AI Bias
Systematic prejudices that a model learns from its training data.
AI bias occurs when a model discriminates or systematically errs regarding certain groups (gender, ethnicity, age). Origin: unbalanced data or reflections of social prejudices. Mitigations: balanced datasets, audits, fairness metrics.
Practical examples
- CV screening that penalizes women
- Facial recognition less accurate for people of color
- Discriminatory credit scoring models
- Differentiated medical suggestions