Bias mitigation is the set of methods used to identify, measure, reduce, or manage unfair or harmful bias in AI systems. It may involve changes to data, model training, evaluation metrics, human oversight, deployment controls, or post-processing of outputs.