回饋迴圈偏見放大

Risk Description

AI system outputs are reincorporated into training data or influence future data collection, creating feedback loops that amplify bias. For example, predictive policing concentrates resources in specific neighborhoods, increasing crime data there, further reinforcing biased predictions. This self-reinforcing mechanism exponentially amplifies initial small biases over time, ultimately causing severe systemic discrimination.