Generative AI生成司法决策的可靠性困境及其应对  

A Study of the Reliability Dilemma of Generative AI in Judicial Decision-Making and Its Countermeasures

作　　者：武振国 周健 Wu Zhenguo;Zhou Jian(School of Humanities and Social Science,University of Science and Technology Beijing,Beijing 100083,China;Second Circuit Court of the Supreme People’s Court,Shenyang 110179,China)

机构地区：[1]北京科技大学文法学院,北京100083 [2]最高人民法院第二巡回法庭,沈阳110179

出　　处：《天津大学学报（社会科学版）》2024年第6期512-520,共9页Journal of Tianjin University：Social Sciences

基　　金：2024年中国博士后科学基金第75批面上资助项目(2024M750168)。

摘　　要：在人工智能时代,Generative AI能否取代人类裁判者已经成为了一个热门话题。但该技术直接生成司法决策却在可靠性方面存在困境。根据信息不对称理论,Generative AI的算法原理和数据库具有黑箱属性,可以通过不对称的信息对人类形成不对称的权力,最终将真实数据导出为不可靠的虚假信息。在技术依赖的前提下,人类自我意识的逐渐封闭将进一步放大Generative AI信息不对称的优势,最终形成司法决策不可靠的困境。为了破解上述困境,需要从人机之间信息对称的视角出发,勘定Generative AI介入司法决策活动的边界,合理界定Generative AI和裁判者之间的功能定位和职能范围,严格审查Generative AI的输入和输出内容。In the era of artificial intelligence,the question of whether Generative Artificial Intelligence(hereinafter referred to as Generative AI) can replace human judges has become a popular topic.However,it faces significant challenges in reliability when Generative AI is used to directly generate judicial decisions.Specifically,according to the theory of information asymmetry,the algorithmic principles and databases of Generative AI exhibit a black-box nature,which can create asymmetric power dynamics in relation to humans based on asymmetric information,ultimately transforming real data into unreliable or false information.Under the premise of technological dependency,the increasing closure of human self-awareness will further amplify the advantages of Generative AI's information asymmetry,leading to a dilemma in unreliable judicial decision-making.To address the dilemma mentioned above,it is necessary to approach the issue from the perspective of information symmetry between humans and machines,delineate the boundaries for Generative AI's involvement in judicial decision-making activities,reasonably define the functional orientation and scope of Generative AI and human judges,and rigorously review the input and output content of Generative AI.

关 键 词：Generative AI 虚假信息 可靠性 司法决策 算法黑箱 

分 类 号：DF01[政治法律—法学理论]