NYU's Center for Social Media and Politics

MAGA forces are once again attacking researchers studying online misinformation and social media manipulation. I write about this outrageous assault on free speech and academic inquiry in this article for Just Security: https://lnkd.in/e4dpZYiH

A few recent articles (all co-authored will brilliant colleagues) 1) Working paper showing that age verification laws focused on stopping minors from accessing adult content are likely ineffective — and may introduce unintended consequences Over the past two years, 19 states have passed age-verification bills, and the Supreme Court recently heard arguments about the policy’s constitutionality. While proponents argue these laws are essential to protect minors and prevent them from accessing online pornography, critics claim they are ineffective and could drive users to less regulated platforms with more dangerous content. In a new working paper using pre-registered synthetic controls & Google Trends data, we find that users in affected states simply shift their habits by searching for non-compliant sites. Specifically, searches for PornHub (a platform that complies with the law) dropped 51 percent, while searches for XVideos (a non-compliant platform) and VPN services (a way to get around the law) increased 48.1 percent and 23.6 percent, respectively. As more states consider these measures and expand digital access laws, understanding these adaptation behaviors is crucial for effective policy design. We hope policymakers can use these findings to refine implementation strategies and future legislation. Working paper: https://osf.io/z83ev Media coverage: https://lnkd.in/e7czirE9 2) Essay in The Brookings Institution arguing that we have limited insights into how AI is being used in elections — and what policymakers and tech companies can do to increase transparency https://lnkd.in/emzeQhM3 3) Analysis of TikTok's Project Texas in The Lawfare Institute, emphasizing the importance of independent data access for auditing of platforms https://lnkd.in/eXsdukRd

CDS Affiliated Faculty jonathan nagler and Joshua Tucker published two new studies examining social media's impact on political information. The first study reveals Latinos using Spanish-language social media for news are more likely to believe false political narratives compared to those who consume English-language content. Their second study shows politicians who share highly polarized content are the most active in sharing political news, with less competitive elections correlating to more polarized sharing. "How reliance on Spanish-language social media predicts beliefs in false political narratives amongst Latinos": https://lnkd.in/ec_2h8ij "News Sharing on Social Media: Mapping the Ideology of News Media, Politicians, and the Mass Public": https://lnkd.in/eFneMd87 #DataScience #PoliticalScience #SocialMedia #PoliticalPolarization

We're deeply concerned about the growing number of attacks on academic researchers studying the online landscape. This is a tremendous overview of how this research works and why it matters. We condemn these campaigns to intimidate and harass researchers, and will continue to stand behind our peers and colleagues who come under attack.

Does generative AI present an unprecedented threat to democracy, as many worried ahead of 2024? Or did it have little impact, as others now claim? The truth is we don't have enough data to draw concrete conclusions. Our latest in The Brookings Institution from Scott Babwah Brennen, Cristina de la Puerta, and Zeve Sanderson explains why. https://lnkd.in/eVjKjytV The AI Political Archive Last year, we partnered with NYU's Center on Tech Policy and the American Association of Political Consultants to collect data ourselves and build the AI Political Archive. Our goal was to combat speculations about AI's impact with robust evidence of the technology's use at scale. While we partnered with AAPC to get content straight from campaigns, because of data collection limitations most of the work was done by research assistants, who had to manually comb through Meta and Google ad libraries looking for examples of AI use in political communications. Examples of AI Use in Elections In total, we found just 71 examples. Many were already covered nationally. But we did find interesting examples of pro- and anti-democratic uses of AI in down-ballot races: county and state campaigns using AI in Meta video ads, a voter pamphlet drafted with help from an LLM, and more. These snapshots are helpful, but the fragmented and opaque nature of the information environment makes it exceedingly difficult to create a comprehensive collection for understanding AI's impact at scale. Political Disclosures Require Metadata This should be a simple fix. Both Google and Meta already collect data on and label AI-generated or augmented content. They also require political ads to disclose these tools. They should make this data available in the metadata researchers use to analyze content at scale. In addition, while it's unlikely tech regulation will pass at the federal level, many states are taking the lead on AI policy. They should consider mechanisms to increase transparency, such as creating public archives for digital political ads or requiring more robust, systematic disclosure. Much of what we think we know about AI is informed by piecemeal snapshots and media narratives rather than comprehensive evidence. Without policies requiring greater cooperation and transparency, researchers, journalists, policymakers, and the public are all working in the dark to understand AI's impact.

