You are on page 1of 4

The Limits of Logic

Dr. Mark Lee mgl@cs.bham.ac.uk

Autological (adj) An adjective is autological if and only if it describes itself. For example the word "short" is short and so describes itself. "Pentasyllabic" and "twelve letter" are also autological. Hetrological (adj) An adjective is hetrological if and only if it does not describe itself. Hence "long" is hetrological as is "monosyllabic." To test if a word "foo" is autological you can ask: "is foo a foo word?" if the answer is yes then "foo" is autological. If the answer is "no" then "foo" is hetrological. If this seems easy then is "hetrological" a hetrological word? Hilbert's Program (A dream in the 1920s) A formalization & Automated Proof of all mathematics Completeness: a proof that all true mathematical statements can be proved in the formalism. Consistency: a proof that no contradiction can be obtained in the formalism of mathematics. Conservation: a proof that any result about "real objects" obtained using reasoning about "ideal objects" (such as uncountable sets) can be proved without using ideal objects. Decidability: there should be an algorithm for deciding the truth or falsity of any mathematical statement. Introduction As the last formal lecture of this module, it's worth reflecting on what we've covered. The module was structured as follows: Natural Language Syntax Meaning (in so far as natural language arguments) Propositional Logic Syntax Truth tables Natural Deduction Predicate Logic Syntax Natural Deduction Natural language we showed was often highly ambiguous and vague. This has allowed us to construct a number of (hopefully) amusing and thought provoking logic puzzles but is clearly not a good basis for a formal account of logic. Propositional Logic is a good first step into formal language. With a small number of operators (, &, V, ~) and a set of symbols for propositions, we were able to represent a wide range of arguments in logic and show if they were valid or invalid. We covered two possible methods to do this: truth tables and natural deduction. Unfortunately, Propositional Logic cannot deal with statements involving generalisations. Predicate logic introduces two quantifiers and the notion of predicates and arguments. In addition, it only requires a small addition to the set of rules for Natural Deduction. Two (and a bit) Proofs Consider the following problem P Q, ZM, P & A, ~M: Q & ~Z It should be easy to prove the validity of this using a truth table (but since there are 5 propositions I'll leave this as homework). The other alternative is to use Natural Deduction.

How should we approach this problem? Recall the three rules of thumb: 1) If there's an implication in the conclusion then hypothesise its antecedent and prove the consequent using the premises. 2) If there's a disjuction in the premises then hypothesise either disjunct and use disjunction elimination. 3) If all else fails - use RAA. So what should we do in this case? Hypothesis ~(Q & ~Z) and provide a contradiction? If so we might be making more work for ourselves. The following proof works fine: (Let's prove Q) 1. P Q Premise {1} 2. P & A Premise {2} 3. P &E 2 {2} 4. Q E 1,3 {1,2} (Now to prove ~Z. The obvious way is to use RAA) 5. | Z H {5} 6. | Z M Premise {5,6} 7. | M E 5,6 {5,6} 8. | ~M Premise {5,8} (Lines 7 and 8 are a contradiction and therefore) 9. ~Z RAA 5,7,8 {6,8} (Now we've proven both Q and ~Z, we can put them together to finish) 10. Q & ~Z &Intro 4,9 {1,2,6,8} Now consider the above problem's big brother: x [Px Qx], x [Zx Mx], x [ Px & Ax], x [ ~Mx] : x [Qx & ~Zx] This shouldn't be any harder than the above. Since the premises include an existential, we need to choose a "typical" individual to reason about as an hypothesis. After proving what we need to about this particular individual, we'll use Existential Introduction to stop talking about this particular individual and then Existential Elimination. The other premises are Universals, so we'll use Universal Elimination to change them to be about our "typical individual". Finally, the conclusion is existentially qualified so we'll use Existential Introduction (and do so before Existential Elimination). This probably sounds complex but it's not if you work through the following carefully. 1. x [Px Qx] 2. | Pa Qa 3. | x [Zx Mx] 4. | Za Ma 5. | x [ Px & Ax] 6. | Pa & Aa 7. | x [ ~Mx] 8. | ~Ma Premise {1} H based on 1 {2} Premise {2,3} UE 3 {2.3} Premise {2,5} UE 5 {2,5} Premise {2,7} UE 7 {2,7}

