The fall semester is upon us and Penn MEDIATED is making some moves! We are looking to build the first public-facing data science institute for the information ecosystem—think the Urban Institute or Center for Security and Emerging Technology, but for media and information technology. We’re going to develop new tools, data infrastructure, and data visualization stories to help the public better understand the information ecosystem. What does that look like so far?

Oh, and did we mention, you can check all of these updates and more in our new website which is now live with our beautiful visual branding refresh!

Let's get into it.

The Media Bias Detector is Live!

This project is foundational to our mission and two years in the making. Led by our Co-Director Duncan Watts, the Media Bias Detector gives researchers, journalists, and the public a scalable way to see how news coverage differs across publishers, tone, political lean, and insights into who the media quotes! To ensure this tool serves our community, the Computational Social Science Lab team is also releasing a dataset of 140,000+ articles from 10 major publishers, along with the underlying code, so others can build on the framework and explore their own questions about news coverage.

Read the Full Paper ⟶

Penn MEDIATED at the Computation and Journalism Symposium in Chicago!

Penn MEDIATED had two contributions at this year's Computation and Journalism Symposium at Northwestern University. Our Executive Director, Alex Engler, presented his paper, co-authored with Professor Danaé Metaxa, on the need for longitudinal monitoring of LLM outputs. Published by the Carnegie Endowment for International Peace, the paper argues that LLMs' Dynamic, Ephemeral, Embedded, and Personalized nature demands ongoing monitoring rather than one-off audits. Read the full paper here.

We also put that argument into practice with a hands-on workshop, led in collaboration with PhD student Ro Encarnación and Professor Metaxa, introducing journalists to Penn's Open Auditor System, a tool built to do exactly this: run repeated audits of AI chatbots on critical social and political topics over time. In a packed room, we heard from journalists about how they use AI in their workflow, and the role tools like OpenAuditor can play in making audits accessible to journalists regardless of their technical background. Ro demoed the tool and walked through how longitudinal auditing reveals shifts in chatbot responses, and we helped participants run their own audits on topics of their choosing, including opinions about the conflict in Iran, the election in Brazil and information about the upcoming US midterms.

New Research Spotlight

Event Recap: LLMs and the Third Wave of Automated Content Moderation

On September 10, we hosted the second convening of our LLMs and Civic Discourse project. Researchers, people building moderation tools and civil society advocates met to discuss how LLMs are changing social media content moderation.

Practitioners from Musubi, Zentropi and Roost discussed some benefits of LLMs for content moderation, and the importance of building open source tools in this space. Research presentations from Penn scholars (Yphtach Lelkes, Danaé Metaxa, and Neil Fasching) explored how these tools impact users once deployed, finding that different models often reach different decisions on the same content, and that identity-related speech is more likely to be wrongly flagged by LLM based moderation tools. Dia Kayyali and Prithvi Iyer closed with an open discussion around the human rights and policy implications of this turn to automation.

Read the Full Recap on the MEDIATED Feed ⟶

Research Seminars are BACK!

We kicked off our Fall series of research seminars with Professor Emily Thorson! Emily’s presentation challenged prevailing beliefs that Americans gravitate only to hyper-partisan news coverage, showing that there may be a more genuine interest in policy focused public-interest journalism. On October 14th, we also will host a special panel of journalists reporting on technology issues, who will discuss something most academic centers wonder: how to make empirical research accessible and informative to journalists! This and a lot more in the works.

And finally, our team is growing!

We are excited to continue the work of ensuring that empirical research on our information ecosystem drives real world change. For this, our team needs to grow, and we are elated to have two amazing colleagues join us this past month!

Pau Torres Pagès Research Analyst Pau Torres Pagès is our new Research Analyst who works across information integrity, global information flows, and human-centered AI. He will work on expanding our research compendium and produce information products that contextualize research to the broader policy and media context.

Manuela Buesa Student Communications Associate Manuela Buesa is our student Communications Associate who will build our social media presence, especially by launching the Center’s instagram and creating short form video content that makes our work accessible to a much wider audience!