旧证据问题及Garber型学习策略探析

The Old Evidence Problem and the Analysis for Garber-type Learning Strategies

由格莱莫尔(Clark Glymour)于1980年提出的旧证据问题,是对贝叶斯确证理论合理性的质疑。自旧证据问题提出以来,多种解决方案被相继提出,最具代表性的是豪森(Colin Howson)的反事实策略和盖贝尔(Daniel Garber)的学习策略。文章通过分析旧证据问题产生的可能原因,从逻辑全能认知的角度和学习策略入手,来梳理Garber型学习策略这类消解方案,探究这类消解方案在哪一层面、何种程度上达到消解旧证据问题的目的。学界对这类方案存在的异议或赞许纷争始终未能达成一致,从而并未实现从根本上消除旧证据问题同时保留贝叶斯理论中的合理成分的目标。对Garber型学习策略的探析或许能为更彻底地解决旧证据问题提供一定的启示作用。

The Old Evidence Problem presented by Glymour in 1980 has been a great challenge for the rationality of the Bayesian Confirmation Theory. Since the Old Evidence Problem has been raised, much attention has been paid to it. A variety of solutions to the problem have been proposed, the most representative of which are Howson＇s Counterfactual Strategy and Garber＇s Learning Strategy. This paper analyzes some possible causes of the Old Evidence Problem. From the perspective of logical omniscience cognition and the Learning Strategy, it then summarizes the development of Garber-type Strategies, and explores how this kind of solutions resolve the Old Evidence Problem. However, these proposals have their own advantages and disadvantages. It is hard to reach an agreement on these disputes among scholars. Thus we haven＇t achieved the goal of both eliminating the Old Evidence Problem and preserving the reasonable components of the Bayesian theory. So the analysis of Garber-type Learning Strategies will provide us with some inspirations for eliminating the Old Evidence Problem thoroughly.