算法歧視的類型界分與規制範式重構

The Type Classification of Algorithmic Discrimination and the Reconstruction of Regulatory Paradigm

對算法歧視進行一體化規制的預設失之抽象，個別化規制的探討則受限於部門法思維，過於宏觀與過於微觀的視角都存在各自盲點，基於類型界分的規制範式重構嘗試探索一條折中路徑。承認算法歧視在根源上與傳統歧視具有同質性是進行分類的前提，複現型算法歧視、加劇型算法歧視及新增型算法歧視是在比對算法決策與傳統決策的基礎上實現的類型化整合，以此為基礎分析規制歧視的傳統範式也顯得有的放矢。面對不同類型算法歧視存在的技術難題與法理障礙，進一步細化具體類型中的特殊情形，提煉規制重點，將深化對算法歧視的認知，實現以制衡算法權力為核心的規制範式重構。

"The presupposition of integrated regulation of algorithmic discrimination is too abstract, while the exploration of individualized regulation is limited by sectoral law thinking. Both too macroscopic and too microscopic perspectives have their own blind spots, and the reconstruction of the regulatory paradigm based on type classification tries to provide a compromise path. The recognition that algorithmic discrimination is homogeneous at its roots with traditional discrimination is a prerequisite for classification, and that reproducible, aggravated and additive algorithmic discrimination is a typological integration based on a comparison of algorithmic decision making and traditional decision making, which is also a useful basis for analyzing traditional regulatory paradigms for discrimination. In the face of the technical difficulties and jurisprudential obstacles of different types of algorithmic discrimination, further concretizing the specific situations in different types and refining the regulatory priorities will deepen the knowledge of algorithmic discrimination and realize the reconstruction of regulatory paradigm with checks and balances on algorithmic power as the core."