Conceptfeedback loop bias / 反馈循环偏差
Feedback Loop Bias
A systemic bias phenomenon in recommender systems and machine learning in which system outputs influence user behavior, and that behavioral data in turn reinforces the original system outputs, forming a self-reinforcing loop that leads to reduced diversity and accumulated bias.
Timeline (last 90 days)
Oct 8
Feedback loop bias in the recommender systems field is highly isomorphic to the model collapse mechanism, and intervention methods such as debiased sampling and adversarial exploration have been developed for it.
Unverified50%