司法人工智能推理辅助的“准三段论”实现路径  

The Ways of Achieving“Quasi-syllogism”in Judicial Artificial Intelligence Ratiocination Assistance

作　　者：王竹[1] Wang Zhu

机构地区：[1]四川大学法学院

出　　处：《政法论坛》2022年第5期28-39,共12页Tribune of Political Science and Law

基　　金：研究阐释党的十九届四中全会精神国家社科基金重点项目“深化司法体制综合配套改革研究”(20AZD029)的阶段性成果。

摘　　要：我国司法人工智能研发面临法律思维、可解释性和可计算性三大瓶颈,应以可解释的司法人工智能推理辅助作为主要突破方向,法律要件解耦是关键。在司法人工智能推理辅助中无法直接嵌入传统三段论逻辑,应当以“争议焦点”替换“结论”,构建“准三段论”逻辑结构,应对大前提的横向杂糅性、争议焦点的类型化和小前提的复杂性与多样性等主要挑战,并实现动态维护。基于“大前提≥争议焦点≥小前提”层次性态势,设置不超过五级的标签体系,可以通过三大“组群”关系合并表达“准三段论”束46种具体链接方式,实现一定程度和范围的智能推理辅助。The research and development of judicial artificial intelligence in China are bottlenecked in three aspects,namely,legal thinking,interpretability and computability.It is assumed that with legal requirements decoupled,breakthroughs can be made in the field of explainable judicial artificial intelligence ratiocination.The classical syllogism cannot be embedded directly in judicial artificial intelligence ratiocination assistance,but when“conclusion”is replaced by“issue”,a logical structure of“quasisyllogism”can be built,which is able to answer the major challenges caused by hybridity of major premise,categorization of issues and complexity and diversity of minor premise,while maintains its dynamic fl exibility.Observing the hierarchical structure of“major premise≥issue≥minor premise”,if we establish a labeling system with no more than 5 levels,then 46 specific ways of linkage of“quasi-syllogism”bundle can be expressed through three major“grouping”relationships,so as to achieve ratiocination assistance by artificial intelligence in certain degree and scope.

关 键 词：司法人工智能 推理辅助 准三段论 民商事判决书 法律要件解耦 

分 类 号：D926[政治法律—法学] TP18[自动化与计算机技术—控制理论与控制工程]