Discrimination on AI and Society Course

10.1 Alternative Data Credit Scoring

Thu, 21 May 2026 00:00:00 +0000

Alternative Data Credit Scoring

Context & Systems Architecture

Traditional credit underwriting frameworks (such as FICO scores) rely on structured financial histories, including repayment logs, existing debt levels, and credit card histories. However, across the fintech and subprime lending landscape, algorithmic credit models have emerged that bypass these traditional guardrails. FinTech platforms and neo-banks utilize non-traditional machine learning classifiers to score the creditworthiness of “credit invisible” or underbanked individuals. These systems ingest digital behavior footprints to predict the statistical probability of loan default, deeply complicating standard consumer protection laws.

6.1 AI Resume Screening & Hiring Discriminators

Thu, 21 May 2026 00:00:00 +0000

AI Resume Screening & Hiring Discriminators

Context & Systems Architecture

Automated hiring and talent acquisition tools are widely used by corporate Human Resources departments to filter through thousands of applicant resumes. Between 2014 and 2017, Amazon developed a proprietary, secret AI recruiting engine designed to automatically rank job applicants and identify top talent. The project was ultimately abandoned after internal engineers discovered that the machine learning pipeline had developed a systematic, algorithmic hostility toward female candidates, uncovering the core vulnerability of training models on historical human decision-making data.