Home / 13. Fairness / Why fairness matters

Why fairness matters

Automated decisions hit loans, benefits, and families. Accuracy alone is not enough when bias can be amplified.

Models trained on historical decisions can replay past discrimination. Even “neutral” quantitative pipelines can create new harms. Fairness is a design goal beside predictive performance.

Hiring screen

If past hiring favoured one group, a model trained to mimic past offers will keep favouring that group — unless you measure and constrain fairness explicitly.

Key takeaways

  • High stakes ⇒ ethical + technical duty.
  • Accuracy alone is incomplete.
  • Fairness needs explicit criteria.