Algorithmic bias is systematic unfairness in AI outputs across groups, a hiring model favoring one demographic, a credit model penalizing certain neighborhoods,
Systematic unfairness across groups usually inherited from training data that encodes historical patterns, or from proxy features that correlate with protected attributes, for example a hiring model favoring one demographic or a credit model penalizing certain neighborhoods.
No, only measured and partially mitigated with fairness metrics, balanced data, and constraint techniques. Fairness definitions conflict mathematically, so choosing which to satisfy is a human, documented, and increasingly legal decision rather than an automatic fix.