Safety, Ethics & Governance · intermediate · concept 169 of 176

🛡️AI Bias & Fairness

AI systems can perpetuate and amplify societal biases present in training data. Understanding, measuring, and mitigating bias is critical for responsible AI deployment. Formal fairness metrics (demographic parity, equalized odds, predictive parity) famously conflict: satisfying all of them at once is mathematically impossible in most real settings.