fear in a population. Terrorism targets noncombatants, inducing a sense of fear that can have wide-ranging consequences on trust in political authorities (Van Der Does et al, 2021). It is a useful topic to examine in relation to the IRA’s influence operations as their behavior suggests they may have anticipated at least one terrorist attack on American soil during the campaign. In particular, there is an incident where the IRA began flooding (Roberts 2020) Twitter (X) with messages about terrorism starting the day before an attack and continuing until the perpetrator was caught several days later. This incident enables comparisons with other cases of terrorism throughout the campaign period. Influence operations function as a “force multiplier” in shaping perceptions, attitudes, and reactions to events (Armistead 2004, 48; US Army Headquarters 2023, 86). Political communication scholars and data scientists have found organized influence operations are effective at raising the salience of content within online networks (Salamanos et al 2023), but we know much less about the impacts of organized influence operations on attention to a topic and reactions to news.
his research addresses the following five research questions:
Does the IRA focus on terrorism before an event increase wider discussion about terrorism in relation to the political candidates?
Is there an effect evident in the period before an attack?
Is there more engagement with IRA tweets at the time of an event and in the aftermath?
How long do these effects endure (if they exist at all)?
How semantically similar is the IRA messaging on a terrorist event? Does greater semantic similarity in IRA tweets result in greater appropriation of IRA phrasing in non IRA tweets?
Do persons who have not previously tweeted about terrorism begin to tweet about terrorism in the “left of boom” period before the terror attack?
Are there differences between far right and Islamist terrorism?
Does the IRA cover far right and Islamist terror events to the same degree?
Is there a tonal equivalence in their coverage?
Does one form of coverage on terror receive greater engagement?
Does terrorism shape candidate evaluations by Twitter users?

The research draws on five datasets. First, it uses the University of Maryland to obtain a comprehensive list of terrorist attacks in the United States during the 2016 campaign. Attacks are coded to indicate the nature of the attacker and whether it was successful. Second, it uses the official IRA datasets released by Twitter to identify IRA tweet content and metadata on IRA tweet engagement (retweets, likes, and replies). Third, it uses an original dataset collected from Twitter’s streaming API in 2016 to capture tweets about the 2016 election. This data shows the extent to which terror events impact discussions about the candidates in the election. Fourth, it uses Google search interest data to determine when there are spikes in attention to terrorism. With this data as a dependent variable, we can predict changes in attention to terrorism over time and obtain a sense of the role of terrorism in shaping views about the candidates based on the related search term data. Finally, the research uses GDELT media coverage (not event) datasets to control for the level of coverage of terrorism as a topic in television news.

Author

• Michael J. Jensen, University of Canberra