人身危险性评估的算法治理——从算法透明与商业秘密冲突展开  

The Algorithm Governance of Risk Assessment of Personal Dangerousness——Based on the Conflicts between Algorithm Transparency and Trade Secret

作　　者：王燃 孙艺桐 WANG Ran;SUN Yitong

机构地区：[1]天津大学法学院 [2]天津大学中国智慧法治研究院 [3]中国人民大学

出　　处：《上海政法学院学报（法治论丛）》2023年第3期94-115,共22页Journal of Shanghai University of Political Science & Law(The Rule of Law Forum)

基　　金：2022年度国家社会科学基金青年项目“回归技术现实主义的刑事证据数智化问题研究”的阶段性研究成果,项目编号:22CFX068

摘　　要：现代技术助推着人身危险性评估从经验化走向精算化、系统化和智能化,其在审前强制措施、量刑结果、刑罚执行等方面影响着司法决策。对于人身危险性评估算法,算法透明诉求的公开性与算法商业秘密保护的秘密性构成了相互矛盾的两个方面。算法透明内蕴的宪法权利、刑事诉讼权利与商业秘密保护的实体、程序规则产生了价值博弈。但两者利益之间并非不可调和,通过积极探索算法治理规则可完成价值平衡:在算法开发者层面,构建可解释的算法模型和算法信息分级公开制度,推动技术透明和程序透明;在算法使用者层面,坚持比例原则与信息公开,逐步完善权利归属约定;在被决策人层面,赋予该群体决策后救济渠道,保障知情权和异议权。三个层面的治理规则构成了符合技术正当程序的算法治理规则框架。Modern technologies have motivated the development of risk assessment of personal dangerousness from empiricalization to actuarialization,systematization,and intelligentization.Risk assessment algorithm has great influences on judicial decision-makings of different stages,including pretrial release,sentencing,and execution,etc..For risk assessment algorithms,the openness of algorithmic transparency claims and the secrecy of trade secret protection constitute two conflicting aspects.The constitutional rights and criminal proceeding rights embedded in algorithmic transparency and the protection of intellectual achievements by trade secret entities and procedural rules have also generated value games.However,the two interests are not irreducible,and the value balance can be accomplished by actively exploring algorithm governance rules.Algorithm developers could construct interpretable algorithm models and hierarchic disclosure system of algorithm information to promote technological and procedural transparency.Algorithm users should adhere to the principle of proportionality and information disclosure,and gradually improve the agreement of right attribution.As for people influenced by risk assessment algorithm,empowering them with post-decision relief channels and guaranteeing right to be informed and right to dissent are essential.The three levels of governance rules above constitute a framework of algorithm governance rules in line with technological due process.

关 键 词：人身危险性评估 商业秘密 技术正当程序 算法治理 

分 类 号：D915.3[政治法律—诉讼法学]