算法偏见及其规制路径研究

算法偏见问题引发了广泛关注,不同程度损害公众基本权利、经营者竞争性利益和特定个体的民事权益,亟需规制。通过分解机器学习过程可知,算法偏见萌芽于数据收集步骤,成熟于模型完善步骤,强化于模型应用阶段。算法偏见的规制,需从社会减少偏见,保证数据可查性与算法可审计性,对算法使用者与设计者课以相应义务,以公平、透明和可责的算法确保算法的规范应用。

Research on Algorithm Bias and Its Regulation Approach

The problem of algorithm bias has aroused widespread social concern, which has harmed the general interests of the public to some degree, the competitive interests of operators and the legitimate rights and interests of specific individuals, so it is necessary to regulate the algorithm bias. By decomposing the machine learning process, the algorithm biases sprout in the data collection step, mature in the model perfect step, and strengthen in the model application stage. To solve the problem of algorithmic bias, we need to minimize prejudice from the society itself, to ensure data accessibility and algorithm audit, the class with the algorithm users and designers corresponding obligations, and the relevant legal regulations to standardize the use of the algorithm, fair, transparent and accountable algorithm to ensure the standard application of algorithm ,and effectively protect the interests of relevant.