Testing the Causal Impact of Social Media Reduction Around the Globe
Stage 1 Registered Report Conditionally Accepted at Nature
Abstract: More than half of the world’s population uses social media. There is widespread debate among the public, politicians, and academics about social media’s impact on important outcomes, such as intergroup conflict and well-being. However, most prior research on the impact of social media relies on samples from the United States and Western Europe, despite emerging evidence suggesting that the impact of social media is likely to differ across the globe. Building on the results of pilot experiments from three countries (n = 894), we plan to conduct a global field experiment to measure the causal impact of reducing social media usage for two weeks across 23 countries (projected n > 8,000). We will then test how social media reduction influences four main outcomes: news knowledge, exposure to online hostility, intergroup attitudes, and well-being. We will also explore how the effects of social media reduction vary across world regions, focusing on three theoretically-informed country-level moderators: income level, inequality, and democratic strength. This large-scale, high-powered field experiment, and the global dataset resulting from it, will offer rare causal evidence to inform ongoing debates about the impact of social media and how it varies around the world
*Led by Steven Rathje, Nejla Asimovic and Tiago Ventura. Co-authors: Sarah Mughal, Claire E. Robertson, Christopher Barrie, …., Joshua A. Tucker, Jay J. Van Bavel.**
Reducing Social Media Usage During Elections: Evidence from a Multi-Country WhatsApp Deactivation Experiment
Under Review
Abstract: Recent research has investigated how social media platforms may spread misinformation and encourage harmful political discourse, which fuels political polarization, prejudice, and offline violence. We deploy online field experiments in Brazil, India, and South Africa to examine how restricting the use of WhatsApp, the world’s most widely used messaging app, affects information exposure, political attitudes, and individual well-being. We incentivize participants to either (1) stop consuming multimedia content on WhatsApp or (2) limit overall WhatsApp usage to 10 minutes per day for four weeks ahead of each country’s elections. We find that our interventions significantly reduced participants’ exposure to misinformation, online toxicity, and uncivil discussions about politics—but at the expense of keeping up with true political news. Using a wide range of measures, we detected no changes to political attitudes, but uncovered substantial gains to individual well-being as treated participants substituted WhatsApp usage for other activities. Results highlight the complex trade-offs associated with the effects of social media use on information consumption and its downstream effects.
Joint Work with Rajeshwari Majumdar, Shelley Liu, Carolina Torreblanco, and Joshua Tucker
- Presented at APSA 2024
The relationship between offline partisan geographical segregation and online partisan segregation
Under Review
Abstract: Social media is often blamed for the creation of echo chambers. However, these claims fail to consider the prevalence of offline echo chambers resulting from high levels of partisan segregation in the United States. Our article empirically assesses these online versus offline dynamics by linking a novel dataset of voters’ offline partisan segregation extracted from publicly available voter files for 180 million US voters with their online network segregation on Twitter. We investigate offline and online partisan segregation using measures of geographical and network isolation of every matched voter-twitter user to their co-partisans online and offline. Our results show that while social media users tend to form politically homogeneous online networks, these levels of partisan sorting are significantly lower than those found in offline settings. Notably, Democrats are more isolated than Republicans in both settings, and only older Republicans exhibit higher online than offline segregation. Our results contribute to the emerging literature on political communication and the homophily of online networks, providing novel evidence on partisan sorting both online and offline.
Joint work with Megan Brown, Tiago Ventura, Joshua A. Tucker, Jonathan Nagler.
- Presented at MPSA and APSA 2023
Understanding Beliefs in Misinformation: Repetition, Partisan Signals and Bayesian Processing
In preparation
Abstract: Partisan motivations and repeated exposure are two dominant explanations for how individuals form beliefs about political misinformation. Yet, there is little research that integrates these processes, despite each pointing to different interventions to combat the spread of false information, especially in online information environments. In this paper, we situate both frameworks within a unified Bayesian model of belief formation and design survey experiments to explore several implications of this theoretical framework. We find that both partisan motivated reasoning and prior exposure (`illusory truth effects’) manifest in our data, and that they exacerbate each other, painting a bleak picture of how the steady drumbeat of partisan-flavored misinformation online influences public beliefs. However, we also find that the duration of these biases attenuates sharply over time and that attaching warning labels to false information mitigates the manifestation of both cognitive biases. The findings suggest that partisan motivations dominate belief formation in political settings, with prior exposure to misinformation playing a secondary role. These results contribute to a deeper understanding of cognitive biases in political information processing and provide a structured way of thinking about how best to understand the phenomenon of online misinformation, shifting the focus from the role of mass-level beliefs for falsehoods to the role of political elites and partisan media spreading rumors.
Joint work with Jim Bisbbe, Sarah Graham and Joshua A. Tucker
Comparing the Humanness of Machine-Generated and Human-Authored Text
Abstract: As chatbots have become more commonplace writing tools, a need exists to understand the breadth of research about the humanness of machine-generated text via techniques that extend beyond the traditional Turing Test, in both dialogue (e.g., conversing with a chatbot) and non-dialogue (e.g., reading a news article) scenarios. To fill this gap and support future work, we synthesize current literature that examines and identifies humanness features of written communication generated with the state-of-the-art generative pre-trained transformer language models, provide a working definition of humanness, propose a text-based humanness taxonomy based on linguistic properties, and identify current research gaps.
Led by: Autumn Toney-Wails. Co-authors: Leticia Bode, Tiago Ventura, Ethan Wilcox, and Lisa Singh
Under Review at ACM Computing Surveys