算法崛起對法律規制提出了挑戰,它可能挑戰人類的知情權、個體的隱私與自由以及平等保護。作為人機交互的算法決策機制,算法并非價值中立,具備可規制性。算法公開、個人數據賦權、反算法歧視是傳統的算法規制方式,但機械適用這些方式面臨可行性與可欲性難題。算法公開面臨技術不可行、公開無意義、用戶算計與侵犯知識產權等難題,個人數據賦權面臨個體難以行使數據權利、過度個人數據賦權導致大數據與算法難以有效運轉等難題,反算法歧視面臨非機器算法歧視、身份不可能完全中立、社會平等難以實現等難題。傳統算法規制路徑面臨困境的根本原因在于忽視算法的場景性,算法可能因為運用算法主體的不同、針對對象的不同以及涉及問題的不同而具有不同的性質。因此,算法規制應采取場景化的規制路徑,根據不同場景類型對算法采取不同的規制方式,以實現負責任的算法為目標。在算法場景化規制原則的指引下,可以構建算法公開、數據賦權與反算法歧視等算法規制的具體制度。

The rise of algorithms poses challenges to their legal regulation as they may challenge people’s right to know and to have individual privacy and freedom and equal protection.As algorithmic decision-making mechanisms involving human-computer interaction,algorithms are not value-neutral and can be regulated.Algorithm disclosure,personal data empowerment,and anti-algorithmic discrimination are traditional methods of regulating algorithms,but the mechanical application of these methods encounters difficulties with feasibility and desirability.Algorithm disclosure faces problems such as technical non-feasibility, meaningless disclosure,user calculations and infringement of intellectual property rights;personal data empowerment faces the problems of individuals being unable to exercise data rights and of excessive personal data empowerment which impedes the effective operation of big data and algorithms;and anti-algorithmic discrimination faces the problems of non-machine algorithmic discrimination,the fact that one’s status cannot be completely neutral,the difficulty in achieving social equality,etc.The fundamental reason for the dilemma of traditional pathways for the regulation of algorithms lies in ignoring their context.Depending on their subjects,objects,and the different problems they involve,algorithms’ properties may vary.Therefore,algorithm regulation should adopt a scenario-based regulatory path and should employ varying regulatory methods depending on the type of scenario,with the aim of achieving responsible algorithms.The principle of scenario-based algorithm regulation can guide construction of specific regulatory systems such as algorithm disclosure,data empowerment and anti-algorithmic discrimination.