Anti-Media Populism in the US Public “Fake News” Twitter Discourse

• In Event: Poster Session: Political Communication

Abstract

Our political information ecosystem is increasingly characterized by the prevalence of misinformation that travels faster, deeper and broader than truth (Vosoughi, 2018). As the term “fake news” becomes a subject of continued scholarly interest, many have called for a closer look into how the term enters the existing social fabric and continues to be interpreted, used, and perpetuated by ordinary citizens.

This study situates “fake news” in the context of the upsurge of anti-media populism. Rising from the antagonistic view of “the people vs. the elites,” anti-media populism extends the traditional definition of elites from political and economic aristocracies to news organizations (Krämer, 2018). Portraying society in a dichotomy of “us vs. them,” populist discourses deny the legitimacy of an objective truth-telling practice long associated with journalism (Waisbord, 2018). The term “fake news” is left to strategic appropriation by the political left and the right to attack opponents and construct alternative realities (Farkas & Schou, 2018).

We examine ordinary citizen’s political utterances about “fake news” by analyzing the presence of populistic languages in the Twitter discourse on the issue. We operationalize populistic languages as the following: 1) conveying an antagonistic view between an ingroup and an outgroup (antagonism), 2) categorizing the ingroup as a unity of the people (people-centrism) and 3) holding negative sentiments towards elites as the outgroup (anti-elitism) (Oliver & Rahn, 2016). We expect that populistic languages are a substantial part of Twitter discourse on “fake news” (H1). We also ask: What are the properties of tweets that contain populistic languages (RQ1)? Are the use of populistic languages contingent on the type of actors a tweet interacts with (RQ2)?

We constructed two datasets. First, we collected 203,442 English-language tweets that contain the term “fake news” or “fakenews” through Twitter REST API. The time frame ranges from December 10th, 2016, when Donald Trump first used “fake news” on Twitter, to December 10th, 2017. We zoomed-in on tweets that “@” mentioned at least one unique twitter account that appeared more than 100 times in our corpus, resulting in 9791 tweets. We also built a second dataset of 155 unique tweets from @realdonaldtrump with the same keywords from the same period.

Two analytical techniques were adopted. We used collocation analysis to identify contiguous sequences of two adjacent words (bigrams) with likelihood ratios greater than one, focusing on key pronouns that suggest populism (we, our, they, their). We also conducted Latent Dirichlet Allocation (LDA), a widely-used topic modeling technique to extract thematic structure of texts.

As expected, populistic languages constitute a substantial part in the public discourse on “fake news” (H1). Words collocate with key pronouns define two opposing groups: “We” are a virtuous unity (we-people, our-nation), while “they” are the malicious media (they-report, their-viewers, they-lie, their-crap). These word collocations also match with those in the Trump tweet dataset.

Topic modeling analysis also reveals the thematic occurrence of populistic languages. Among the four topics emerging from the data, two of them are especially telling of the populist view of “us vs. them.” We label them as “We are the people” (e.g. “We as True Americans must and will chastise the lying, opinionated, fake, and evil news coming from CBS, CNN”) and “They are lying media” (e.g. “@cnn = fake news who cares what they say, want or do, fake is fake”). These two topics dominate over half of our sample. Tweets in these topics also show an ideological use of “fake news,” with fewer use of hyperlinks referring to specific news stories than tweets in other topics (RQ1).

Finally, we observe similar topical distributions in tweets mentioning left- and right-leaning media accounts. Nevertheless, tweets interacting with Trump-related accounts differ from the media accounts in terms of showing a stronger collocation between “we” and words with positive affects and a higher proportion of topic “We are the people” (RQ2).

Based on these initial findings, we propose to advance this study in terms of 1) extending our analyses to a broader sample, e.g. tweets without “@” mentions, 2) using time series analysis to study temporal adoption of populistic languages, and 3) using social network analysis to examine elite influence on public discourse. Our findings are useful to illustrate the (un)democratic implications of the term “fake news,” which produces new battlegrounds for partisans to consolidate ingroup identity through labeling “who our enemy is.” Such usage is often accompanied by negative affects towards the outgroup and lack of reference to specific evidence of falsehoods. Our study shows how the use of the term “fake news” poses a serious challenge for collective deliberation and action on the real problem - the spread of misinformation.

Authors

• Jianing Li, University of Wisconsin-Madison

• Min-Hsin Su, University of Wisconsin-Madison