Open Access Preprint · v1.0.0 · 1 mai 2026
Constitutional Oversight of Hybrid ML-RAG Credit Scoring Systems: Scalable Governance for Safe and Fair Financial Inclusion in Emerging Economies
Koffi Kan Marc-Ephrem · DOI : 10.5281/zenodo.19955522
Accessible Abstract
A three-tier oversight framework to enforce faithfulness and fairness in hybrid ML-RAG credit scoring systems: financial constitution, AI referee, and human-in-the-loop audit.
Hypothesis & Methodology
An algorithmic score computed by an ML model may be paired with a RAG-synthesized natural-language explanation that appears fluent yet deviates from true mathematical attributions. This preprint introduces a machine-readable financial constitution, an automated AI referee, and human-in-the-loop oversight to audit explanation faithfulness and demographic fairness.
Limitations & Real-world Applications
This framework is an academic research preprint. Empirical validation and independent third-party audits are essential prior to live lending deployment. The Fvundi project serves as an operational testbed for this research in explainable financial intelligence. Fvundi.