法律思维中的“软计算”和贝叶斯语言方法  

‘Soft Computing'and the Bayesian Linguistic Approach in Legal Thinking

作　　者：刘东亮 Liu Dongliang

机构地区：[1]西安交通大学法学院

出　　处：《中国法学》2024年第6期108-126,共19页China Legal Science

基　　金：2023年度教育部社科基金规划项目“法律模糊性问题的解码与计算”(项目批准号:23YJAZH078)的阶段性成果。

摘　　要：“法律人如何思考”是学界一向关注但一直没有阐释清楚的问题。从认知科学角度而言,人类的思维是一种心智计算。由于人脑的语言并非数学语言而是统计性质的语言,并且主要是自然语言,因此心智计算是不追求精确解、富有灵活性的“软计算”,尤其是在信息不完备的条件下,软计算体现为贝叶斯方法和语言方法的综合运用。贝叶斯方法的本质是“用新证据修正概率”,而概率可通过语言概率(模糊语言值)来表示,心智计算模型因而可以概括为:软计算=贝叶斯法则+语言概率。简言之,人类的思维遵循贝叶斯法则,在认识事物时基于自身经验通常会有一个主观的先验概率之判断(初始置信度),然后根据新证据提供的信息不断修正调整先验概率而得到客观化的后验概率(经改进的更全面的置信度),并在后验概率判断的基础上,作出最终决策。这种心智计算模型,与法律人的思维过程特别是法官的判案过程高度契合。‘How do lawyers think'is a question that the academic circle has always been paying attention to but has not been explained clearly.From the perspective of cognitive science,human thinking is a kind of mind computation.Since the language of the human brain is not mathematical language but statistical language and mainly natural language,mind computation is soft computing that does not pursue accurate solutions and has great flexibility.Especially under the condition of incomplete information,soft computing is the comprehensive application of the Bayesian method and linguistic approach.The essence of the Bayesian method is updating the probability with new evidence,and the probability can be expressed by linguistic probability(fuzzy language value),so the mind computation model can be summarized as:soft computing=Bayes Rule+linguistic probability.In short,human thinking follows Bayes'Rule.We revise constantly the subjective prior understanding of things according to new evidence information,conclude the objective posterior understanding,and then make the final decision.This mind computation model is highly consistent with the thinking process of lawyers,especially the judgment process of judges.

关 键 词：软计算 词语计算 贝叶斯法则 语言概率 法律思维 

分 类 号：TP18[自动化与计算机技术—控制理论与控制工程] D90[自动化与计算机技术—控制科学与工程]