60. Fairness and responsible machine learning
Measure unequal errors, representation gaps, accessibility harms, and feedback loops across affected groups. Compare fairness criteria, document tradeoffs, and involve stakeholders in deciding what acceptable performance means.