Abstract：Legal personalization uses data profiling, algorithmic prediction, and real-time feedback to transform abstract rules addressed to unspecified subjects into differentiated decisions and arrangements of obligations. This provides new technical conditions for addressing the overinclusiveness and underinclusiveness of general rules, improving the adaptability of legal norms, and achieving individualized justice. In the practice of algorithmic governance, however, platforms and public authorities can rely on precise identification, risk classification, and behavioral tracking to impose more targeted obligations and restrictions on specific subjects. Rights protection, by contrast, still depends largely on uniform provisions, standardized appeal procedures, and abstract declarations of rights. The allocation of obligations is becoming increasingly precise, while rights protection has not been refined to the same extent. This gives rise to a “reverse value” tendency in legal personalization and may cause it to depart from its original institutional purpose of pursuing individualized justice.This imbalance follows a relatively stable logic of formation. Historical data may absorb existing institutional biases and reproduce them through variable selection, model training, and risk classification. Differentiated decisions based on statistical correlations can also turn group characteristics into adverse inferences about particular individuals, causing correlation to be mistaken for causation. Technological and institutional black boxes further obstruct explanation, challenge, and review. The entrenchment of labels and liability by association in credit sanctions, the construction of a “high-risk” persona in predictive policing, behavioral manipulation and differentiated pricing in consumer settings, and the quasi-legislative structures created by platform rules all display similar asymmetries between rights and obligations.Traditional legality review focuses on legal basis, legitimacy of purpose, and the appropriateness of means, but pays insufficient attention to whether a reasonable relationship is maintained between the capacity for technological intervention and the capacity for rights protection. The proportionality principle provides a basic analytical framework of legitimacy of purpose, necessity, and the balance of interests. Symmetry review further brings the relationship between technological intervention capacity and rights protection capacity into the proportionality analysis. The legitimacy of purpose requires that burdens imposed on specific subjects be supported by reasons commensurate with the degree of specificity of those burdens, and that the relevant risks and their substantive connection to the differentiated measures be explained. The assessment of necessity should not only examine whether equally effective but less restrictive alternatives are available. It should also examine whether existing technological capabilities can be used at the same time to improve mechanisms for explanation, evidence presentation, and review, and should distinguish whether inadequate remedies result from “technical inability” or “commercial unwillingness”. Proportionality in the strict sense should compare the actual burden imposed on the individual with the demonstrable gains in governance. The more general the asserted public interest is, the more fully its importance, likelihood of realization, and connection to the differentiated measures need to be explained.A corresponding accountability framework should cover the entire process of algorithm design, operation, and remediation. Before an algorithm is put into operation, impact assessments and the retention of design documentation should be used to identify bias and potential impairment of rights. During the operation, algorithmic outputs should be continuously checked through dynamic monitoring, third-party audits, tiered transparency, and individualized explanations. After the harm occurs, substantive human review, a reasonable allocation of the burden of proof, and external judicial and administrative oversight should be used to change the pattern of self-review created by closed internal appeals on platforms. The legitimacy assessment of legal personalization should accordingly move from asking “whether differentiation is permissible” to asking “whether precision governance is matched by rights protection of equal precision”, so that the growth in technological capacity is translated into the corresponding growth in rights protection capacity.
陈林林, 杨璐萌. 比例原则视角下的算法个性化治理[J]. 浙江大学学报(人文社会科学版), 2026, 56(9): 101-114.
Chen Linlin, Yang Lumeng. Governance of Algorithmic Personalization from the Perspective of the Proportionality Principle. JOURNAL OF ZHEJIANG UNIVERSITY, 2026, 56(9): 101-114.
