Duke Team Receives $100,000 Laude Moonshots Award to Make Social Media More Civil

What if coordinated AI interventions could heal the internet’s civility problem? A team led by Jun Yang, the Knut Schmidt Nielsen Distinguished Professor of Computer Science, received an honorable mention from the Laude Moonshots competition, along with a $100,000 grant to develop an AI-powered system for improving civic discourse on social media.  

Yang’s team brings together a multidisciplinary group of Duke faculty. Christopher Bail, Professor of Sociology, and Alexander Volfovsky, Associate Professor of Statistical Science, co-direct the Duke Polarization Lab; Yang, along with Bhuwan Dhingra, Assistant Professor of Computer Science, and Brandon Fain, Assistant Professor of the Practice of Computer Science, contribute expertise in database systems, natural language processing, and algorithmic fairness; and Cynthia Rudin, the Gilbert, Louis, and Edward Lehrman Distinguished Professor of Computer Science, adds a specialization in interpretable machine learning. 

The team’s approach is motivated by a growing body of social science evidence showing that existing moderation strategies — from removing extreme content to adding warning labels — tend to be short-lived or counterproductive when applied in isolation. Research by members of the Polarization Lab has shown that simply exposing social media users to opposing viewpoints can sometimes increase, rather than reduce, polarization, underscoring the need for more carefully coordinated, personalized interventions. 

The Moonshots grant will support the development of Orchestrate+, an open-source protocol and infrastructure for coordinating AI-powered interventions on social network platforms. Rather than applying a single, blanket fix, Orchestrate+ uses a two-tier architecture: a layer of AI agents that interact with individual users through tailored interventions — such as recommending content from other perspectives, prompting users to reconsider inflammatory phrasing before posting, or surfacing bridging content approved across political lines — and a coordinating “conductor” that orchestrates these agents and optimizes for healthier discourse at the network level, while maintaining user engagement and ensuring fairness across users. 

The $100,000 grant will fund an initial phase of development. Later in 2026, Laude will convene a dedicated showcase for Moonshots runners-up and honorable mentions in front of funders assembled specifically to support these projects. 

The Laude Institute backs computer science researchers turning research into real-world impact, combining grantmaking with hands-on support from early-stage ideas to open-source projects and multi-year labs. Yang’s team was among 25 groups recognized among 125 Moonshots proposals submitted by more than 600 researchers at 47 leading institutions across the United States and Canada. The selection committee was chaired by David Patterson, a Turing Award winner and founding board chair of Laude Institute, and included Nobel laureate John Jumper, Turing Award winner John Hennessy, and Jeff Dean, chief scientist of Google DeepMind.