Publications

Digital technologies have created a new playing field for social movements—they have impacted how movement actors organize, how we think about social movements, and how we study them. In terms of organizational strategies, activists leverage online affordances, such as efficiency and scalability, to spread their messages and coordinate action. As social movements have adapted to the digital environment in these ways, new theories have been introduced to describe them, including the “connective action” and “networked counterpublic” frameworks. Finally, the digital environment offers new opportunities for the study of social movements, from the kinds of data available to the techniques applied. It is important that scholars take up this work so that social movements may be documented, and their tactics may be shared; it is equally important that they do so thoughtfully, without reproducing the injustices these movements aim to resist.

The deaths of Black victims of police brutality, such as George Floyd, Breonna Taylor, Sandra Bland, and Philando Castile, have become focusing events and symbols for the Black Lives Matter (BLM) movement, catalyzing wide-spread public attention to racial injustice. While prior studies on hashtag activism predominantly focus on single and widely known cases, less is understood about why some incidents draw massive public attention while others do not. Addressing this gap, our study investigates the factors influencing the likelihood and size of public attention on Twitter (now X) following extrajudicial police killings. We analyzed 1.5 million tweets in response to 795 police killings between January 1, 2015, and December 8, 2016, in the United States. By examining cases on all scales, from unnoticed to prominent, we provide large-scale empirical evidence on disparities in public attention to police killings and their victims. Results indicate two distinct processes in the emergence of focusing events: While victims’ attributes such as race, age, and gender increased likelihood of receiving any attention (thresholding), variables of context and social construction were related to overall wave size (focusing).

The murder of George Floyd by police in May 2020 sparked international protests and brought unparalleled levels of attention to the Black Lives Matter movement. As we show, his death set record levels of activity and amplification on Twitter, prompted the saddest day in the platform’s history, and caused his name to appear among the ten most frequently used phrases in a day, where he is the only individual to have ever received that level of attention who was not known to the public earlier that same week. Importantly, we find that the Black Lives Matter movement’s rhetorical strategy to connect and repeat the names of past Black victims of police violence—foregrounding racial injustice as an ongoing pattern rather than a singular event—was exceptionally effective following George Floyd’s death: attention given to him extended to over 185 prior Black victims, more than other past moments in the movement’s history. We contextualize this rising tide of attention among 12 years of racial justice activism on Twitter, demonstrating how activists and allies have used attention and amplification as a recurring tactic to lift and memorialize the names of Black victims of police violence. Our results show how the Black Lives Matter movement uses social media to center past instances of police violence at an unprecedented scale and speed, while still advancing the racial justice movement’s longstanding goal to “say their names.”

In this work, we leverage a panel of over 1.6 million Twitter users matched with public voter records to assess how a standard keyword-based approach to social media data collection performs in the context of participatory politics, and we critically examine the speech this method leaves behind. We find that keyword classifiers undercount young people’s participation in online political discourse, and that valuable political expression is lost in the process. We argue that a mainstream keyword approach to collecting social media data is not well-suited to the participatory politics associated with young people and may reinforce a false perception of youth political apathy as a result.

Stance detection, which aims to determine whether an individual is for or against a target concept, promises to uncover public opinion from large streams of social media data. Yet even human annotation of social media content does not always capture “stance” as measured by public opinion polls. We demonstrate this by directly comparing an individual’s self-reported stance to the stance inferred from their social media data. Leveraging a longitudinal public opinion survey with respondent Twitter handles, we conducted this comparison for 1,129 individuals across four salient targets. We find that recall is high for both “Pro” and “Anti” stance classifications but precision is variable in a number of cases. We identify three factors leading to the disconnect between text and author stance: temporal inconsistencies, differences in constructs, and measurement errors from both survey respondents and annotators. By presenting a framework for assessing the limitations of stance detection models, this work provides important insight into what stance detection truly measures.

Core-periphery structure, the arrangement of a network into a dense core and sparse periphery, is a versatile descriptor of various social, biological, and technological networks. In practice, different core-periphery algorithms are often applied interchangeably, despite the fact that they can yield inconsistent descriptions of core-periphery structure. For example, two of the most widely used algorithms, the k-cores decomposition and the classic two-block model of Borgatti and Everett, extract fundamentally different structures: the latter partitions a network into a binary hub-and-spoke layout, while the former divides it into a layered hierarchy. We introduce a core-periphery typology to clarify these differences, along with Bayesian stochastic block modeling techniques to classify networks in accordance with this typology. Empirically, we find a rich diversity of core-periphery structure among networks. Through a detailed case study, we demonstrate the importance of acknowledging this diversity and situating networks within the core-periphery typology when conducting domain-specific analyses.

