Structuring and Activation of Knowledge for Intelligent Legal Systems in Smart Cities: Towards Transparent and Fair Justice

This paper proposes a conceptual architecture for structuring and activating legal knowledge in intelligent systems that could assist first-instance courts and digital public services. It asks how formal rules and ontologies can be combined with large language models while preserving human responsibility for legal interpretation and decisions. The analysis draws on the author’s earlier work in logic programming, selected literature on legal AI, and documented digital-justice initiatives in Europe, the United States, and China. Exploratory conversations with language models informed the design, but the paper does not report a working prototype, a comparative performance test, or measured effects on judicial outcomes. The proposed architecture has four linked functions: extracting facts from documents; representing relevant norms, exceptions, and case relationships; checking inferences against those representations; and presenting reasons, uncertainties, and an audit record to a human user. Language models could help process unstructured text and support dialogue, while rule-based components could make selected reasoning steps traceable. The comparison distinguishes cross-border information exchange and public-sector algorithmic oversight from actual automated adjudication. Estonia provides an example of digital infrastructure, but its Ministry of Justice has stated that it has not developed an AI judge for small claims. For the Republic of Moldova, the paper recommends a phased evaluation of narrowly defined assistance tasks, starting with transparent rules, verified data extraction, judicial review, and regular updates to the knowledge base. Open-textured norms, contested facts, bias, privacy, and errors in language-model output limit what formalization can establish. Any gains in consistency, processing time, cost, or public trust remain hypotheses that require case-level testing and independent assessment before deployment.