Submind YouTube summaries
Thumbnail for Proofs & Semantics

Proofs & Semantics

Watch on YouTube

Video summary

In deductive logic, proofs and models represent two distinct yet interconnected methods for analyzing sentences within Quantified Logic (QL), distinguished by their syntactic and semantic nature respectively. Syntactic analysis relies on proof theory, indicated by the single turnstile symbol ($\vdash$), which asserts that a statement is derivable or provable based solely on logical rules without reference to truth values; this establishes whether something is a theorem. Conversely, semantic analysis uses model theory, denoted by the double turnstile symbol ($\models$), to determine if an argument holds true across all possible interpretations, thereby establishing validity or tautology. While these approaches differ in their fundamental mechanisms—one manipulating symbols according to rules and the other evaluating truth conditions—they are deeply related through two critical properties of formal systems: soundness and completeness. A system is considered sound when every syntactic proof corresponds to a semantically valid argument, ensuring that no invalid conclusions can be derived; for instance, the rule of conjunction introduction preserves truth because if premises $A$ and $B$ are true in any model where they hold, their combination must also be true. The relationship between these methods is further defined by completeness, which guarantees that every semantically valid argument has a corresponding syntactic proof within the system; both QL systems discussed possess this property, allowing logicians to use proofs or models interchangeably depending on convenience. However, not all logical frameworks share this balance, as demonstrated by hypothetical rules like "modus pocus," which would allow deriving $C$ from an implication where only one case is true but fails semantic entailment in others, rendering such a system unsound. This distinction highlights that while QL is both sound and complete for basic logic, stronger systems capable of expressing arithmetic are inherently incomplete due to Gödel's incompleteness theorems, meaning there will always be truths about natural numbers that cannot be proven within those specific proof systems. Consequently, understanding when to apply a model versus a proof becomes essential: models are most efficient for demonstrating invalidity by providing a single counterexample or showing consistency with one true instance, whereas proofs are necessary for establishing universal validity, tautology, or contradiction where reasoning must hold across all possibilities. Practical application of these concepts allows logicians to efficiently address various logical questions such as equivalence, contingency, and inconsistency by choosing the most appropriate tool for the task at hand. For example, proving that a statement is contingent requires showing it can be true in one model while false in another, whereas demonstrating invalidity only demands constructing a single countermodel where premises are true but the conclusion is false. Similarly, establishing logical equivalence involves deriving each sentence from the other via proof or exhibiting models with differing truth values for non-equivalent pairs. The transcript illustrates these strategies through specific exercises involving quantifiers and predicates; one example shows how to prove an autology by using universal introduction after handling free variables in a subproof, while another demonstrates that a seemingly consistent statement is actually contingent only if two different models are constructed. In cases of inconsistency or invalidity, the process often shifts toward model construction to find specific assignments for constants and extensions of predicates that satisfy premises but falsify conclusions, effectively bypassing lengthy derivations when semantic counterexamples suffice. Ultimately, mastering the interplay between proof-theoretic and semantic concepts enables a flexible approach to solving complex logical problems without reinventing methods from scratch. By leveraging soundness and completeness, students can confidently alternate between deriving proofs in software like Carapace or constructing models with limited domains to verify truth conditions, ensuring that their conclusions are robust regardless of the method used. This dual capability is particularly valuable for exam scenarios where one must determine if a set of sentences is consistent by finding a model where all are true, or prove an argument invalid by identifying a scenario where premises hold but the conclusion fails. The ability to switch seamlessly between these perspectives not only deepens understanding of logical structure but also provides practical strategies for tackling questions about validity, contingency, and equivalence efficiently. As emphasized in the lecture, keeping both tools in mind allows for optimal problem-solving techniques that will serve well in advanced logic courses and examinations, reinforcing the idea that while proofs offer rigorous derivation paths, models provide intuitive checks on truth across interpretations.