NEW REPORT – Better Feeds: Algorithms That Put People First. As state, federal, and global policymakers grapple with how to address concerns about the link between online algorithms and various harms, KGI’s new report from a distinguished group of researchers, technologists, and policy leaders offers detailed guidance on improving the design of algorithmic recommender systems that shape billions of users’ online experiences. Drawing on the latest research documenting these harms and evidence demonstrating the effectiveness of alternative design approaches, this guide can help shift platform recommendation systems away from attention-maximizing designs toward optimizing for long-term user value and satisfaction. https://bit.ly/3QzxVzq This report is the product of an incredible group of expert authors: Alex Moehring, Alissa Cooper, Arvind Narayanan, Aviv Ovadya, Elissa R., Jeff Allen, Jonathan Stray, Julia Kamin, Leif Sigerson, Luke Thorburn, Matt Motyl, Ph.D., Motahhare Eslami, Nadine Farid Johnson, Nathaniel Lubin, Ravi Iyer, and Zander A.. ???? The Problem Some platforms optimize their recommender systems to maximize predicted “engagement” – the chance that users will click, like, share, or stream a piece of content. This design aligns well with the business interests of tech platforms monetized through advertising. And it has been linked to a range of individual and societal harms, including the spread of low-quality or harmful information, reduced user satisfaction, problematic overuse, and increased polarization. 2?? A False Choice In policy circles, chronological feeds and blanket bans on personalization are common go-to solutions to these concerns, but they have important limitations and unintended consequences and can reward spammer behavior. They fail to take advantage of better designs in existence that put people’s? interests front and center. ? The Path Forward Platforms and policymakers can help to address the harms associated with recommender systems while preserving their potential to enhance user experiences and societal value. There is a clear path forward: designing algorithms to promote long-term user value instead of short-term engagement. The report outlines how policymakers and product designers can help make this happen by: - Promoting detailed transparency - Giving users meaningful choices and better defaults - Assessing long-term impacts of design changes ?? Learn More Better Feeds serves as a roadmap and how-to guide for policymakers and technology companies interested in creating algorithmic systems that put users' long-term interests front and center. Read the report: https://bit.ly/3QzxVzq

Will spinning off Reels into a standalone app help Meta take on TikTok? While many have complained it's not as good as TikTok, creating a separate app reflects a broader effort to prioritize Reels and invest more in its recommendation system, Sol Messing tells CNET's Samantha Murphy Kelly. https://lnkd.in/ejCbvj2p

As Trump dismantles the agencies that monitor foreign influence operations, the national security concerns around TikTok remain. But how could we even know if TikTok's algorithm has been influenced, either by China or another adversarial actor? Zeve Sanderson, Sol Messing, and Scott Babwah Brennen unpack this question in our latest piece at The Lawfare Institute.

The false claims about USAID funds going to Politico and other news outlets are Trump's latest attempts to discredit the media. Our Co-Director Joshua Tucker spoke to NPR's All Things Considered about this story and the participatory nature of disinformation online.

We are grateful to Professor Joshua Tucker at New York University and Yael Eisenstat for leading a compelling discussion yesterday on the intersection of artificial intelligence and democracy. ? They explored the challenges posed by synthetic media, disinformation, and the erosion of public trust in democratic societies. From deepfake technology to the role government can play in mitigating harm, both offered invaluable insights into navigating a dynamic and complex digital landscape. We are deeply grateful to everyone who joined us for our Virtual Briefing Series, always open to all interested individuals. If you missed it, catch up on the discussion here: https://lnkd.in/edx3_MqC Together, we will keep exploring solutions to global challenges. #AIDemocracy #ElectionIntegrity #Network2020