The MISTRAL project develops a dynamic, AI-enabled toolkit for Health Impact Assessment (HIA), to anticipate the health and socioeconomic effects of environmental exposures in industrial urban areas. Central to MISTRAL is the co-creation of interactive HIA dashboards through participatory processes that actively involve citizens, policymakers, and local stakeholders. This approach ensures that decision-support tools are not only technically robust but also ethically grounded, socially legitimate, and context-sensitive. Drawing on environmental, socioeconomic, and clinical data processed through federated learning architectures, the platform enables privacy-preserving analytics while supporting evidence-based policymaking. Piloted across three European case studies - Taranto (Italy), Rybnik (Poland), and Hasselt/Genk (Belgium) - MISTRAL contributes to the development of a scalable Urban Health Impact Assessment platform. This paper positions MISTRAL within the field of citizen science, proposing co-creation as both a methodological and ethical framework for trustworthy AI in public health. It introduces a participatory governance model addressing contextual trust, inclusion, data governance, and actionable intelligence, and reports on the ethnographic, stakeholder engagement, and co-design activities implemented across the three pilot regions.
Rather than presenting a formal impact evaluation, the paper describes the methodological foundations and early implementation outputs of a citizen-informed approach to trustworthy AI in environmental health assessment.