2023 Year in Review: Our Research & Impact

For more than a decade, the landscape around social media and politics has been dominated by a handful of platforms — Facebook, Instagram, and Twitter — based largely on social networks. But in the past year we have seen a rapid transformation in the online environment.

While legacy platforms increasingly focus on algorithmically-driven, video-based content, the social aspect of social media is fracturing across niche platforms. Generative AI continues to proliferate, raising important questions about how it will influence the online information environment heading into 2024. Against this backdrop, social media is also becoming harder than ever to study, with platforms becoming less transparent.

Amid this splintering ecosystem, lawmakers are reshaping the laws governing the online sphere, tech companies are building new products, and civil society and journalists are working to better understand and improve our digital lives. Updated policies are necessary to ensure a healthy democracy, but it’s critical for these policies to be informed by high-quality empirical evidence. That’s where NYU’s Center for Social Media and Politics excels.

At the center of our research is our robust data infrastructure and the diverse substantive expertise of CSMaP scholars. Together, our community continues to push the boundaries of academic study while working to nurture the next generation of scholars and experts. Our team now comprises 18 full-time researchers and operations staff, along with numerous research assistants, affiliated faculty, and graduate students. We are deeply grateful for their dedication and collaboration, and to our community of funders and internal partners whose support makes our work possible.

See below for a short overview of our research and impact in 2023, and read our full annual report for the 2022-23 academic year here.

With gratitude,

Zeve Sanderson, Jonathan Nagler, and Joshua A. Tucker

Our Research

CSMaP’s primary focus is the production of rigorous academic research and advancing scientific knowledge in public discourse. In 2023, we released nine peer-reviewed publications, posted three public working papers, pushed forward on ongoing research, and launched a number of new initiatives. 

Here’s a selection of research from the past year, and previews of upcoming projects:

• U.S. 2020 Facebook and Instagram Election Study (Science and Nature): This summer, a team of close to two dozen external academic researchers — which was co-led by CSMaP Faculty Co-Director Joshua A. Tucker — published the first articles from an unprecedented multi-year collaboration to examine the impact of Facebook and Instagram in the context of the U.S. 2020 elections. Not only were algorithms found to be extremely influential in terms of users’ on-platform experiences, but the research also demonstrated that Facebook’s users were significantly ideologically segregated in their exposure to political news. However, despite the influence of algorithms, the team found that modifying several critical features of the platform experience did not notably affect political attitudes, including political polarization. Read coverage of the studies in The New York Times, NPR, Wired, The Atlantic, The Verge, and Bloomberg. Listen to Tucker discuss the findings on podcasts from Crooked Media and Lawfare.

• Exposure to the Russian Internet Research Agency Foreign Influence Campaign on Twitter in the 2016 Election and Its Relationship to Attitudes and Voting Behavior (Nature Communications): Numerous investigations have found clear evidence that Russia interfered in the 2016 election, with the twin goals of influencing voting behavior and undermining American democracy. Despite this massive effort, however, most voters actually had little or no direct exposure to Russia's campaign on Twitter in 2016. In this paper, we found that Russian Twitter campaigns primarily reached a small subset of users, most of whom were highly partisan Republicans. In addition, there were no measurable changes in attitudes, polarization, or voting behavior among those exposed to this disinformation. Read media coverage of the paper at The Washington Post, Columbia Journalism Review, Vox, and The Atlantic. Listen to Tucker discuss the findings on the Tech Policy Press and Moderated Content podcasts.

• Generative AI: We’re exploring two lines of research related to new AI technologies. First, we’re leveraging recent innovations in LLMs to augment how we study politics. Measurement of core political concepts has remained a key challenge in quantitative research. In a recent working paper, we used ChatGPT to successfully estimate political ideology. We’re now exploring how to use these tools to better measure new concepts, such as political sectarianism, at scale. Second, we’re analyzing how AI will impact politics. We’re in the opening stage of a project with researchers at the University of California San Diego and Princeton University examining how ChatGPT answers prompts posed in English compared to Chinese. Early results suggest that prompts posed in Chinese often produce answers that differ from those posted in English, suggesting the possibility of “propaganda bias” in LLMs.

