Sociotechnical Governance of AI-Driven Credit Systems: Integrating Fairness Metrics, Blockchain Transparency, and Stakeholder Dynamics in Traditional and Decentralized Finance

Abstract

This paper proposes a sociotechnical framework to address these issues by integrating fairness metrics, explainable AI (XAI), and game-theoretic models. We adapt statistical fairness criteria (demographic parity, equal opportunity) to audit bias, extend SHAP values to blockchain data for transparency in DeFi, and simulate stakeholder dynamics using agent-based models. Novel contributions include a governance-aware fairness metric that combines technical parity with stakeholder trust scores and a multi-layer agent model linking AI behavior to decentralized governance. Our findings reveal that DeFi systems exhibit narrower bias gaps than traditional systems but introduce new risks (e.g., collateral volatility), while profit-driven DAO governance often prioritizes short-term gains over systemic stability. This work advances interdisciplinary approaches to AI governance, emphasizing the need to reconcile technical robustness with social accountability.