A common task in computational text analyses is to quantify how two corpora differ according to a measurement like word frequency, sentiment, or information content. However, collapsing the texts’ rich stories into a single number is often conceptually perilous, and it is difficult to confidently interpret interesting or unexpected textual patterns without looming concerns about data artifacts or measurement validity. To better capture fine-grained differences between texts, we introduce generalized word shift graphs, visualizations which yield a meaningful and interpretable summary of how individual words contribute to the variation between two texts for any measure that can be formulated as a weighted average. We show that this framework naturally encompasses many of the most commonly used approaches for comparing texts, including relative frequencies, dictionary scores, and entropy-based measures like the Kullback-Leibler and Jensen-Shannon divergences. Through several case studies, we demonstrate how generalized word shift graphs can be flexibly applied across domains for diagnostic investigation, hypothesis generation, and substantive interpretation. By providing a detailed lens into textual shifts between corpora, generalized word shift graphs help computational social scientists, digital humanists, and other text analysis practitioners fashion more robust scientific narratives.

When researchers collect and aggregate social media data, they are making explicit decisions about the populations and behaviors under study. However, there is little available guidance to ensure that these methodological choices are conceptually and empirically grounded. For example, how should researchers conceptualize a topical sample of social media content? Can it be understood as a self-contained world? Can we interpret individual accounts as participating in the same discourse or do we need to consider the ways in which different subpopulations engage? Should we disaggregate the specific mechanisms of user activity and engagement? In short: when do researchers need to consider the breadth of variation in user experience and behavior, and when can they meaningfully aggregate over such behavior? Leveraging a panel of 1.6 million Twitter accounts matched to U.S. voting records we provide empirical guidance on these questions through the conceptual lens of public sphere theory. We focus on the first nine months of 2020, giving particular attention to the Black Lives Matter movement and the COVID-19 pandemic. Examining the demographics, activity, and engagement of 800,000 American adults who collectively posted nearly 300 million tweets during this time, our findings provide practical and empirical guidance for researchers aiming to establish meaningful bounds around populations and behaviors to study. Specifically, we find that topics are imperfect but useful bounds, though topically selected tweets must be understood to be capturing segments of numerous, overlapping, and disconnected conversations. We further find that researchers should always conduct a dissaggrated analysis of tweet activity, separately examining behavior around authored tweets, retweets, quote tweets, and replies. Additionally, we find retweets and quote tweets appear to be used in distinctly different ways, potentially reflecting that retweets amplify content while quote tweets modify that content. Finally, we find that while temporal bias is inherent to social media data, its effects are manageable within our period of study. Overall, this work paints a picture of Twitter as a fluid, contextual environment best conceptualized as networked publics and characterized by enormous variety in user identity, activity, and engagement. While there are no self-contained “Twitter publics” around which perfect boundaries can be drawn, our findings provide valuable empirical guidance to researchers grappling with the conceptual implications of their methodological choices.

In the absence of clear, consistent guidelines about the COVID-19 pandemic in the United States, many people use social media to learn about the virus, public health directives, vaccine distribution, and other health information. As people individually sift through a flood of information online, they collectively curate a small set of accounts, known as crowdsourced elites, that receive disproportionate attention for their COVID-19 content. However, these elites are not all created equal: not all accounts have received the same attention during the pandemic and various demographic and ideological groups have crowdsourced their own elites. Using a mixed methods approach with a panel of Twitter users in the United States over the first year of the COVID-19 pandemic, we identify COVID-19 crowdsourced elites. We distinguish sustained amplification from episodic amplification and demonstrate that crowdsourced elites vary across demographics with respect to race, geography, and political alignment. Specifically, we show that different subpopulations preferentially amplify elites that are demographically similar to them, and that they crowdsource different types of elite accounts, such as journalists, elected officials, and medical professionals, in different proportions. In light of this variation, we discuss the potential for using the disproportionate online voice of crowdsourced COVID-19 elites to equitably promote public health information and mitigate misinformation across networked publics.

