Academic Research

How can one understand the spread of ideas across text data? This is a key measurement problem in sociological inquiry, from the study of how interest groups shape media discourse, to the spread of policy across institutions, to the diffusion of organizational structures and institution themselves. To study how ideas and narratives diffuse across text, we must first develop a method to identify whether texts share the same information and narratives, rather than the same broad themes or exact features. We propose a novel approach to measure this quantity of interest, which we call “narrative similarity,” by using large language models to distill texts to their core ideas and then compare the similarity of claims rather than of words, phrases, or sentences. The result is an estimand much closer to narrative similarity than what is possible with past relevant alternatives, including exact text reuse, which returns lexically similar documents; topic modeling, which returns topically similar documents; or an array of alternative approaches. We devise an approach to providing out-of-sample measures of performance (precision, recall, F1) and show that our approach outperforms relevant alternatives by a large margin. We apply our approach to an important case study: The spread of Russian claims about the development of a Ukrainian bioweapons program in U.S. mainstream and fringe news websites. While we focus on news in this application, our approach can be applied more broadly to the study of propaganda, misinformation, diffusion of policy and cultural objects, among other topics.

It is often claimed that social media platforms such as Facebook and Twitter are profoundly shaping political participation, especially when it comes to protest behavior. Whether or not this is the case, the analysis of “Big Data” generated by social media usage offers unprecedented opportunities to observe complex, dynamic effects associated with large-scale collective action and social movements. In this article, we summarize evidence from studies of protest movements in the United States, Spain, Turkey, and Ukraine demonstrating that: (1) Social media platforms facilitate the exchange of information that is vital to the coordination of protest activities, such as news about transportation, turnout, police presence, violence, medical services, and legal support; (2) in addition, social media platforms facilitate the exchange of emotional and motivational contents in support of and opposition to protest activity, including messages emphasizing anger, social identification, group efficacy, and concerns about fairness, justice, and deprivation as well as explicitly ideological themes; and (3) structural characteristics of online social networks, which may differ as a function of political ideology, have important implications for information exposure and the success or failure of organizational efforts. Next, we issue a brief call for future research on a topic that is understudied but fundamental to appreciating the role of social media in facilitating political participation, namely friendship. In closing, we liken the situation confronted by researchers who are harvesting vast quantities of social media data to that of systems biologists in the early days of genome sequencing.

As more than a billion people had done previously, on November 21, 2013, Ukrainian journalist and activist Mustafa Nayem wrote a Facebook post; this post, however, would have a much larger impact on subsequent political developments than most that had preceded it. Frustrated with President Viktor Yanukovych’s decision not to sign a long-promised association agreement with the European Union, Nayem asked others who shared his frustration to comment on his post. Even more importantly, Nayem wrote that if the post received at least 1,000 comments from people willing to join him, they should all go to Independence Square to protest. And indeed they did: starting with just a few thousand people, the protests would swell to be the largest since Ukraine’s independence, particularly after police used force against protesters at the end of November 2013. Eventually, these protests led to the resignation of the government, the exile of the former president, and indirectly to the secession of Crimea and the ongoing conflict in the eastern part of the country.

Why and when do group identities become salient? Existing scholarship has suggested that insecurity and competition over political and economic resources as well as increased perceptions of threat from the out-group tend to increase the salience of ethnic identities. Most of the work on ethnicity, however, is either experimental and deals with how people respond once identity has already been primed, is based on self-reported measures of identity, or driven by election results. In contrast, here we examine events in Ukraine from late 2013 (the beginning of the Euromaidan protests) through the end of 2014 to see if particular moments of heightened political tension led to increased identification as either “Russian” or “Ukrainian” among Ukrainian citizens. In tackling this question, we use a novel methodological approach by testing the hypothesis that those who prefer to use Ukrainian to communicate on Twitter will use Ukrainian (at the expense of Russian) following moments of heightened political awareness and those who prefer to use Russian will do the opposite. Interestingly, our primary finding is a negative result: we do not find evidence that key political events in the Ukrainian crisis led to a reversion to the language of choice at the aggregate level, which is interesting given how much ink has been spilt on the question of the extent to which Euromaidan reflected an underlying Ukrainian vs. Russian conflict. However, we unexpectedly find that both those who prefer Russian and those who prefer Ukrainian begin using Russian with a greater frequency following the annexation of Crimea, thus contributing a whole new set of puzzles – and a method for exploring these puzzles – that can serve as a basis for future research.