3 ms·
> Yes, there’s a “self-hating theory,” F+Not(Con(F)), which believes in its own inconsistency. And yes, by Gödel, this self-hating theory is consistent if F its
by skybrian 6y ago
> Yes, there’s a “self-hating theory,” F+Not(Con(F)), which believes in its own inconsistency. And yes, by Gödel, this self-hating theory is consistent if F itself is—which means that it has a model (involving “nonstandard integers,” formal artifacts that effectively promise a proof of F’s inconsistency without ever actually delivering it). But this self-hating theory can’t be sound: I mean, just look at it! It’s either unsound because F is consistent, or else it’s unsound because F is inconsistent.
I’m not following. What does “unsound” mean here? How can a theory be consistent but unsound?
- pdonis 6y ago> How can a theory be consistent but unsound? "Consistent" means that there is no proposition P such that both P and not-P have proofs within the theory. If F itself is consistent, then F+Not(Con(F)) will also be consistent, since Con(F) will have no proof in this theory. (Conversely, if F itself is inconsistent, then Con(F) will have a proof in this theory (since any proposition will have a proof in an inconsistent theory), so F+Not(Con(F)) will also be inconsistent.) "Sound" means that there exists some semantic model of the theory that the theorems of the theory accurately describe. This cannot true of the theory F+Not(Con(F)), since there are only two possibilities: (1) F itself is consistent--which means that, while F+Not(Con(F)) is also consistent, it can't be sound, because the theorem (by virtue of being an axiom) Not(Con(F)) cannot accurately describe any semantic model of the theory; or (2) F itself is inconsistent--which means that the theory F+Not(Con(F)) has no possible semantic model at all (since F, a subset of the theory, cannot--no inconsistent theory can have any semantic model), and therefore cannot be sound (since to be sound a theory must have some semantic model). Possibility #1 above answers your question.
- dwohnitmok 6y ago1. is not true. There are semantic models of `F + Not(Con(F))` that accurately describe the theory. The difference is that these models contain nonstandard natural numbers (these nonstandard numbers become the Godel numberings of the proofs of `Not(Con(F))`) if `F` is "truly" consistent (according to the standard model of PA). The fact that there must be semantic models of any consistent theory is a cornerstone of first-order logic and is e.g. not true of second-order logic. This follows from Godel's Completeness Theorem (not Incompleteness Theorems).
- pdonis 6y ago> There are semantic models of `F + Not(Con(F))` that accurately describe the theory. That depends on what you allow to count as "accurately describing" the theory. See below. > The difference is that these models contain nonstandard natural numbers (these nonstandard numbers become the Godel numberings of the proofs of `Not(Con(F))`) if `F` is "truly" consistent (according to the standard model of PA). These "proofs" are only "proofs" if you accept "proofs" with an infinite number of steps. In the post I originally responded to in this subthread, there is a quote from the Aaronson article that describes it this way: "it has a model (involving “nonstandard integers,” formal artifacts that effectively promise a proof of F’s inconsistency without ever actually delivering it)" Aaronson doesn't accept these "proofs" as being valid, which is why he takes the position that F + Not(Con(F)) is unsound--or, to put it the way I was putting it, that Not(Con(F)) does not accurately describe the theory since F is consistent but Not(Con(F)) says it isn't. That is also the position I have been taking. The position you are taking is that Not(Con(F)) does accurately describe the theory--and therefore that the theory F + Not(Con(F)) is in fact sound--because we can reinterpret the formula Not(Con(F)) to mean, not "there is some standard natural number that gives a proof that F is inconsistent", but "there is some non-standard natural number that gives a proof that F is inconsistent". Then, as noted, you would also have to reinterpret the notion of "proof" to allow "proofs" that contain an infinite number of steps and never actually end, since that is what a proof whose Godel number is a non-standard natural number would look like. And since such a proof never actually ends, it never actually delivers a formula that says "F is not consistent", so it never actually makes the formal system inconsistent. In the end this is not really a dispute about the math--we agree on what the formal theory is, what its possible semantic models are, and what properties those models have--but on the use of words like "proof" and "accurately describe". So there isn't really a way to resolve such a dispute; we just have to be aware that there are two different viewpoints on how such ordinary language words can be used to describe the math. > The fact that there must be semantic models of any consistent theory is a cornerstone of first-order logic and is e.g. not true of second-order logic. This follows from Godel's Completeness Theorem (not Incompleteness Theorems). Yes, I'm aware of that; see above.
- skybrian 6y ago
- dwohnitmok 6y agoAaronson is using a slightly less common definition of soundness (arithmetical soundness) than the usual one you'll see. Usually we talk about the soundness of entire logical systems (e.g. the soundness of first-order logic) rather than particular theories within that system. Here arithmetical soundness is tightly coupled to the idea of Godel encodings of theorems and proofs. That is a system like ZFC does not have a direct notion of theorems and proofs. Instead a statement such as `Con(ZFC)` is actually a statement about the (non-)existence of a certain natural number fulfilling certain conditions. More specifically `Con(F)` is usually a statement stating that the natural number corresponding to a Godel encoding of a proof of a clearly contradictory statement (e.g. `1 = 0`) does not exist, whereas `Not(Con(F))` is a statement that such a Godel encoding does exist (which usually makes `Not(Con(F))` the more fundamental statement of the two and `Con(F)` is really `Not(Incon(F))` but I digress). Arithmetical soundness requires the notion of "standard natural numbers." In particular, as stated elsewhere in this thread, first-order Peano Arithmetic can be satisfied by many different models, only one of which we designate as our familiar "standard natural numbers." Arithmetical soundness then is the property that natural numbers which satisfy the conditions of these statements encoded via Godel encodings are all standard natural numbers. So for example `Not(Con(F))` corresponds to a statement about the existence of a natural number. If `F` is "truly" consistent then `Not(Con(F))` can only be satisfied by a non-standard natural number. On the other hand if `F` is "truly" inconsistent then `Not(Con(F))` can be satisfied by a standard natural number. You've noticed that I've put the word "truly" in quote marks. This is for two reasons: 1. The usual way to rigorously define "truly" is to appeal to some standard model. Hence the arrow of definition should be going the other way; the concept of a standard natural number that underlies arithmetical soundness is what lets talk about "true" consistency. 2. Whether it makes sense to talk about "the" standard model of a given theory or whether there are multiple possible standard models becomes a matter of philosophy. The former position corresponds to essentially a version of Platonism. The latter position is what is often known as the "multiverse" theory of mathematical logic. Through the philosophical viewpoint of the latter, we can never say absolutely whether a given theory is arithmetically unsound or not, only that it is arithmetically unsound relative to a given model of the natural numbers.