A team of interdisciplinary scholars at Columbia University has unveiled a framework that reimagines how artificial intelligence (AI) can be ethically and effectively used in community-engaged public health research. Published in Implementation Science, “The Practical, Robust Implementation and Sustainability (PRISM)-capabilities model for use of Artificial Intelligence in community-engaged implementation science research” introduces a human-centered approach to integrating AI into complex, real-world settings. Led by University Professor Dr. Nabila El-Bassel of Columbia School of Social Work, and Dr. Tian Zheng, Chair of the Department of Statistics, the paper develops a framework for improving data collection, analysis, insight, and engagement with AI, while preserving research ethics.
The framework builds on the HEALing Communities Study, the largest federally funded implementation science effort to reduce opioid overdose deaths in the U.S., with a model that demonstrates how AI can be used not only to analyze large datasets but also to co-design, monitor, and adapt interventions with communities in real time. The PRISM-Capabilities model centers on six interconnected components ranging from equity audits to ethical oversight that equip researchers and community coalitions with actionable tools such as natural language processing, simulation modeling, and explainable AI methods. By prioritizing transparency, shared decision-making, and inclusivity across all phases of research, the PRISM-Capabilities model offers researchers, funders, and policymakers a concrete roadmap for embedding AI within socially just, community-driven science.
This work also anchors Columbia’s newly launched AI for Social Good and Society Initiative, an ambitious, university-wide effort to ensure AI innovations reflect community values, promote equity, and drive measurable improvements in public health and social work and related field.
To learn how the model is being applied to validate AI tools across real-world community data, read the full article at Implementation Science.