- 學位論文
指示條件句的兩個難題: 亞當斯論題與合理的推論
Two Problems of Indicative Conditionals: Adams' Thesis and Reasonable Inference
摘要
本論文主要針對指示條件句的兩個難題: 第一個的難題是, 亞當斯論題(Adams' thesis)該如何解讀? 第二個難題是, 一個好的條件句推論該具有什麼性質? 亞當斯論題是由亞當斯對指示條件句所提出的一個主張, 此論題在條件句的文獻裡得到很多人的支持。而它通常被理解為, 在前件的主觀機率不等於0的條件下, 指示條件句的主觀機率等於給定其前件下其後件的條件機率。形式化地說: P(A → B) = P(B|A), 在 P(A) ≠ 0 的條件下。 許多學者認為這個論題在直覺上是成立的, 但對其明確意義及為何成立, 則有不同的立場。本文將論證亞當斯論題不僅僅只是一個關於指示條件句的假說或假定, 而且在基於一個恰當的指示條件句三值語意論下, 它可以被恰當地說明及導出。 當一個推論中有指示條件句時, 我們要用什麼判準來評價它是一個好的推論或是不好的推論? 亞當斯和史東內克都試圖為這個難題提供解答, 亞當斯對此提出機率的健全性準則; 史東內克則是提出 合理的推論說法。亞當斯認為一個好的推論是, 其結論的不確定性不能超過其前提的不確定性之總和。然而, 史東內克認為一個好的推論是, 在其前提可以被適當地斷說或預定的每個脈絡中, 一個人在接受其前提時, 不可能不去接受其結論。本文將論證史東內克的合理推論說法比亞當斯的說法更可信。 筆者主張指示條件句是三值語句, 並對它提出一個一般性的機率理論。在這樣的機率理論下, 當我們只針對簡單條件句時, 亞當斯論題就可以被恰當地說明和導出。換句話說, 亞當斯論題是筆者理論下一個特殊的情況。這其中最重要的想法是, 我們得去區分機率與可斷說性, 才能對亞當斯論題所代表的意義有更精確的掌握。筆者試圖從三值語意論給可斷說性一個清楚的定義, 在有了清楚的可斷說性概念後, 我們就能再去區分可斷說性保留性與可斷說性健全性以解決第二個難題。在這樣的區分下我們會發現: 亞當斯所支持的是可斷說性健全性; 史東內克所支持的是可斷說性保留性。最後筆者結論: 亞當斯論題應被解讀成, 簡單條件句的可斷說性等於其相對應的條件機率; 一個好的條件句推論是一個合理的推論, 也就是, 一個具有可斷說性保留性的推論。
並列摘要
This dissertation deals with two problems of indicative conditionals. First, what is an appropriate interpretation of Adams' thesis"? Second, what is the property that a "good" inference involving indicative conditionals has? "Adams' thesis" is that the it probability of indicative conditionals is the conditional probability of the consequents given the antecedents ─ schematically: P(A → B) = P(B|A) (provided P(A) ≠ 0). Many scholars believe that Adams' thesis is intuitively correct, but they disagree on its exact meaning and why it is correct. This paper argues that Adams' thesis is not only a hypothesis, but also one that can be properly explained and derived by an appropriate semantics of indicative conditionals. Adams proposes that a good inference should be a it probabilistic sound inference such that the uncertainties of its conclusion cannot exceed the sum of the uncertainties of its premises. However, Stalnaker proposes that a good inference is a reasonable inference such that in every context in which the premisses could appropriately be asserted or supposed, it is impossible for anyone to accept the premisses without committing himself to the conclusion. Adams and Stalnaker both try to capture the property that good inferences involving indicative conditionals have. This paper argues that Stalnaker's account is more plausible than Adams'. I provides a generalized probability theory of 3-valued indicative conditionals, and given the generalized probability theory it shall be shown that Adams' thesis can be properly explained and derived as a special case when the indicative conditionals under consideration are simple indicative conditionals. The idea is that we need to make a distinction between probability and assertability in order to have a better understanding of Adams' thesis. Then we can make a distinction between assertibility preservation and assertabilitic soundness. In this sense, Adams' probabilistic sound inferences are assertabilitic sound inferences, and Stalnaker's reasonable inferences are assertability preservation inferences. So I conclude that Adams's thesis should be interpreted as the assertabilities of a simple indicative conditional equals the corresponding conditional probabiliity, and that a good inference is a reasonable inference which has the property of assertability preservation.