Use of Fairness Metrics in Evaluating Bias

Fairness metrics have become a popular method for assessing bias in training datasets. By providing quantifiable measures to evaluate bias, they offer a clear baseline for comparison. However, the inherent limitation of fairness metrics is their dependency on the chosen metric; different metrics can provide vastly different assessments of bias for the same dataset. Thus, while useful, fairness metrics must be chosen and interpreted carefully to effectively reflect bias.