1 paper · 1 filter
Soheil Ghili, Ehsan Kazemi, Amin Karbasi
How can we control for latent discrimination in predictive models? How can we provably remove it? Such questions are at the heart of algorithmic fairness and its impacts on society…