Demographic Parity

Demographic parity, a fairness metric in machine learning, aims to ensure that algorithms produce equal outcomes across different demographic groups, mitigating biases present in training data. Current research focuses on developing algorithms and post-processing techniques that achieve demographic parity while maintaining prediction accuracy, often employing methods like optimal transport and Wasserstein barycenters, or leveraging large language models for improved performance in complex tasks. This work is crucial for addressing algorithmic bias in high-stakes applications such as loan applications, hiring, and criminal justice, promoting fairer and more equitable outcomes.