A bank introduces an AI system to pre-screen loan applications.
The model was trained on historical data and reduces processing time by 40 %.
After one year, a pattern emerges:
• Self-employed applicants receive loans significantly less often.
• Salaried employees with comparable income are approved more frequently.
The AI uses the monthly income volatility as a strong indicator.
Self-employed people naturally have more irregular earnings.
Question:
Is this a legitimate risk assessment — or structural discrimination?
Solution follows tomorrow.