To me (but I might be wrong) Rice's theorem asserts that it's not possible to formalise the demonstration of a non-trivial property of a recursively enumerable language within the same given language. The modus ponens being the basis for demonstrating, Rice therefore says in substance that you will necessarily need a metalanguage to define it.

What I call Carroll's paradox is referred in Wikipedia under What the Tortoise Said to Achilles.

Wikipedia explains the following:

The Wittgensteinian philosopher Peter Winch discussed the paradox in The Idea of a Social Science and its Relation to Philosophy (1958), where he argued that the paradox showed that "the actual process of drawing an inference, which is after all at the heart of logic, is something which cannot be represented as a logical formula ... Learning to infer is not just a matter of being taught about explicit logical relations between propositions; it is learning to do something" (p. 57). Winch goes on to suggest that the moral of the dialogue is a particular case of a general lesson, to the effect that the proper application of rules governing a form of human activity cannot itself be summed up with a set of further rules, and so that "a form of human activity can never be summed up in a set of explicit precepts" (p. 53).

In other words, Carroll's paradox shows that it's not formally possible to prove a property of a language purely with propositions of this language. You will need to admit a rule (or a principle) that "the modus ponens means X" with X using everyday words. I.e. you will need a meta language.

I will concede that it is a philosophical way to ask the question. But Carroll's paradox can be described in a formal (mathematical) way (done here).

Hence am I mistaken in principle to say that Caroll's paradox implies Rice's theorem?

• I think that 'similar' is vague enough to be said of many things to the point that it is meaningless. I'm afraid that this question isn't very suitable for this Q and A format unless you have a more concrete way of comparing the two statements. Apr 9 '18 at 10:17
• Perhaps this question is more appropriate for Philosophy. Apr 9 '18 at 13:10
• Ok I've edited it Apr 9 '18 at 14:16
• I think trying to express precise mathematical statements in imprecise vague English language, and then reason about their similarities by analogy of the language, is just asking for trouble. Rice's theorem is a statement of mathematics, with a precise meaning, and I don't think its meaning matches the description you gave. Nor do I think that your description of Carroll's paradox fully captures what's going on with it. And what's the difference between "mistaken in principle" and "mistaken"?
– D.W.
Apr 9 '18 at 17:10

Well, any true statement implies every other true statement, so in that vacuous sense, I suppose one implies the other.

But no, I wouldn't say that Carroll's paradox implies Rice's theorem in any interesting or meaningful way. Carroll's Paradox is an explanation of why we need modus ponens as an inference rule in propositional logic and can't replace it with a finite list of axioms. It's nothing fancier than that. As such, it refers to what is true in a logical theory. It refers to what inference rules or axioms we need.

Rice's theorem is different. It talks about what is decidable, not what is true. It talks about what can be computed on a Turing machine, not about the inference rules or axioms we need.

Rice's theorem doesn't say "it's not possible to formalise the demonstration of a non-trivial property of a recursively enumerable language within the same given language". It has nothing to do with "within the same given language" (it's not even clear what that would mean; a language is a subset of $\{0,1\}^*$ and doesn't demonstrate anything), but rather with that can be computed with a Turing machine. Rice's theorem talks about what a Turing machine can decide.

Rice's theorem doesn't say "you will necessarily need a metalanguage to define modus ponens". (I don't even know what a metalanguage would mean in this context.) It has nothing to do with what you can define. Rather, it has to do with what a Turing machine can decide.

In general, I warn you against trying to reason about mathematical statements using English language and analogies. That is notoriously error-prone. Rice's theorem is a mathematical statement with a precise mathematical meaning. At a suitable level of imprecision and vagueness, I suspect everything can seem similar and connected -- but when you look at the specifics, the situation often becomes richer and more interesting than that.

• Right let me write it this way then. The hare says there is a finite method to prove B is TRUE when we know A is TRUE and A implies B. In other words the hare is trying to write a finite axiomatic theory. The tortoise shows him that if such a theory is possible, then the hare's algorithm (written in his finite blotter) never produces TRUE or FALSE. We know that any finitely axiomatic theory is recursively enumerable. So what the turtle is really saying is that there is no decision algorithm which terminates on the problem of knowing if B is TRUE or FALSE. Which is the halting problem Apr 11 '18 at 14:21
• @Jerome, I realize it all sounds plausible in English, but again, a lot of false things about math can sound plausible if stated imprecisely enough in English. You need to learn the math -- you can't reason about this by talking in English sentences. Anyway, your reasoning is faulty. When you say "what the turtle is really saying", no, that's not what the turtle is really saying, and it's not at all equivalent. The turtle is saying "if you don't have the modus ponens inference rule, you can't prove all true things". That's not the same as Rice's theorem or the halting problem.
– D.W.
Apr 11 '18 at 16:40
• We know that even if you do have modus ponens as an inference rule, then the language of true statements is still not decidable. So that's a stronger, more powerful statement than anything found in Carroll's paradox, because Carroll's paradox is limited to exploring the consequences if we don't have modus ponens as an inference rule.
– D.W.
Apr 11 '18 at 16:42