• Fracturing Social Media Environment: As noted above, after more than a decade of dominance by a few major platforms, the social media landscape appears to be splintering. In the past few years, we’ve seen the rise of broadcast-style entertainment apps, private messaging apps, and other niche platforms, ranging from burgeoning Twitter alternatives to right-wing networks. We’re setting up new research infrastructure to explore these online environments, including projects analyzing which communities use Nextdoor, examining image-based communication among far-right groups on Telegram, and collecting video data to understand how politics is talked about on TikTok.

• Understudied Populations: Most social media research — as well as resources at the platforms — focuses on the United States. But most of the world’s social media users live outside the U.S. and/or speak non-English languages. We have several projects examining these understudied populations. In a working paper posted this year, we conducted an experiment in Brazil, finding that reducing users’ WhatsApp activity reduced exposure to false rumors, but had limited effects on beliefs or political polarization. We also continued our Bilingual Election Monitor project, which pairs regular surveys with digital trace data to examine media consumption, discussion of political issues, and the spread of misinformation among English- and Spanish-speaking Latinos.

Public Impact

Beyond academia, our team also focused on increasing the public impact of the Center’s work, through policy engagement, events, and more.

Advancing Public Policy & Discourse

Leveraging our research, CSMaP experts added scientific rigor to media coverage and debate on several important topics about social media and politics in 2023.

• Artificial Intelligence: With the rise of generative AI and ChatGPT in the past year, there’s been a particular focus on how new AI tools may lead to an increase in misinformation in the 2024 election. “Just as social media reduced barriers to the spread of misinformation, AI has now reduced barriers to the production of misinformation,” wrote Joshua A. Tucker in The Hill. Whether misinformation is produced by AI or humans, social media platforms will remain the means by which it spreads. To that end, CSMaP submitted comments to the White House’s Office of Science and Technology Policy on how to mitigate the risks of AI. See more commentary in The New York Times and Vox. Also see our proposal in Tech Policy Press for AI companies to make it easier for users to donate their data for research.

• Changing Social Media Landscape: Twitter has been an important part of American political life for more than a decade. But that shifted in the past year, as the platform continued to degrade under Elon Musk’s leadership, leading to an exodus of users and advertisers. At Barron’s, Zeve Sanderson analyzed Twitter’s possible downfall. “Just like Twitter’s under-resourced technical infrastructure worked until it didn’t, network effects mean that Twitter might remain a large and diverse platform until it doesn’t,” he wrote. Sol Messing also provided insights on several Twitter alternatives, including Bluesky and Threads, for NPR, Vanity Fair, Reuters, and the Associated Press.

• Researcher Access to Data: Musk’s ownership of Twitter has also had drastic consequences for social media research. Earlier this year, Twitter stopped providing free access to its API, which enabled thousands of academics, journalists, and civil society actors to study some of the most important issues impacting our society today. In an open letter organized by the Coalition for Independent Technology Research, CSMaP joined more than 600 organizations and individuals calling on policymakers to act to protect this vital infrastructure. Data access also continues to be an issue in the European Union, which began implementing the Digital Services Act this year. CSMaP met with senior European Union officials in March and submitted comments to the European Commission in May outlining suggested standards for data access mechanisms. Read more commentary from CSMaP in The Washington Post, NBC News, Politico, and ABC News.

Academic and Public Events

In 2023, CSMaP directors and experts presented at nearly 40 external events, ranging from academic conferences and workshops to public-facing lectures. Internally, we also ran three events for public audiences, including a research presentation on foreign interference and propaganda, and two discussions on AI and democracy. Finally, we hosted our fourth annual academic conference, which welcomed more than 30 scholars from around the globe to present research, discuss future opportunities for collaboration, and network.

Data Engineering

At the heart of CSMaP’s work is our massive data infrastructure. Leveraging NYU’s high-performance computing cluster, we can analyze billions of diverse data points, ranging from social media posts to donated digital trace data. We also build and maintain open-source software tools and modeling processes that enable the broader research community. Here are key statistics regarding advancements over the past year. 