1 陈景辉：《作为法律本质的一般性》，《法学》2025年第7期，第3-20页。 2 德］克里斯托弗·布施：《个性化经济中的算法规制和（不）完美执行》，《环球法律评论》2019年第6期，第5-19页。 3 Coglianese C.， “Moving toward personalized law，” 2022-03-09， University of Chicago Law Review Online， https：//lawreview.uchicago.edu/online-archive/moving-toward-personalized-law， 2025-11-13. 4 Burk D. L.， “Algorithmic fair use，” University of Chicago Law Review， Vol. 86， No. 2 （2019）， pp. 283-307. 5 Casey A. J. & Niblett A.， “A framework for the new personalization of law，” University of Chicago Law Review， Vol. 86， No. 2 （2019）， pp. 333-358. 6 张凌寒：《算法权力的兴起、异化及法律规制》，《法商研究》2019年第4期，第63-75页。 7 Verstein A.， “Privatizing personalized law，” University of Chicago Law Review， Vol. 86， No. 2 （2019）， pp. 551-580. 8 汪怀君、汝绪华：《人工智能算法歧视及其治理》，《科学技术哲学研究》2020年第2期，第101-106页。 9 高进、刘聪：《算法歧视的敏捷治理——以D短视频平台为例》，《西安交通大学学报（社会科学版）》2024年第1期，第121-132页。 10 姜野：《算法的规训与规训的算法：人工智能时代算法的法律规制》，《河北法学》2018年第12期，第142-153页。 11 岳平、苗越：《社会治理：人工智能时代算法偏见的问题与规制》，《上海大学学报（社会科学版）》2021年第6期，第1-11页。 12 曹博：《算法歧视的类型界分与规制范式重构》，《现代法学》2021年第4期，第115-126页。 13 Wachter S. & Mittelstadt B.， “A right to reasonable inferences： re-thinking data protection law in the age of big data and AI，” Columbia Business Law Review， Vol. 2019， No. 2 （2019）， pp. 494-620. 14 Binns R.， “Fairness in machine learning： lessons from political philosophy，” Proceedings of Machine Learning Research， Vol. 81 （2018）， pp. 149-159. 15 郭伟栋、周志中、乾春涛：《手机App列表信息在信用风险评价中的应用——基于互联网借贷平台的实证研究》，《中国管理科学》2022年第12期，第96-107页。 16 谭九生、范晓韵：《算法“黑箱”的成因、风险及其治理》，《湖南科技大学学报（社会科学版）》2020年第6期，第92-99页。 17 张欣：《算法影响评估制度的构建机理与中国方案》，《法商研究》2021年第2期，第102-115页。 18 Habermas J.， “Three normative models of democracy，” Constellations， Vol. 1， No. 1 （1994）， pp. 1-10. 19 彭錞：《论信用监管的制度逻辑及其合法性与有效性》，《华东政法大学学报》2024年第3期，第39-54页。 20 戴昕：《信用法治研究：“瓶颈”与前景》，《地方立法研究》2025年第2期，第39-56页。 21 印子：《乡村社会诚信建设中的连带责任机制》，《中国农村观察》2022年第6期，第38-55页。 22 马龙：《“司法+大数据”控制视域下失信被执行人信用修复研究》，《法律适用》2023年第2期，第58-68页。 23 陈永生：《大数据预测警务的运作机理、风险与法律规制》，《中国法学》2024年第5期，第165-184页。 24 Van Brakel R. E. & Govaerts L.， “Exploring the impact of algorithmic policing on social justice： developing a framework for rhizomatic harm in the pre-crime society，” Theoretical Criminology， Vol. 29， No. 1 （2025）， pp. 91-109. 25 张全涛：《大数据警务算法规制：基于正义理论的视角》，《中国人民公安大学学报（社会科学版）》2025年第3期，第95-113页。 26 郑玉双：《人工智能嵌入法治：智慧司法的法理反思》，《江海学刊》2024年第3期，第172-180，256页。 27 马长山：《算法治理的正义尺度》，《人民论坛·学术前沿》2022年第10期，第68-76页。 28 郭明瑞、娄逸骅：《算法消费者信息茧房的治理困境与制度因应》，《深圳大学学报（人文社会科学版）》2025年第4期，第86-95页。 29 苏新建、沈运峰：《个别化价格歧视的规制难题及其纾解——基于“禁止差别待遇”规则的分析》，《商业经济与管理》2022年第11期，第60-70页。 30 宋玉茹：《平台算法推荐下消费者理性决策的理论依据与实践路径》，《价格理论与实践》2025年第6期，第106-111页。 31 唐惠敏、范和生：《论互联网平台治理规则体系的构建》，《出版发行研究》2024年第12期，第82-90页。 32 王鹏飞：《电商平台算法歧视法律救济的困境与出路》，《科学决策》2023年第5期，第213-223页。 33 中国司法大数据研究院课题组：《大数据视角下社交平台涉诉问题的司法应对》，《数字法治》2024年第6期，第158-174页。 34 刘权：《目的正当性与比例原则的重构》，《中国法学》2014年第4期，第133-150页。 35 Kroll J. A.， Huey J. & Barocas S. et al.， “Accountable algorithms，” University of Pennsylvania Law Review， Vol. 165， No. 3 （2017）， pp. 633-705. 36 Pasquale F. A.， “Platform neutrality： enhancing freedom of expression in spheres of private power，” Theoretical Inquiries in Law， Vol. 17， No. 2 （2016）， pp. 487-513. 37 王丽洁：《个人信息处理中比例原则审查基准体系的建构》，《法学》2022年第4期，第49-63页。 38 Kaminski M. E. & Urban J. M.， “The right to contest AI，” Columbia Law Review， Vol. 121， No. 7 （2021）， pp. 1957-2048. 39 吕炳斌：《论个人信息处理者的算法说明义务》，《现代法学》2021年第4期，第89-101页。 40 Gritsenko D. & Wood M.， “Algorithmic governance： a modes of governance approach，” Regulation & Governance， Vol. 16， No. 1 （2022）， pp. 45-62. 41 张凌寒：《算法评估制度如何在平台问责中发挥作用》，《上海政法学院学报（法治论丛）》2021年第3期，第45-57页。 42 清华大学人工智能国际治理研究院：《我国算法治理政策研究报告》，2022年10月，https：//aiig.tsinghua.edu.cn/info/1025/1759.htm，2025年10月20日。 43 张欣、宋雨鑫：《算法审计的制度逻辑和本土化构建》，《郑州大学学报（哲学社会科学版）》2022年第6期，第33-42页。
2026, Vol. 56 