Read the full video transcript
hello and welcome back to Phil 320 deductive logic I'm professor Matthew Brown this is the third in our lectures on proofs in ql and the final lecture of the semester um and today we're going to be talking about proofs and models in ql how to use our proof theoretic and our semantic Concepts to accomplish various tasks I want to start by just talking about the relationship between proofs and models the these are two very different ways of evaluating sentences of ql we use these different symbols the single Turn Style and the double Turn Style to indicate when we're doing semantic and syntactic Analysis the single Turn Style tells us that a proof is possible right um we denote this with a single Turn Style and we know this is a syntactic style of analysis right it's not the same as semantic entailment which we wrote with this double Turn Style on the right here we talked a bit in the last unit about the difference between proof theoretic or syntactic and semantic analyses so um you remember on the Left Right single Turn Style a says that a is a theorem a is a theorem of SL or ql depending on what we're talking about um the right says with the double Turn Style that a is a tautology right similarly the left here says that we can derive B from a right B can be proved on the basis of a the right on the other hand says that a semantically entails B if a is true then B must be true we don't have to mention the truth values of A and B to understand derivability or provability but we do have have to mention it if we're going to account for semantic entailment right so now I've I'm I'm saying all of these are different Notions and they are because one is syntactic and the other is semantic um but how are they connected how do they relate to one another well the property of some formal systems that we call soundness means that whatever is derivable is also semantically entailed right a pro system is sound if there are no proofs of invalid arguments right if every proof implies that there is semantic entailment there is validity so consider the conjunction introduction rule right suppose up to this point you have a proof um where you've derived A and B and suppose also that you uh have determined semantically speaking that the proof is valid right so A and B are either premises of the argument or they're valid consequences of the premises in any model where the premises are true A and B are true right so we've done the semantic analysis of that right given the definition of Truth in ql right um specifically part three of that definition A and B must also be true right um so the conjunction introduction rule when applied to um sentences we know to be true preserves that truth right so any application of the conjunction introduction rule not only creates a proof uh where we derive A and B but we know that that must be also valid a valid argument because it's semantically entailed right consider a new rule that we might add to our our proof system let's call it modus Pocus right um here's how I'll Define that rule suppose on line M uh we have a conditional of the form if B then C right modus Pocus as a rule allows us to conclude C right and you might think that doesn't sound like a great rule how did I get C well yes that's part of the point now let's look at our uh definition of Truth in ql right um if a sentence a has the form if B then C for some wolfs B and C then we know that a is not satisfied if um B is satisfied but C is not and it's satisfied otherwise right that's how we Define the truth of the sentence on line M right and and let's suppose for the sake of argument that we know that that line is true but we know that can be true in the case where B is false and C is false if we allow the modus Pocus rule in then we have a derivation from a to c right but a does not semantically entail C so the system of ql plus modus Pocus is not a sound system so conjunction introduction is sound sound but modus Pocus is not sound okay so that's soundness we also have the property of completeness right a proof system is complete if there is a proof of every valid argument right which is to say if a semantically entails B that implies that b is derivable from a now both SL and ql are complete and sound but not every logical system is in fact any system that is strong enough to express the basic arithmetic of natural numbers is incomplete right that was uh that was proved in the early 20th century by Kurt gerell right there will always be statements about natural numbers that are true but are unprovable within a proof system that is uh is powerful enough to express those truths because ql is sound and complete you can use proofs or models interchangeably to establish things like validity tautology and so on and sometimes it's more convenient to use a proof other times it's more convenient to use a model let's go through the cases right let's start with the question of is a a topology if it is a toy the easiest way to show that is just to prove uh that a is derivable that a is a theorem right um but if it is not a topology the easiest way to do that is just to give a model where a is false right same deal with the contradiction except the negation is there so the easiest way to show that a is a contradiction is to prove that not a is a theorem right whereas the easiest way to show that it is not a contradiction is to give a model where a a is true the question of whether a is contingent the easiest way to show that in the affirmative is to give a model where a is true and another one where is a is false um whereas the easiest way to show that it is not contingent is to prove that it's a theorem or that its negation is a theorem right now you may be catching on to a pattern here wherever you can answer the question with a single model or a pair of models models are the easiest way to get wherever you would have to reason about all possible models it may be easier just to do a proof right let's look at some other cases you want to know whether A and B are logically equivalent right you can show this through proof by proving by deriving B from a and vice versa right you can show that they're not equivalent by giving a model where they have different truth values right to show that a set of sentences a is consistent right you can show that by just giving a model in which all of the sentences are true you can show that they're inconsistent by taking all of the sentences as a as premises and proving a contradiction and then finally to show that the argument if with premises p and conclusions C is valid all you need to do is prove C on the basis of P to show it's invalid it's easier to just give a model where p is true and C is false right and so in this way you can uh you can combine what we learned in unit six about models with what we've learned in this unit as well as unit four about proofs um in order to answer any of these types of questions about contingent or logical truths um about equivalent y consistency and validity and it would be really to your benefit to keep all of these things in mind when you come to exam s right let's try to look at a number of examples of applying this sort of proof or model approach right so here I have six questions um that I want you to try to answer either using a proof or a model so take a moment to pause the video and work through all six of these questions and we'll come back and look at them together okay let's find out how you did uh going through uh each one right uh first we want to know is for all X and for all y lxy or not lxy autology I think it probably is it certainly has that sort of um a or not a form which is which suggests topology and so our our way to show that it's a topology is just to prove that it's a theorem right to prove that we can derive it without any premises so let's go over to carap and see if we can make that work so here we go we have for all