The COVID-19 pandemic created an unprecedented disruption of everyday life in America that has been exacerbated by the infodemic of COVID-19 misinformation spread on social media. In this report, we examine the tweets of 1.6 million registered American voters to ask: who is sharing COVID-19 fake news and what are they sharing? We find that older people and Republicans are more likely to share URLs from fake news web domains. In addition, we show that the most shared fake news web domain is The Gateway Pundit, which received far more shares than even the second most popular fake news domain.

Social media protest networks involve many participants, from long-time activists to individuals who are engaged only in a particular protest event. We propose a new approach to studying how various communities of users participate in protest events. Our approach combines two methodological innovations. First, rather than study tweets central to one event, we collect full timelines of user activity leading up to participation in a focal event. Second, we propose bi-spectral clustering as a scalable computational method for rapidly identifying communities of users by the words (hashtags) they use. Using a large sample of tweets from users who discussed the 2016 protests in Charlotte, North Carolina following the extrajudicial killing of Keith Lamont Scott as a case study, we demonstrate how bi-spectral clustering can be applied to sort, sample, and identify ideologically and thematically coherent clusters whose members participated in the protest on Twitter. Our proposed approach provides another tool in the mixed methods scholar’s toolkit to computationally sort and cluster large-scale network data by allowing researchers to look beyond focal hashtags or keywords and situate protest messages within the broader communication context of participating users.

The social stigma looming over disclosures of sexual violence discourages many women from publicly sharing their stories, limiting their ability to seek support and obscuring the epidemic of sexual violence against women. By inviting women to share their ordinarily silenced stories, the hashtag #MeToo surfaced a network of survivors to confront this stigma. Through a mixed-methods analysis of over 1.8 million tweets posted during the first two weeks after #MeToo gained widespread popularity in 2017, we map the landscape of disclosures that emerged and disentangle the effects of network-level reciprocal disclosures, or disclosures made in reaction to seeing others disclose. We detail how survivors disclosed a diversity of sexual violence experiences in solidarity with others, composing nearly half of all authored tweets and comprising a disproportionate number of interactions within the #MeToo network. Further, we show that the more disclosures an individual potentially saw prior to disclosing, the more likely they were to share details with their disclosure. We argue that such network-level reciprocal disclosures may have reduced stigma, creating a counterpublic space safe for disclosure which, subsequently, generated more disclosures. Our work illustrates how feminist hashtag activism, like #MeToo, can unify individual and collective narratives to dismantle the stigma surrounding disclosures of sexual violence.

Since the shooting of Black teenager Michael Brown by White police officer Darren Wilson in Ferguson, Missouri, the protest hashtag #BlackLivesMatter has amplified critiques of extrajudicial killings of Black Americans. In response to #BlackLivesMatter, other Twitter users have adopted #AllLivesMatter, a counter-protest hashtag whose content argues that equal attention should be given to all lives regardless of race. Through a multi-level analysis of over 860,000 tweets, we study how these protests and counter-protests diverge by quantifying aspects of their discourse. We find that #AllLivesMatter facilitates opposition between #BlackLivesMatter and hashtags such as #PoliceLivesMatter and #BlueLivesMatter in such a way that historically echoes the tension between Black protesters and law enforcement. In addition, we show that a significant portion of #AllLivesMatter use stems from hijacking by #BlackLivesMatter advocates. Beyond simply injecting #AllLivesMatter with #BlackLivesMatter content, these hijackers use the hashtag to directly confront the counter-protest notion of “All lives matter.” Our findings suggest that Black Lives Matter movement was able to grow, exhibit diverse conversations, and avoid derailment on social media by making discussion of counter-protest opinions a central topic of #AllLivesMatter, rather than the movement itself.

While generative models such as Latent Dirichlet Allocation (LDA) have proven fruitful in topic modeling, they often require detailed assumptions and careful specification of hyperparameters. Such model complexity issues only compound when trying to generalize generative models to incorporate human input. We introduce Correlation Explanation (CorEx), an alternative approach to topic modeling that does not assume an underlying generative model, and instead learns maximally informative topics through an information-theoretic framework. This framework naturally generalizes to hierarchical and semi-supervised extensions with no additional modeling assumptions. In particular, word-level domain knowledge can be flexibly incorporated within CorEx through anchor words, allowing topic separability and representation to be promoted with minimal human intervention. Across a variety of datasets, metrics, and experiments, we demonstrate that CorEx produces topics that are comparable in quality to those produced by unsupervised and semi-supervised variants of LDA.