A model can look strong overall and still perform poorly for specific groups. Aggregate metrics often hide these gaps, making the system seem better than it actually is.

This is how fairness issues show up in practice. Not as obvious failures, but as uneven outcomes across different groups.

This chapter focuses on how bias enters ML systems, how to measure fairness, and what you can do to address it.

Note: This chapter is about fairness bias in ML systems, not the statistical bias–variance tradeoff. Statistical bias is about underfitting. Fairness bias is about systematic disadvantage to certain groups.