X for all y lxy or l or not lxy I think the best way to do this is to start with a conditional introduction inter indirect proof we did something very similar in SL back in unit 4 I'm going to start by assuming La I want to get La again I'm using LA because I can't use l XY I can't have Unbound variables free variables so I need to use some constants that's easy just reiteration right now I have if La then la that is um conditional introduction one to two right to get that into a disjunction form I just need to use the material conditional rule that's not lab or lab material conditional on line three we can just shift that around through commuity on line four and now I want to start introducing my Universal quantify fires I'm going to first replace B with y That's Universal introduction on line five now I'm going to place the a with an X lxy or not lxy it's Universal introduction on line six and I I forgot my quantifier there that's done okay did you get something similar when you tried this on your own uh let me know let's move on to number two we want to know is there exist an X PX and not PA a contradiction you might think well it sort of seems like it right um but pay attention to the scope of the quantifier right the existant X PX is only over the first part right so there is something that's PX but it's not PA actually that seems like pretty consistent right so let's see so if we think that it is not a contradiction all we have to do is show a model where it comes out true right so let's see if we can do that um let's start with a universe of discourse that has two items I'm thinking two items because we have to have one thing that is PX and we have to have a be not PX right and not BP not satisfy P so um if we make the extension of P0 that satisfies the left hand side there is an X PX if we make the reference of A1 then a is not P right a is not in the extension of P so that makes the right hand side of the conjunction true so the conjunction is true that means it's not a contradiction is that how you did it let me know um our third one is that is the question is for all X PX and not PA contingent if it is contingent we just have to show two models one where it's true and one where it's false but in this case I think it is not contingent because if all X are PX then a is one of all the things right so not PA couldn't be true I think this is a contradiction and so to prove that we have to derive um the negation we have to prove that the negation of it is a theorem right so let's head over to carap and give that a go to prove a theorem we don't have any premises we just we just begin with an indirect proof of some kind I am going to suggest because negation is our main connective in what we're trying to prove that we try a negation introduction proof um so we need to assume the thing that we want to uh negate the thing we want to derive a contradiction from and that's for all X PX and not PA right so we're going to assume that for reductio we got a conjunction here we can definitely use our conjunction elimination rule on line one I'm seeing Universal elimination here is the obvious possibility Universal elimination online two gets us PA remember Universal elimination you can use any constant you want even if it is in an assumption I've got PA I can get not PA from uh our conjunction elimination on line one right PA and not PA are the contradiction that we need for a reductio proof and so I can I can close out my sub proof and conclude not ax PX and not PA through negation introduction lines one through four and that has done it I hope you had a similar answer let's look at number four um here we want to know whether these two are equivalent and although you might think that they're equivalent because they have a very similar form we've just subbed out the um predicates because the predicates mean different things or can mean different things they don't have to be equivalent and this simple model can show it let's look at number five right number five we ask is this set of four sentences consistent right um and uh if they are consistent we just need to provide a model where all of them are true but looking at this set of sentences I kind of doubt that they're consistent right um we've got for all X PX in one in one sentence we've got a not PX there and another sentence I think that they're probably uh not consistent inconsistent and so I'm going to try to prove a contradiction based on this set so let's head over again to carap to see if we can show that so here we are I've loaded all of the sentences in our um in our set in as premises in carap and what I want to do is I want to try to find a contradiction I want to derive a contradiction let's look at what we've got we've got a universal right so one thing we could do is we could eliminate the universal on line one that's an obvious one I've got three existentials now um and so so we're going to need to use existential elimination I think to get anywhere with this let's see what we can do so let's start with line three that's where our not p is at um so what I'm going to try here is is QA and not PA that's an assumption I can get not PA through conjunction elimination on line six I can get there exists an X not PX through existential introduction on line seven okay that's gotten rid of my a from the Assumption and and so I can bring that out of the sub proof that is existential elimination on existential 3 and sub proof 6 through 8 okay how can I turn that into a [Music] contradiction I think I see it right I can do quantifier negation to move that negation out front here not for all X PX that's a quantifier negation online n right whenever I uh do that exchange of the negation and the quantifier I I flip it from one to the other not ax PX um if we combine it with ax PX we get a straightforward contradiction there and that is what I'm trying to prove that's just conduction introduction on lines one and 10 um so actually I didn't really need this line five that was was unnecessary that was a a dead end which is okay that happens there may have been another way to get a contradiction but that is the way that I try to do it now let's see about this uh last question here we want to know whether this argument for all X if PX then QX QA therefore PA is valid this looks invalid to me this is uh what we might call the fallacy of affirming the consequent right um if we had PA we could derive QA if we had not QA we could derive not PA but I don't think we can derive anything from Q from QA and this conditional let's show this is invalid by coming up with a model like I've said before I'm going to start with a model with just one item in it I might have to introduce additional items as we go through um in this case I think I won't I can set the extension of P to the empty set right that means that the universal for all X PX and QX is true um trivially right um uh all PS are Q's when there are no PS no matter what we might think P represents um if we set the uh extension of Q to one and the referent of a to one then QA is true but PA is false because there's nothing in the extension of P including the reference of a and so this shows that the argument is invalid right what we've been doing today is just using our knowledge from unit six on semantics and models with our uh Knowledge from Unit Seven here on proofs combining them together due to the soundness and completeness of ql to to solve questions about logical truth contingency validity consistency Etc um using both models and proofs as is most appropriate you should practice this because it is going to feature centrally in the last exam exam 7 so good luck with the practice exercises please get in touch if you need some help um and uh good luck with it I hope you've enjoyed uh this semester and um I look forward to seeing your final progress in the class um and uh please don't hesitate to reach out if you need to bye