CosmicAI Faculty at Responsible AI in 2025 Conference
The virtual two-day Responsible AI in 2025: State of the Art and Future Directions Conference, hosted by Good Systems at the University of Texas at Austin, brought together researchers and AI leaders to assess where Responsible AI stands today and where it needs to go next. Sessions explored research-informed perspectives on work, robotics, smart cities, and the information environment, while underscoring the need for practical safeguards such as operational controls, policy constraints, human oversight, and transparency for impacted communities. As Meg Young of Data & Society highlighted, responsible deployment is not a simple yes-or-no choice—it requires thoughtful sociotechnical design.
CosmicAI faculty members Matt Lease, Jessy Li, and Greg Durrett contributed to these conversations, adding expertise on trustworthy and interpretable AI. Across the event, participants emphasized that interdisciplinary and cross-sector collaboration is essential for building accountable, human-centered AI systems.
In his welcome remarks and keynote introduction, Matt Lease, co-Director of CosmicAI, framed the two-day virtual conference as a chance for researchers and AI leaders to assess both progress and persistent challenges in Responsible AI. He noted the rapid growth of AI capabilities and their increasing use in high-impact settings, emphasizing the need to capture benefits while carefully managing risks. He underscored Good Systems’ view that Responsible AI is a societal challenge requiring input from many stakeholders, not only technologists, and highlighted the importance of grounding research in real-world problems to reveal gaps in capability and governance. By focusing on these gaps, he explained, Responsible AI research can better support specific societal needs while improving practices for the broader public good.
From bottom right moving left: Maria De-Arteaga, Associate Professor, Data, Analytics, Technology, and Artificial Intelligence (ESADE), Greg Durrett, Associate Professor, Computer Science (NYU), Moderator: Matt Lease, Professor, School of Information (UT), and Jessy Li, Associate Professor, Linguistics (UT).
In a panel titled, Responsible AI and the Information Environment, Matt Lease, along with CosmicAI faculty researchers Jessy Li and Greg Durrett (NYU), as part of their project, Designing Responsible AI Technologies to Protect Information Integrity, focused on how access to reliable information underpins personal decisions, public health, financial stability, and democratic processes, and how misinformation can quickly erode all of these. They pointed to real-world cases, from COVID-19 confusion to natural disasters and foreign influence campaigns, to show how unreliable information can trigger broad harm. They also noted that AI plays a dual role: it can strengthen the resilience of the information environment, but it also introduces new risks, since generative systems can produce convincing but inaccurate content or be misused to create deepfakes. To conclude, they emphasized that AI can help analysts, journalists, and public agencies work more effectively and monitor information spaces at scale, offering faster detection and response when new threats emerge.