Abstract: Building the DREAM-KG: A Community-Engaged Approach to Addressing Homelessness through Knowledge Graphs (Society for Social Work and Research 31st Annual Conference)

Schedule:

Friday, January 22, 2027

Golden Gate 4 (Hilton San Francisco Union Square)

* noted as presenting author

AnneMarie Tomosky, MPH, Research Project Manager, Temple University, Philadelphia, PA

Omar Martinez, JD, MPH, MS, Professor, University of Central Florida, Orlando, FL

Yuzhou Chen, PhD, MS, Assistant Professor of Statistics, University of California, Riverside, Riverside, CA

Chiu Tan, PhD, Associate Professor, Temple University, Philadelphia, PA

Huanmei Wu, PhD, Professor and Chair, Temple University, Philadelphia, PA

Background: Homelessness is a complex, multidimensional challenge shaped by intersecting social, economic, health, and policy factors, requiring innovative, cross-sector solutions. This study presents the Dynamic, Responsive, Engaged, Adaptive, and Multifaceted Knowledge Graph (DREAM-KG), a graph-based artificial intelligence (AI) system designed to enhance service navigation, coordination, and data integration across homelessness response systems. DREAM-KG integrates fragmented information across housing, healthcare, behavioral health, legal services, and community resources to improve access and continuity of care for individuals experiencing homelessness.

Methods: The system was developed using a community-engaged approach guided by a 20-member Community Advisory Board (CAB) composed of clinicians, nonprofit leaders, service providers, and policy advocates. Across three formal meetings and over 40 informal consultations, CAB members provided iterative input on system design, usability, and governance.

Results: Rapid qualitative analysis identified four key themes: (1) the need for expanded cross-sector partnerships to support sustainability; (2) the importance of positioning DREAM-KG as trust-building infrastructure, including features such as transportation information, weather alerts, trauma-informed language, and accessibility for users with low digital literacy; (3) concerns related to ethical data governance, including algorithmic bias, surveillance risks, and protection of sensitive information; and (4) challenges in integrating fragmented datasets, with recommendations for standardized data harmonization frameworks. CAB members also contributed to a replication and sustainability toolkit to support broader implementation.

Conclusion: Building on these findings, we introduce the DREAM-AI Engagement Framework to guide the development and scaling of community-engaged AI systems addressing complex social challenges. This work advances social work science by positioning the field at the forefront of ethical, community-informed AI innovation and has direct implications for policy and practice related to cross-sector data integration, governance, and equitable homelessness response systems.