推进旧证据问题的逻辑学习方案研究：一种新视角

Advancing Research on a Logical Learning Approach to the Old Evidence Problem: A New Perspective

摘要: 由格里默最先提出的旧证据问题是对贝叶斯确证理论所提出的最关键的挑战之一。诸多贝叶斯主义者针对该问题提出了不同的解决方案。其中的一种标准回应是允许逻辑学习，即逻辑学习方案。但逻辑学习方案也面临着许多问题，从而造就了逻辑学习方案的不同版本，它们都各有特点与优势。其中一个较新的版本是由伊娃和哈特曼于2020年提出的。该版本继承哈特曼和菲特森的思路，主张放宽原子命题、引入理论间的竞争关系，并创新地使用理论对证据的解释效力优势刻画确证关系，这都是以往版本所不曾考虑到的。这些观点都颇为符合实际，但仍旧有其缺陷，即忽略确证活动中背景证据所发挥的作用以及理论、证据、背景证据之间的动态关系，而这是与理论解释效力优势密切相关的。然而，当我们引入背景证据，就会发现推进逻辑学习方案的另一种视角，这也将成为看待旧证据问题本身的新视角。并且，在这种视角中，逻辑学习方案的一些版本以及另一些解决旧证据问题的方案也可免于部分批评。

Abstract: The old evidence problem, first proposed by Glymour, is one of the most critical challenges to Bayesian confirmation theory. Many Bayesians put forward different solutions to this problem. One of standard responses is to allow logical learning, naming logical learning approach. However, logic learning approach also faces many problems, thus creating different versions of logic learning approach, which have their own characteristics and advantages. One of the newer versions was proposed by Eva and Hartmann in 2020. This version inherits the ideas of Hartmann and Fitelson, advocating relaxing the atomic propositions, introducing competitive relationship between theories, and innovatively using the superiority of theory’s explanatory effect on evidence to depict the confirmatory relationship, which has not been considered in previous versions. These viewpoints are quite in line with reality, but they still have their defects; that is, they ignore the role played by background evidence in confirmation activity and the dynamic relationship between theory, evidence and background evidence, which is closely related to the superiority of theory’s explanatory effect. However, when we introduce background evidence, we will find another perspective to advance the logic learning approach, which will also become a new perspective to look at the old evidence problem itself. In addition, in this perspective, some versions of logical learning approach and other solutions to old evidence problems can also be exempted from partial criticism.