Lee este artículo en voz alta
Enciende el micrófono y lee a tu ritmo. Las palabras se iluminan sobre la marcha y nadie te detiene por un error.
Tu audio nunca llega a MovaReader: lo reconoce tu navegador. Política de Privacidad
Inicia sesión para guardar esta sesión en tu historial de lectura.
Digital Ethics and Algorithmic Bias
Priya wanted the hiring tool to be fair in two ways at the same time. She wanted each group of applicants to be selected at the same rate. She also wanted the tool to make the same kinds of mistakes in each group: the same share of suitable people rejected, and the same share of unsuitable people accepted. Both demands sounded like plain justice. She asked the engineer to satisfy both.
The engineer could not, and the reason was not a missing line of code. The records the model had learned from did not show the same share of suitable applicants in every group. When those base rates differ, the two demands pull apart. Equal selection rates are one definition of fairness, often called demographic parity. Matching those mistake rates across groups, the same true