AI Bias & Algorithmic Discrimination

Why AI systems can discriminate systematically and what's being done about it.

10 min·4 sections·2 quiz questions

How Bias Enters AI Systems

AI learns from historical data. If historical data reflects discrimination — and most does, because human history is full of discrimination — AI systems will learn and perpetuate those patterns. A hiring algorithm trained on resumes of successful employees at a historically male-dominated company will learn to prefer male-coded features. A criminal risk assessment tool trained on arrest data will reflect racially biased policing patterns.

Biased input data produces biased AI output — garbage in, garbage out, at scale.

Key Takeaways

✓AI systems learn and amplify the biases present in their training data
✓Algorithmic discrimination has caused documented harm in criminal justice, hiring, and healthcare
✓Black-box AI in consequential decisions violates principles of fairness and accountability
✓Regulation and advocacy are essential complements to technical solutions