yakew7/Fair-Code — reverse-engineered prompt

Build me a project called Fair Code that exposes bias in AI decisions and shows how to reduce it with real data. I want a simple public website that feels like an educational audit dashboard, with five case studies: COMPAS criminal justice, AI hiring, German credit lending, insurance denial, and benefits denial.

Each case should explain the harm in plain English, show the protected group being tested, show the unfair result, then show the improved result after removing protected attributes or proxy variables. Include clear charts or tables for the fairness gap before and after, plus short explainers for concepts like proxy variables, demographic parity, calibration, equalized odds, and SHAP values.

Also create runnable Python audit scripts and notebook walkthroughs for each case so someone can reproduce the results. Keep it approachable, but make it credible for researchers, students, and activists. Include a landing page, citations, contribution info, and basic automation that can run the audits. If you need current docs, look them up online.