摘要
随着人工智能深度赋能刑事司法,刑事司法中产生了算法治理的新问题.算法解释是刑事司法中算法治理的核心内容之一,但面临三重制度选择:在技术层面,算法的可解释性优于性能,应在算法设计之初植入可解释的理念,选择相对简单的算法模型;在规范方面,权利规范比禁令规范更契合刑事司法的场景化特点,应以权利作为主要的规范形式,在个别场景中禁止算法决策;在主体方面,国家机关的算法解释义务居于主导性地位,主要以算法模型作为解释对象,应健全事前、事中、事后的配套措施;公民个体的算法解释权居于回应性地位,能够保障公民对具体算法决策的知情、理解与救济,应建构相应的运行流程.
Abstract
With the deep integration of artificial intelligence into criminal justice,new issues of algorithmic governance have emerged in this field.Algorithmic interpretability stands as one of the core components of algorithmic governance in criminal justice,yet it confronts a tripartite institutional choice.At the technical level,the interpretability of algorithms takes precedence over their performance,necessitating the incorporation of interpretability principles into the initial stages of algorithm design and opting for relatively simple algorithmic models.In terms of norms,rights-based norms are more compatible with the contextualized characteristics of criminal justice than prohibitive norms,advocating for rights as the primary normative form while prohibiting algorithmic decision-making in specific scenarios.Regarding the subject,the algorithmic interpretation obligations of state organs occupy a dominant position,primarily focusing on algorithmic models as the objects of interpretation,and requiring the improvement of supporting measures before,during,and after the process.Meanwhile,the algorithmic interpretation rights of individual citizens occupy a responsive position,ensuring citizens'right to know,understand,and seek redress for specific algorithmic decisions,necessitating the establishment of corresponding operational procedures.
关键词
刑事司法/算法解释/算法解释权/算法解释义务Key words
criminal justice/algorithmic interpretation/right to algorithmic interpretation/obligation of algorithmic interpretation引用本文复制引用
出版年
2024
NETL
NSTL
万方数据