作者：单勇
作者介绍：南京大学法学院教授
出版日期：2026年4月9日
内容提要 犯罪之算法治理已成为平安中国建设的基本方略，其应用覆盖地点评分、维稳风控、大数据扫黄、网络犯罪风控等场景。以“监视—筛选”为特质的犯罪治理算法在极大提升治理能力的同时，也存在法律依据不健全、对个体生活形成挤压和算法权力失范等问题。为推动算法治理从“基于技术的治理”转向“基于规则的治理”，应引入以结果公平和过程公平为内涵的算法公平原则，作为建构“算法之法”的正当性基础。依循回应性规制和预防性规制路径，通过算法救济和问责等回应性机制修复算法偏误对结果公平的侵蚀，依托算法影响评估和算法公开等预防性机制防止算法失范对过程公平的偏离，以促进算法向善、保障算法治理始终运行于法治轨道。
关键词 犯罪治理算法 算法治理 算法公平 回应性规制 预防性规制
Abstract: The algorithmic governance of crime has become a fundamental strategy of the Peaceful China Initiative, with applications covering scenarios such as location scoring, stability maintenance risk control, big-data-based crackdown on prostitution, and cybercrime risk control. The crime governance algorithm characterized by ‘monitoring and screening’ has significantly enhanced the ability of crime governance. However, it has also led to problems such as an incomplete legal basis, intrusion into individual lives, and misconduct of algorithmic power. The principle of algorithm fairness with both consequential and procedural dimensions shall be the legitimacy basis of algorithm regulation to promote the transition of algorithmic governance of crime from ‘technology-based governance’ to ‘rule-based governance’. This should follow a dual path of responsive regulation and preventive regulation. Responsive regulation can correct algorithmic errors to prevent the erosion of result fairness through algorithmic relief and accountability. Preventive regulation can prevent deviations of algorithms from process fairness through algorithmic impact assessment and algorithmic transparency. They develop algorithms for good and guarantee that algorithmic governance consistently operates within the framework of the rule of law.