Tweety is a bird.
Birds fly.
Therefore, Tweety flies.

What if we also know that:

Penguins are birds.
Penguins do not fly.
Tweety is a penguin.
Therefore, Tweety does not fly!

Default reasoning provides a solution. A default rule specifies a plausible inference:

Typically birds fly.

I.e., In the absence of evidence to the contrary, all birds fly. Since it can now be shown that Tweety is a penguin, and penguins do not fly, and hence Tweety does not fly, the default rule will not be applicable.

In the absence of the fact that Tweety is a penguin, the default inference would still be possible. Later on, if it is asserted that Tweety is a penguin, the earlier inference would have to be retracted. Such logics are called nonmonotonic. One has to extend FOPC to include such default rules.

所以應該是一個 “預設規則” (default reasoning)的問題
後來還有人發明circumscription…也就是預設一些東西不成立
(好像又太簡化…不過大致的意思是這樣子)
第一個提出circumscription的人似乎是McCarthy的樣子…
東西好多有空有興趣在繼續研究吧

那個所謂的反例，當中「鳥會飛」這個前提本身是錯誤的。
前提假設是錯的話，推論方式是正確的也沒用。但不能說三段式論證有反例。

實際上那個「鳥會飛」是屬於「歸納法」的問題。
「歸納法」取之於經驗，一但有超出經驗的例外，所歸納出的法則就破滅。
比如大多數人所見的天鵝都是白色的，所以依經驗歸納出「凡是天鵝都是白色的」
然而最後發現澳大利亞有一種黑天鵝，這個例外使「凡是天鵝都是白色的」破滅。

所以文中「鳥會飛」是屬於「歸納法」的問題，不能指涉邏輯推論錯誤。