Data Representations Role in Intersectional Gender Bias

AI systems learn from data, and if datasets lack diversity, especially regarding intersecting identities, the AI will inherit and amplify these biases. For instance, facial recognition tools have shown higher error rates for women of darker skin tones compared to white men or white women, as these groups are underrepresented in training data, leading to skewed gender bias outcomes.