At this point, we can now use the same approach as directly above. 9. | Pa 10| Qa 11| | Za 12| | Ma 13| | ~Ma 14| ~Za &E 6 { 2,5} E 2,9 {2.5} H {2.11} 11, 4 {2.3.11} i 7 {2,7} RAA 11,12,13 {2,3,11}

15| Qa & ~Za &I 14, 10 {2,5,3,11} Finally EI before EE 16| x [Qx & ~Zx] EI 15 {2,5,3,11} 17 x [Qx & ~Zx] EE 16 {5,3,11} Soundness and Completeness (This is recap from earlier lecture notes) At the start of the module we defined Validity as follows: Semantic Validity An inference is valid iff it cannot be the case that all the premises are true and the conclusion false. In propositional logic, the obvious way to prove this would be to construct a truth table and show that under all possible combinations of truth, the inference is valid. Under Natural Deduction, we've used a different notion of validity Syntactic Validity An inference is syntactically valid iff the conclusion can be derived from the premises by means of stipulated rules of inference. This notion appears weaker than semantic validity, In fact we use a different notation for semantic and syntactic validity. P Q, P |- Q Syntactically valid theorem P Q, P |= Q Semantically valid theorem Soundness Given any logic, it's necessary to ask whether every syntactically provable argument is also semantically valid. This is called Soundness. Soundness |- S => | = S Both Propositional Logic and Predicate Logic with ND (happily) have this feature. Completeness Soundness guarantees that any syntactically valid proof can also be shown to be semantically valid. Completeness is the complementary concept: Given a semantically valid theorem, can we guarantee that a syntactic proof exists? i.e. Completeness |= S => |- S Both Propositional Logic and Predicate Logic with ND (happily) have this feature. Decideability The third and final property we'd like to see in a formal logic is decideability. Decideability For every argument, there is a procedure to decide whether or not it is valid or invalid in a finite number of steps. Propositional Logic is decidable. Predicate Logic with single arity relations is also decidable (i.e. restricted to just predicates such as x [ Fx Mx]). Predicate Logic with multiple arity relations is semi-decidable (i.e. you can show validity but not invalidity). Further restrictions can be placed multiple arity predicate logic to make it decidable. If you study Prolog in your second year then this is largely how prolog works.

Beyond Logic (This section is purely for interest. The material is far beyond this module, but I hope it will motivate you to study further). During the end of the 19th Century, a German Mathematician called Frege attempted to provide the foundations for all mathematics in Logic. His starting point was to define numbers in terms of (naive) set theory. For example, Frege defined the number "1" to be "One is the set of numbers such that some X is in the set and if any Y is a member then X = Y" The number "2" can be defined as "Two is the set of all sets such that X and Y are in the set, X does not equal Y and if Z is a member then Z = X or Z = Y" Set theory was chosen because Mathematicians regarded that 1) it was well understood 2) logically sound. Unfortunately Bertrand Russell (1901) proposed the following paradox regarding sets: Divide sets into two sorts: Sets that are self membering and sets that are non-self membering. For example the set of all fruits is non-self-membering (because it's a set, not a fruit) but the set of all sets is self membering (because it's a set). What about the set of non-self-membering sets? Is it self-membering? This is a paradox and shows that naive set theory isn't as well founded as previously assumed and therefore Frege's project was doomed. Godel's Incompleteness Theorems It turns out however that Godel (1931) (using a similar argument involving paradox but beyond this module) showed the following: Godel's 1st theorem Any effectively generated theory capable of expressing elementary arithmetic cannot be both consistent and complete. In particular, for any consistent, effectively generated formal theory that proves certain basic arithmetic truths, there is an arithmetical statement that is true, but not provable in the theory. Godel's 2nd theorem For any formal recursively enumerable (i.e., effectively generated) theory T including basic arithmetical truths and also certain truths about formal provability, T includes a statement of its own consistency if and only if T is inconsistent. That is to say that any any formal system that includes enough of the theory of the natural numbers is either incomplete or inconsistent. In other words Hilbert's program (see above) is (mostly) impossible. Since then current research in logic, computer science and mathematics have redefined Hilbert's program. It's still an unsolved question, to what degree the foundations of mathematics can be fully formalized.

You might also like