Collecting Data from Platforms

• Twitter - 65 billion tweets going back to 2016

• YouTube - 500 million videos and 3 billion comments from 178 million channels

• Reddit - 1.65 billion submissions and 12.66 billion comments since 2005

• Meta (Facebook & Instagram) - 3.5 million posts by candidates for Congress, from 2016 to present

• Gab - 393 million posts from 6.4 million users

• Gettr - 153 million posts from 6.6 million unique users

• Rumble - 80 million videos from nearly 6.5 million users

• BitChute – Infrastructure in progress to collect metadata for all videos on the platform

• TikTok – Initial stages of data collection

• News/Media - 3 million unique news articles from dozens of sources, which we aim to publish as an open-source tool for researchers

Collecting Data from People

• One unique aspect of CSMaP’s work is our ability to pair traditional surveys with digital trace data from respondents. This allows us to understand respondents’ views — and the online information environment that influences those views.

• Last year we ran two panels (YouGov and Bilingual Election Monitor) surveying nearly 6,000 Americans, in both English and Spanish. From this group, we collected data on more than 1 million Facebook posts and likes, and 5.9 million YouTube videos.

• We are also developing an Android-based mobile application for survey data collection. The app will measure users’ individual app usage and collect granular data about the content users see on social media.

Our People

By training students and postdocs, CSMaP develops a new generation of scholars and experts to explore some of the biggest questions at the intersection of social media and democracy. In the 2022-23 academic year, our faculty co-directors taught both undergraduate and graduate courses, mentoring more than three dozen undergraduate, masters, and PhD students. 

Since last fall, we also welcomed several new team members:

• Solomon Messing joined CSMaP as a Research Associate Professor. Prior to joining NYU, Sol founded data science research teams at Pew Research Center, Acronym, and Twitter. His published work spans digital media and politics, advertising and elections, and congressional communication.

• Benjamin Guinaudeau joined as a Postdoctoral Fellow. Benjamin's research leverages methodological advances to unravel the mechanisms of democratic representation in the 21st century. He is currently interested in understanding how AI systems interact with political content. 

• Wei (Rocio) Zhong joined as a Postdoctoral Fellow. Wei specializes in the application of computational methods to the study of social media and its impact on politics. Her research interests include voter behavior, the dynamics of online extremism, and the mechanisms of political polarization.

• Ben Boehme joined as a Research Engineer. Ben has a background in computational biology and software platforms for scientific collaboration, and is interested in the influence of technology and social media on participatory democracy and resistance movements.

We also continued to support and collaborate with our alumni network, which includes three dozen researchers across academia and industry. We helped two postdoctoral fellows and one research engineer transition into new roles, and said goodbye to three students who received their PhDs. 

• Megan A. Brown started her PhD at the School of Information at the University of Michigan

• Maggie Macdonald joined the University of Kentucky as an Assistant Professor in Political Science

• Tiago Ventura started as an Assistant Professor in Computational Social Science at Georgetown University’s McCourt School of Public Policy

• Nejla Ašimović joined the University of Pennsylvania’s Institute for the Study of Citizens and Politics as a Postdoctoral Fellow

• Angela Lai graduated with her PhD from the Center for Data Science at NYU in September

• Swapneel Mehta joined Boston University and MIT as a Postdoctoral Associate

In the News

Here’s a selection of additional stories citing our work and researchers in 2023:

• The New York Times: How Bad is Antisemitism Online? It’s Increasingly Hard to Know.

• NBC News: Israel-Hamas war misinformation on social media is harder to track, researchers say

• The Economist: AI will change American elections, but not in the obvious way

• PolitiFact: How Meta, TikTok, Twitter and YouTube plan to address 2024 election misinformation

• The New York Times: The Titanic Truthers of TikTok

• NPR: One mom takes on YouTube over deadly social media blackout challenge

• Associated Press: Nearly three-quarters of Americans blame media for dividing nation, poll says

• CBS News: TikTok CEO to Testify Before Congress

• Tech Policy Press: The Problem with TikTok’s New Researcher API is Not TikTok

• The Washington Post: Two years after Jan. 6, Facebook mulls if Trump is still a threat

Support CSMaP

Gifts and grants fund everything we do, from undertaking ambitious research to building out a talented team of researchers. Please consider supporting our work.