From Russia with Spin: Foreign Propaganda in U.S. 2020 Campaign News

• In Event: Gimme Some Truth: The Dynamics of Political Misinformation

Abstract

This paper uses a combination of human coding and artificial intelligence to tag and trace how Russian propaganda affects U.S. news coverage of the 2020 Presidential candidates. This research analyzes the framing of the candidates in Russian English-language propaganda outlets, defines these frames linguistically via topic modeling, and uses linguistic markers to search for Russian framing in U.S. media content. While the case study is focused on the timely issue of foreign influence on U.S. elections, it contributes to deeper understanding of the relationship among news framing, foreign propaganda, and campaigns in broader methodological and theoretical ways. The project demonstrates how human content analysis can be augmented via machine learning by using “human-in-the-loop” coding in which human coders can iteratively train machines by correcting automated codes. In addition, this system generates linguistic markers of specific media frames that can be used to detect these frames in other media content, i.e. matching the Russian propaganda frames to U.S. news coverage of U.S. candidates. Earlier use of this system showed it can define specific narratives used to characterize U.S. Democratic presidential election candidates in the primary campaign stage (Oates, Gurevich, Walker, and di Meco, 2019). On a theoretical level, this paper is a useful approach to tracing specific media messages back to a foreign propaganda origin, which addresses the problem of transparency in media content. Based on preliminary work, we expect to see key frames of U.S. candidates from Russian propaganda outlets echoed in some far-Right U.S. media outlets. This addresses a concern raised by Jamieson (2018) about the difficulty of separating Russian propaganda from pro-Trump narratives on certain topics. With this system, we should be able to ascertain links from specific US stories to specific Russian content. At the same time as this work addresses a specific puzzle in the U.S. media ecosystem -- i.e. to what extent do some U.S. outlets echo Russian propaganda -- it also provides a useful way to tag and trace how messages in general spread through a news ecosystem. The project will use Russian English-language propaganda sites including RT (formerly Russia Today) and Sputnik, while matching the content against major U.S. newspapers and television news transcripts. The analysis will be carried out by the MarvelousAI StoryArc system, which measures news narratives by combining human coding, natural language processing, computational linguistics, and machine learning. This paper addresses a key strand in this year’s conference theme of “Democracy, Difference, and Destabilization” by examining how the U.S. media systems can facilitate threats to democracy through the spread of foreign propaganda as well as how making this propaganda visible can help to detect and deter this threat.

Sources: Oates, Sarah, Olya Gurevich, Chris Walker, and Lucina di Meco. 2019. Running While Female: Using AI to Track How Twitter Commentary Disadvantages Women in the 2020 U.S. Primaries. Paper presented at the Political Communication Pre-Conference, American Political Science Association Annual Meeting, Washington, D.C.

Jamieson, Kathleen Hall. 2018. Cyberwar: How Russian Hackers and Trolls Helped Election a President: What We Don’t, Can’t, and Do Know. New York: Oxford University Press.

Authors

• Sarah A. Oates, University of Maryland

• Olya Gurevich, Marvelous AI