AI for Public Benefit Lab

This project examines how generative artificial intelligence can be responsibly used to address public-interest challenges, with a particular focus on social service delivery. One goal is to develop a clear roadmap for researchers, practitioners, and policymakers that outlines how to determine good opportunities for generative AI, key risks of using generative AI in public interest settings, and how to implement solutions. This roadmap is grounded in the project’s accompanying paper, "Addressing Public Problems with Generative AI," which also reviews emerging applications of generative AI in high-stakes public settings.

A second goal is to apply this framework to the delivery of social services such as SNAP, TANF, and community-based assistance. In particular, we are interested in how generative AI can reduce administrative burden, wait times, and errors by supporting both clients and frontline staff. This includes client-facing tools that help people navigate complex benefits systems and connect with appropriate resources, as well as caseworker co-pilots that assist with policy interpretation, eligibility determinations, and application processing. We are in conversations with state governments, non-profits, and public benefits corporations about building and/or evaluating generative AI tools for social service delivery.