This closed-loop degradation is driven by a foundational flaw in algorithmic auditing: the unchecked amplification of systemic bias. When machine learning models train on synthetic data generated by ...