Hello friends and colleagues–

Over the past few months, we’ve expanded our research lines, partnerships, and Knight Fellows, plus we have a job opening an upcoming event to share. Here's what we've got for you: 

Penn MEDIATED invites applications for a Research Analyst to help expand our research compendium and support translational research projects on the information ecosystem and its impact on democracy. Apply before August 17, 2026!

We are excited to host a second online convening as part of our LLMs and Civic Discourse project. This convening brings together researchers, technologists, and practitioners to examine what this third wave of automated content moderation means in practice. We will explore how LLMs are being deployed by platforms today, how they compare to prior automated approaches, and what the emerging research in this space tells us. We will hear from experts who are building their own custom systems using LLMs for content moderation and researchers who study their implications for information integrity. 

You can see more details and register for this event via this link.

The Penn MEDIATED Research Compendium now has 62 entries to help you make sense of our increasingly chaotic information environment—here are a few essential new findings:

After the 2020 George Floyd protests, there was widespread hope that America's racial reckoning would translate into lasting change. But new research from Professor Matt Levendusky tells a different story. Analyzing cable news coverage, congressional social media posts, and a longitudinal survey of swing-state voters, Levendusky finds that conservative media successfully reframed racial justice as discrimination against white people—while Democrats largely failed to counter the narrative.

Lyle Ungar’s new paper finds that LLMs overestimate how harshly people respond to norm violations, miss how much those responses depend on social relationships, and blur the line between describing behavior and prescribing it.

Part of what drives media narratives in today’s world is who is actually creating content on social media. Research from Professor Yphtach Lelkes finds that only 7% of X users are "creators"—those who actively post and shape discourse. That small group skews older, more Republican, and consumes far more low-credibility content than lurkers or interactors. The online conversation may look far more representative than it actually is.

So what can be done about growing polarization amid a fractured media environment? Changing minds, it turns out, is less about making the right argument and more about how you have the conversation. Research from Professor Emily Falk uses fMRI and natural language data to show that pairs instructed to compromise reached higher agreement than those told to persuade—not because they stayed in sync, but because they were more willing to explore new ground together.

We have been actively tracking the emerging research on how LLMs are reshaping civic discourse, that is, how LLMs discuss political topics, refer to politicians, and relate election information. Our LLMs & Civic Discourse dashboard now includes 87 papers and the latest update surfaced some interesting findings:

Oh, and, LLMs are frequently wrong when estimating public opinion about political preferences. All this and more in our dashboard, equipped with thematic tags and short summaries! Explore it here.

Penn MEDIATED and the Atlantic Council’s Digital Forensics Research Lab (DFRLab) are partnering for the 2026 Digital Sherlocks program. This initiative trains practitioners––members of civil society, journalists, researchers—on foundational concepts as well as open-source investigative methodologies and tools to understand the information landscape and especially online influence operations. Penn's Professors Danaé Metaxa, Jane Esberg, and Dolores Albarracín are now contributing to the program’s training, directly informing practitioners on the front lines of information integrity and democratic resilience.

Penn MEDIATED is delighted to announce our membership with the Global Network Initiative—a diverse community of institutions and experts working to advance a free and democratic technological future. On July 20, we gave a presentation on how LLMs are shaping social media content moderation at GNI’s Policy Committee. We look forward to more such dialogues with the GNI community, to further integrate Penn's cutting-edge science of the information ecosystem, such as on LLMs and civic discourse as well as computational social science research investigating narrative formation, into the vital discussions on policies and practices relevant to technology and human rights.

We're thrilled to welcome three new Knight Fellows to our community of researchers studying the information ecosystem and its impact on democracy. Each fellow will work closely with our affiliated faculty throughout their fellowship.

Thomas Chamberlain is a quantitative political scientist studying elections, voting behavior, and how political actors and media manipulate issue salience.

Burak Ozturan's research focuses on understanding how modern technologies shape the production and circulation of information, and its implications for democratic deliberation.

Jacky Liang's research examines how communication strategies shape political trust and democratic accountability.

We also want to congratulate outgoing Knight Postdoctoral Fellow Amir Tohidi, who is moving on to a new postdoctoral role at the University of Basel's Faculty of Business and Economics! 

Read more about our Knight Fellows and their contributions in our latest blog.