The Promises and Perils of using LLMs for Effective Public Services

要旨

Governments are the primary providers of essential public services and are responsible for delivering them effectively. In high-stakes decision-making domains such as child welfare (CW), agencies must protect children without unnecessarily prolonging a family’s engagement with the system. With growing optimism around AI, governments are pushing for its integration but concerns regarding feasibility and harms remain. Through collaborations with a large Canadian CW agency, we examined how LocalLLM and BERTopic models can track CW case progress. We demonstrate how the tools can potentially assist workers in opportunistically addressing gaps in their work by signaling case progress/deviations. And yet, we also show how they fail to detect case trajectories that require discretionary judgments grounded in social work training, areas where practitioners would actually want support to pre-emptively address substantive case concerns. We also provide a roadmap of future participatory directions to co-design language tools for/with the public sector.

著者
Erina Seh-Young Moon
University of Toronto, Toronto, Ontario, Canada
Matthew Tamura
University of Toronto, Toronto, Ontario, Canada
Angelina Zhai
Georgia Institute of Technology, Atlanta, Georgia, United States
Nuzaira Habib
University of Toronto, Toronto, Ontario, Canada
Behnaz Shirazi
Child Welfare Institute, CAS of Toronto, Toronto, Ontario, Canada
Altaf Kassam
Child Welfare Institute, CAS of Toronto, Toronto, Ontario, Canada
Devansh Saxena
University of Wisconsin-Madison, Madison, Wisconsin, United States
Shion Guha
University of Toronto, Toronto, Ontario, Canada

会議: CHI 2026

ACM CHI Conference on Human Factors in Computing Systems

セッション: Generative AI in Design and Practice

P1 - Room 116
7 件の発表
2026-04-17 20:15:00
2026-04-17 21:45:00