论贝叶斯条件化和杰弗里条件化之关系--评休伯对贝叶斯主义的批评  

On the Relation Between Rules of Bayesian Conditionalisation and Jeffrey Conditionalisation--Comments on Huber’s Criticism of Bayesianism

作　　者：袁继红 陈晓平[3] Yuan Jihong;Chen Xiaoping

机构地区：[1]广东财经大学马克思主义学院,广东广州510320 [2]广东财经大学智能社会与人的发展研究中心,广东广州510320 [3]华南师范大学政治与公共管理学院,广东广州510006

出　　处：《学术研究》2022年第11期35-42,197,共9页Academic Research

基　　金：国家社科基金一般项目“计算社会科学的解释研究”(19BZX033);国家社科基金一般项目“语境主义反怀疑论方案批判研究”(18BZX040);广东省社科规划项目“心身问题的功能主义进路研究”(GD20YZX02)的阶段性成果。

摘　　要：在概率演算公理系统之内把概率解释为主观置信度的意义上,杰弗里条件化常被纳入贝叶斯理论的范围,但是它与贝叶斯条件化有着本质的区别,即贝叶斯条件化只考虑确凿证据对假说的认证度的影响,而不考虑不确定证据对认证度的影响。贝叶斯条件化是杰弗里条件化在Pr´(e)=1时的特例,似乎不如杰弗里条件化来得普遍。但是,贝叶斯条件化的这种“狭隘”对于科学检验来说不但不是缺点,反而是科学之严格性和可靠性的体现。然而,休伯却忽视了贝叶斯条件化和杰弗里条件化之间的这一区别,把从杰弗里条件化推出的违反科学直觉的结论强加到整个贝叶斯理论之上,这种做法是有失公允的。In the sense of interpreting probability as subjective degree of belief within the axiom system of probability calculus, Jeffrey conditionalisation is often included in the scope of Bayesian theory. However, Jeffrey conditionalisation is substantially distinct from Bayesian conditionalisation, that is, Bayesian conditionalisation considers only the effect of reliable evidence on the confirmation of a hypothesis, and not the effect of uncertain evidence on the confirmation. Bayesian conditionalisation is a special case of Jeffrey conditionalisation at Pr’(e) =1, and seems to be not as general as Jeffrey conditionalisation. However, this “narrow” of Bayesian conditionalistion is not a disadvantage for scientific test, but a embodiment of the rigor and reliability of science. Huber, nevertheless, ignores this distinction between Bayesian and Jeffrey Conditionalisation, and imposes counterintuitive conclusions derived from Jeffrey Conditionalision over Bayesian theory. This is unfair.

关 键 词：贝叶斯条件化 杰弗里条件化 认证 老证据问题 

分 类 号：N031[自然科学总论—科学技术哲学]