From environmental data to equitable action: E-PHIA, an equity-centred framework for AI-enabled public health intelligence

Environmental health inequalities arise from interactions among environmental hazards, social disadvantage, population susceptibility and unequal capacity to respond. Public-health systems increasingly generate environmental, demographic, geospatial and health data, yet these information streams are often fragmented across surveillance and decision-making systems. Geospatial artificial intelligence (GeoAI) creates opportunities to integrate complex spatial data and identify patterns of environmental and population-health risk, but technical capability alone does not ensure equitable or actionable public-health decision-making. This conceptual paper proposes E-PHIA (Equity-Centred Public Health Intelligence with AI), an evidence-informed framework for translating environmental, population-health and social-vulnerability data through responsible AI and GeoAI into geographically actionable intelligence. E-PHIA comprises six connected layers: environmental intelligence; population-health and social-vulnerability intelligence; data integration and quality; responsible AI/GeoAI; actionable public-health intelligence; and action, evaluation and learning. Equity, explainability, fairness, privacy, governance and human oversight operate as cross-cutting safeguards. The framework introduces Equity Checkpoints across the intelligence lifecycle and distinguishes model performance from decision performance and population impact. It also proposes a conceptual prioritisation function in which public-health priority depends on hazard, exposure, vulnerability and actionability rather than environmental exposure alone. E-PHIA positions AI as an augmentation of professional public-health judgement, not a substitute for it. Empirical work is required to test data feasibility, technical and spatial validity, subgroup fairness, decision utility and population impact. The framework offers a testable pathway for moving from fragmented environmental data and retrospective reporting towards earlier risk identification, geographically targeted prevention and continuous evaluation of whether public-health action narrows or widens inequalities.Keywords: artificial intelligence; GeoAI; environmental health; public health intelligence; health inequalities; health equity; environmental epidemiology; public health surveillance; responsible AI; climate health