assume an unbiased predictive model. In this paper, we address the fact that prediction bias can be nonnegligible in large VA … portfolio valuation and investigate the impact of prediction bias in both the modeling and sampling stages of active learning …. Our experimental results suggest that bias-based sampling can rival the efficacy of traditional ambiguity-based sampling …

that may arise from ML predictions. We define a new source of bias related to incompleteness in real time inputs, which may … bias. Our observational and experimental analyses in the patent examination context support this conjecture. In the face of …

algorithms are informed by data, if these are corrupted, from a social bias perspective, good machine learning algorithms would …