These notes are the author's personal study summaries for the course. They do not contain, refer to, or disclose any assignment or examination materials. All information is for individual learning and revision only.
经典逻辑 Classical Logic
经典逻辑其实就是命题逻辑。惊喜的发现这一部分已经在离散数学里学过了!
argument: premise(s) + conclusion.
an argument is valid just in case if the premises are True, then the conclusion is True.
an argument is sound just in case if it's valid and its premises are True.
Language - surface form (in case of a fire, do not use the lift) - logical form
Interpretation - a function which assigns True or False to each propositions.
条件逻辑引入了 ceteris paribus ("other things being equal") clause:当我在说 明天不下雨 时,我实际说的是 在明天不下雨,且其他事物不变的世界里,我们去打羽毛球吧。
在模态逻辑 Kv 下 (不考虑世界之间 accessibility 的逻辑,其实是 S5),条件逻辑 C 把 if A then B 定义为 A>B:
vw(A>B)=1 iff for all w′ such that wRAw′,vw′(B)=1
RA 即是 ceteris paribus,它表示世界 w′ 中有 A 为真,且除此之外的事物均与 w 相同。
此外,再引入 notations fA(w)={x∈W:wRAx} 与 [A]={w:vw(A)=1}。使用这个表示,我们可以把条件逻辑 C 写为:Vw(A>B)=1 iff fA(w)⊆[B]。
C+ 则在 C 的基础上添加了两个符合直觉的限制。
fA(w)⊆[A]. it's natural to require A to be true at w′ if wRAw′.
If w∈[A], then w∈fA(w). if A is already true in w, then, presumably, the worlds that are essentially the same as w except A is true there, must include w itself.
Similarity Sphere
我们说 fA(w) 是 the worlds essentially the same as w except that A is true there;这个 essentially 的程度值得商榷。即是说,在 ceteris paribus clause RA 中同样也分重要程度。
藉此我们不妨把所有世界按照与 w 的相似程度 (similarity) 划分成若干个相互包含的球面 (sphere) S1,S2,...,并重新定义 fA(w)=Si∩[A] for the smallest i。定义 essentially 的程度为:应当有一个世界的集合,它与 [A] 有重合,且尽量使得该世界集合尽量与 w 相似。
(6) implies that for any world, w, if there are any worlds at which A is true, then there is a unique world closest to w at which A is true.
该约束可以很容易的基于世界的对称性反驳:Bizet-Verdi couterfactuals。Bizet 与 Verdi 是同时代人,但一个是法国人,一个是意大利人。那么,一个 Bizet 与 Verdi 同时是法国人,或者同时是意大利人的世界显然与我们所在的世界有着同样的相似度 (一个他们同时是德国人的世界则离我们的世界更远)。
C2 条件逻辑中,Conditional Excluded Middle 条件排中律 (A>B)∨(A>¬B) 成立 (在 S 中不成立)。
条件排中律是有问题的。令 A 为 it will rain tomorrow or it won't, B 为 it will rain tomorrow。显然 A>B 与 A>¬B 均为假,但条件排中律却判断 (A>B)∨(A>¬B) 为真。
David Lewis 发现了这个问题,他定义了 C1,将 (6) 替换成了 (7):
If w∈[A] and w′∈fA(w), then w=w′.
(7) implies that if A is true at w, then the most similar worlds at which A is true comprise just w itself (Any world is more similar to itself than any other world).
Paradoxes of Self-reference (罗素悖论中的集合既属于它自身又不属于它自身;也有人使用 neither 来解释,但这样 Contradiction Law p∧¬p 将会失效).
此外,和 L3 之于 K3 一样,LP 也有其修改版 RM3 (解决了 positive, negative paradox,获得了 modus ponens,仍不具有 modus tollens)。
一阶蕴涵 First Degree Entailment
关于 First-Degree Entailment 最重要的一点是: Being false (relate to 0) does not mean being untrue (not relate to 1). Also, being untrue does not mean being false。
逻辑实证主义运动作为具有广泛影响力的哲学运动,存在严重的缺陷 (the verifiability criterion of meaning was itself unverified),饱受各路哲学家批评。A.J. Ayer 甚至打趣说「逻辑实证主义最重要的缺陷在于它几乎全部是错误的」。该运动式微于 20 世纪 70 年代;1967 年 John Passmore 宣称其已经「死了,就像它曾经是一场哲学运动一样死了」。
逻辑实证主义将所有 statements 分为三类 (因此它也能被三值逻辑表达):
Tautologies (Analytic Statements). Their truth is self-contained and independent of the external world. Often reveal something about the logical or linguistic structure. e.g., "All bachelors are unmarried", or "2+2=4". These statements are cognitively meaningful.
Verifiable by Experience (Empirically Verifiable). Can be confirmed or falsified through observation or experiment. e.g., "It's raining outside", or "Water boils at 100C under standard atmospheric pressure". These statements are scientifically meaningful.
Nonsense (Metaphysical). Neither be verified by experience nor are they tautologically true, thus lack cognitive significance. e.g., "God exists", or "The universe has a purpose".
经典逻辑实证主义将神学,伦理学与美学均视为形而上学的范畴,因此是无实在含义的。这些领域内的 statements 被逻辑实证主义者视为 emotionally significant, but not literally significant.
逻辑树 Tableau
证明 A⊨B 的利器。
递归的对 A 与 ¬B 进行原子化,遇 ∧ 继承,遇 ∨ 分叉。逻辑树的每一个枝杈是 argument 的一个 interpretation。若枝杈发展到最后,存在 p 与 ¬p 的 contradiction,我们标记该枝杈为 closed。若逻辑树的每一个枝杈都 closed,说明每种 interpretation 都能保证在 A 为真的前提下 B 一定非假;这符合 validity 的定义。
证明 A 是 tautology ⊨A:对 ¬A 进行展开。证明 A 是 contradiction ⊨A:无需 negate。
模态逻辑树则稍微复杂一点。我们需要维护一个由 possible world set,并在每一个世界都检查是否有矛盾出现。