Fact-Grounded Explainable Analytics for Rural Governance

Pre-print version. Rural development is underrepresented in e-Government research, where planning decisions rely on fragmented data, opaque tools, and limited explanatory support. We present a framework that unifies heterogeneous geospatial and socio-economic data to generate verifiable, human-readable explanations for rural policy analysis through an auditable pipeline combining indicator computation, spatial aggregation, and fact-grounded language generation. We introduce a formal model that constrains explanation generation through fact selection, enabling quantifiable coverage, utility, and factual precision. Evaluation across six European regions shows (i) stable composite indicators under up to 30% missing data, (ii) high factual precision (0.93), and (iii) broader explanatory coverage than existing reporting sources. A user study with municipal practitioners shows improved comprehension accuracy and reduced interpretation time. This positions the proposed approach as a step toward treating AI-generated outputs in public-sector systems as verifiable evidence rather than opaque or advisory text.