Credit scoring

Credit scoring is the algorithmic assessment of a borrower's likelihood of repaying a loan, producing a numerical score that determines access to credit, housing, and sometimes employment. The practice appears to be a neutral risk assessment tool, but it encodes historical patterns of discrimination into mathematical form. Variables like zip code, educational background, and employment history function as proxies for race and class, allowing models to replicate redlining without explicitly referencing protected categories.

The systems-theoretic critique is that credit scoring does not merely predict risk — it manufactures it. By excluding certain populations from credit markets, the algorithm concentrates poverty and reduces the very economic mobility that would lower default rates. The model then interprets this concentrated poverty as validation of its risk predictions. Like predictive policing, credit scoring operates as a WMD because the feedback loop between prediction and outcome is closed, and the subjects of the model cannot see or contest the reasoning that shapes their life chances.

The opacity of credit scoring models is typically defended as proprietary trade secrets, but this defense masks a deeper structural issue: the scoring system is not a measurement of an independent reality but an active force in constructing economic stratification. The mathematical veneer of objectivity conceals the fact that every variable is a choice about what counts as creditworthiness, and those choices reflect the power relations of the financial system they serve.\n\nThe relationship between credit scoring and redlining is not merely historical analogy but structural continuity. The shift from explicit geographic exclusion to algorithmic proxy discrimination is what scholars call algorithmic redlining — a phenomenon that reveals the inadequacy of legal frameworks designed to combat 20th-century discrimination for 21st-century mathematical harm.