Уровень 0 · материалов: 2
В кластер входят документы о влиянии расовых факторов на работу автоматизированных систем и алгоритмов, и не входят тексты о статистическом анализе смертности в результате действий полиции.
Общие признаки: расовая предвзятость, анализ данных, системное неравенство, алгоритмическая дискриминация
Группа выше: Алгоритмическая дискриминация и ложные метрики
Смысл: The main idea is that predictive algorithms used in the US judicial system to assess recidivism risk exhibit systemic racial bias against African Americans, despite claims of neutrality, due to the use of racially correlated proxy variables in a proprietary, non-transparent system.
An investigation revealed that Northpointe's risk-assessment software used by US courts is racially biased and inaccurate, disproportionately flagging Black defendants as high-risk.
Смысл: The main idea is that AI can detect human race from medical imaging with high precision through patterns invisible to humans, which highlights a dangerous potential for hidden racial bias in automated healthcare and other critical systems.
AI can identify a person's race from X-rays with up to 98% accuracy, raising urgent concerns about hidden algorithmic bias and systemic discrimination in medical AI.