‘Killing robbers is not killing people’: the Later Mohists on inferences that fail
One passage in the Later Mohist texts seems to me to belong in any discussion of what predicate logic was invented to handle.
In the chapter of the Mozi usually called the "Lesser Selection" (Xiaoqu), the Mohists discuss a kind of inference by parallel construction: add the same words to both sides of a true statement and see whether the result is still true. Sometimes it works. A white horse is a horse, and riding a white horse is riding a horse. But sometimes it fails. Robbers are people. Yet, the text says, having many robbers is not having many people, and killing robbers is not killing people.
That last line was notorious. Xunzi, in his chapter on the correct use of names, cites it as an example of confusing names, and he was not being friendly. The Mohists had a stake in it: they condemned killing people, and they also accepted that robbers could be executed.
What fascinates me is that they noticed the pattern at all. They saw that a form of inference which seems obviously safe breaks down for certain verbs, and they tried to classify when it works and when it does not. That is exactly the territory of relational and intensional inference that syllogistic could not reach.
Is "killing robbers is not killing people" a confusion, a moral evasion, or an early insight about how descriptions of actions work?