FRL: A Normative Agent Layer for Liability-Aware Administrative Decision Support

The FRL framework introduces a norm-aware agent architecture designed to support administrative decision processes under complex legal and institutional constraints.

The system implements a machine-readable governance layer that integrates legal norms, policy rules, and procedural safeguards into a structured decision-support workflow. Instead of relying on purely mathematical scoring, FRL incorporates normative semantics derived from legal frameworks (e.g., administrative law, fundamental rights protections, and emerging AI governance regimes) to guide decision routing and option evaluation.

The architecture is implemented as a modular Python and Streamlit prototype using JSON-based rule objects, policy hooks, and webhook connectors to external systems. It is designed as a decision-support layer rather than an automated decision engine, ensuring compatibility with regulatory constraints concerning automated decision-making and human oversight.

The repository contains a reference implementation, schema definitions, and an extensible rule architecture intended for experimentation by public institutions, researchers, and governance engineers exploring norm-aware decision infrastructures.

https://bw-ruah.de/