# Is mathematics done in an arbitrary model of ZFC?

Following up a previous thread I posted, I have tried to refine my questions. I would be happy with answers simply confirming that I have understood matters correctly, but of course I would also be happy to read longer and more elaborate answers!

Suppose I want ZFC (as formulated in first-order logic) to be my foundations of mathematics, i.e. I want it to be possible to formalise all of my reasoning about for example number theory as consequences in first-order logic of the ZFC axioms.

These are my questions:

1. When I consider a "set", for example $\mathbb{N}$, when doing mathematics, is it reasonable to view $\mathbb{N}$ as living in a fixed model of ZFC? So if I wanted to be very pedantic, I would refer to $\mathbb{N}$ as $\mathbb{N}^\mathcal{A}$ where $\mathcal{A}$ is my fixed model of ZFC?

2. Suppose I am doing mathematics and discussing properties of $\mathbb{N}$ with a friend of mine. Can any "problems" arise from me (implicitly) thinking of $\mathbb{N}$ as being $\mathbb{N}^\mathcal{A}$ living in my ZFC-model $\mathcal{A}$, while my friend thinks of $\mathbb{N}$ as being $\mathbb{N}^\mathcal{B}$ where $\mathcal{B}$ is his ZFC-model? I suppose since all assumptions we have made about $\mathcal{A}$ and $\mathcal{B}$ is that they are ZFC-models, no "substantial" problems can arise, so my friend and I can therefore safely agree to ease notation by referring to $\mathbb{N}^\mathcal{A}$ and $\mathbb{N}^\mathcal{B}$ as simply $\mathbb{N}$?

3. Given that my understanding of the above questions is correct, it seems reasonable to view mathematics in a "semantic" way as "being done in an arbitrary model of ZFC"? Alternatively, if I insist on first-order logic ZFC to be my foundations, I could view doing mathematics as a sort of informal natural deduction, i.e. I could have a "syntactic" view on mathematics. The soundness and completeness theorems ensure that the "semantic" and "syntactic" views in some sense are the same, but perhaps one of the approaches has some philosophical advantages? I obviously find the semantic view more pleasing, since the semantic view is "consistent" with my view of a sets as living in models of ZFC...

Edit:

I realise that perhaps "the standard model of ZFC" might be relevant to my questions. That is, I should perhaps view $\mathbb{N}$ as always referring specifically to $\mathbb{N}^\mathcal{S}$ where $\mathcal{S}$ denotes the standard model of ZFC? Similarly, I should perhaps also view mathematics as being done specifically in the standard model of ZFC. If we disregard problems with even defining which model of ZFC is the standard one, this view also seems to give us problems with translating our informal mathematical proofs to natural deduction: If mathematics is done in a very specific model of ZFC, we can no longer rely on completeness of first-order logic to guarantee existence of a derivation in natural deduction.

To summarise my edit: If the standard model of ZFC is relevant to my questions, please illuminate how. I do not think the standard model is relevant for my questions in any other way than being an "intuitive model" when thinking of ZFC.

• I believe soundness means that if your proofs boil down to formal proofs with ZFC axioms alone, the only conclusions you can draw are those true in all models, so it shouldn't matter what model you're thinking of. But if you're thinking only of a model where CH holds, and you want to make use of CH, then you're not "working in ZFC" anymore, but ZFC+CH instead. – Mark S. Jun 2 '17 at 22:41
• You should read Benecerraf's delightful paper "What numbers could not be" (which will show you that even the definition of $\Bbb{N}$ within a given formalization of set theory is a nuanced question). – Rob Arthan Jun 2 '17 at 22:59
• Just wondering ... what if your friend said that he looked at the natural numbers as urelements - would you refuse to discuss math with him? – CopyPasteIt Jun 2 '17 at 23:06
• It is a good thing you have at least a ZFC mindset. Otherwise some set theory purists would be pouncing on your question with down votes and going for eventual deletion. – CopyPasteIt Jun 2 '17 at 23:10
• @MikeMathMan If we were actually solving some math problem, I would happily treat the natural numbers as urelements. However, I would also like to feel safe about my foundational questions, to know "what's going on in the background" :) – SimonSimon Jun 2 '17 at 23:25

Given this one model we can construct other models with different properties. For example we can construct a model $\mathcal A$ where the Continuum Hypothesis holds and another $\mathcal B$ where it doesn't hold. So if are doing maths over $\mathcal A$ and your friend is doing maths over $\mathcal B$ it is true for you that $\aleph_1= 2^{\aleph_0}$ but not for your friend.
The solution is, rather than imagine you are working over an arbitrarily chosen model, to imagine you are working over ALL models of ZFC at once. When you say $1+1=2$ you do not mean $1^\mathcal A+1^\mathcal A=2^\mathcal A$ for your favourite model $\mathcal A$. You mean $1^\mathcal A+1^\mathcal A=2^\mathcal A$ for each and every model $\mathcal A$.
An alternate approach might be to try and define a standard model as one which obeys all the axioms of ZFC and does not have any additional sentences that could be added as axioms, and do your maths there. This avoids the continuum problem above. However it is impossible because we can take any non-axiom sentence $P$. Either $P$ or $\neg P$ will hold in the model. Therefore it is also a model of either $ZFC + P$ or $ZFC + \neg P$.