Self Reported Race

Self-reported race, a commonly collected demographic variable, is increasingly scrutinized for its implications in various fields, particularly healthcare and AI. Research focuses on understanding how readily machine learning models can infer race from seemingly race-neutral data (like medical records or names), highlighting privacy concerns and potential for algorithmic bias. Current efforts involve developing methods to mitigate this bias, including fairness-aware loss functions and disentanglement techniques within models like variational autoencoders and transformers, while also exploring alternative, less sensitive proxies for race like skin reflectance metrics. These investigations are crucial for ensuring fairness and privacy in AI applications and for accurately assessing disparities in healthcare outcomes.