Consider ∃x∀yAttendClass(x,y) and ∀y∃xAttendClass(x,y).
If there is a student s attending all classes, then each class is attended by at least a student, i.e. s; if each class is attended by a student, this does not imply that there is a student attending all classes.
Relationships between quantifiers
Relationships between quantifiers
Consider the formula, “no university has no student”, it is equivalent to “all universities have at least one student”, which shows that quantifiers are linked.
Problem: represent and negate “there is a student attending the ELA module”
Encode as a formula: ∃xAttendClass(x,ela)
Obtain negation: ¬∃xAttendClass(x,ela)
Use rules demonstrated earlier to obtain ∀x¬AttendClass(x,ela)
This produces two different statements:
Negation: “there is no student attending the ELA module”
Generated from rule: “all students do not attend the ELA module”
Example 2: Interpreting Negated Formulae 2
Problem: represent and negate “all universities have at least one student”
Encode: (∀x)(University(x)→(∃y)Student(y,x))
Negate: ¬(∀x)(University(x)→(∃y)Student(y,x))
Transform using rules:
¬(∀x)(Uni(x)→(∃y)Student(y,x))≡(∃x)¬(Uni(x)→(∀y)Student(y,x))≡(∃x)¬(¬Uni(x)∨(∃y)Student(y,x))≡(∃x)(Uni(x)∧¬(∃y)Student(y,x))≡(∃x)(Uni(x)∧(∀y)¬Student(y,x))
The last formula reads “there is a university without any student”.
Example 3: Showing Equivalence
Problem: show that the following two formulas are equivalent.
Vacuous quantification is when a variable x has no occurrence in a formula ϕ, hence then binding x with a quantifier in ϕ has no effect. Or when the variable x is already bound by another quantifier in a formula ϕ, hence binding x again has no effect.
Similarly to propositional logic, we can obtain logical consequences in FOL using deduction rules. All natural deduction rules involving the propositional connectives and seen for the propositional logic case are still valid.
New deduction rules need to be provided for the quantifiers.
Universal Instantiation (∀-elimination)
Universal instantiation
A universal instantiation means for each and every entity $c$ in the domain we have a correct inference rule of the form: $$\boxed {\begin{array}{c}\underline{\forall xP(x)} \\P(c)\end{array}}$$
For example,
∴Everyone in the class is feeling miserable today.Agi is feeling miserable today.
By universal instantiation, given the premise (∀x)ϕ(x), we may derive the conclusion ϕ(x/t), where t is any object in the domain:
ϕ(x/t)(∀x)ϕ(x)(UI)
If ϕ(x) holds for all the elements of the domain, it holds in particular for the element t as well. Universal instantiation can be seen as ∀-elimination.
title: ExampleValidate the following inference:- Every boy likes Mary- John is a boy- **Therefore**: John likes Mary$$ \begin{aligned} &1. &&\forall x(\text{Boy}(x) \rightarrow \text{Likes}(x, mary)) &&\text{data} \\ &2. &&\text{Boy}(john) &&\text{data} \\ &3. &&\text{Boy}(john) \rightarrow \text{Likes}(john, mary) &&\text{from } (1.) \text{ and } \forall E: x / john \\ &4. &&\text{Likes}(john, mary) &&\text{from } (2.), (3.), \text{ and } (\rightarrow E) \end{aligned}$$
Universal Generalisation (∀-introduction)
Universal generalisation
A universal generalisation is $$ \boxed{\begin{array}{c} \underline{P(c) \textsf{ for any arbitrary entity } c \textsf{ in the domain}} \\ \forall x P(x)\end{array}} $$
By universal generalisation, given the premise ϕ(x/t) takes place for a generic object t of the domain, we may derive (∀x)ϕ(x):
(∀x)ϕ(x)ϕ(x/t)(UG)
The object t in the premise of (UG) must not be specific but arbitrary, i.e. we cannot make any assumption about t other than it comes from the domain. Universal generalisation can be seen as ∀I.
title: ExampleValidate the following inference:- All humans are mammals- All mammals are vertebrates- **Therefore**: All humans are vertebrates$$ \begin{array}{l} \begin{aligned} &1. &&\forall x (\text{Human}(x) \rightarrow \text{Mammal}(x)) &&\text{data} \\ &2. &&\forall x (\text{Mammal}(x) \rightarrow \text{Vertebrate}(x)) &&\text{data} \\ &3. &&\text{Human}(a) \rightarrow \text{Vertebrate}(a) &&\text{subcomputation} \end{aligned} \\ \boxed{ \begin{aligned} & && && &&\underline{\text{Vertebrate}(a)} \\ &3.1 &&\text{Human}(a) &&\text{assume} \\ &3.2 &&\text{Human}(a) \rightarrow Mammal(a) &&\text{from } (1.), \text{ and } \forall E: x/a \\ &3.3 &&\text{Mammal}(a) &&(3.1), (3.2), \text{ and } \rightarrow E \\ &3.4 &&\text{Mammal}(a) \rightarrow \text{Vertebrate}(a) &&(2.), \text{ and } \forall E: x/a \\ &3.5 &&\text{Vertebrate}(a) &&(3.3), (3.4), \text{ and } \rightarrow E \end{aligned} } \\ \begin{aligned} &4. &&\forall x (\text{Human}(x) \rightarrow \text{Vertebrate}(x)) &&(3.), \text{ and } \forall I: a / x \end{aligned} \end{array}$$
Existential Generalisation
Existential generalisation
An existential generalisation means that for every entity c in the domain, we have a correct inference rule of the form $$ \boxed{ \begin{array}{c} \underline{P(c)} \\ \exists x P(x) \end{array} } $$
For example:
∴Marla is reading the lecture notes.There is someone in the class reading the lecture notes.
By existential generalisation, given that ϕ(x/t) is known to hold for a particular object t of the domain, we may derive (∃x)ϕ(x):
title: ExampleValidate the following inference:- Anne likes Bob- **Therefore**: Someone likes Bob$$ \begin{aligned} &1. &&\text{Likes}(ann, bob) &&\text{data} \\ &2. &&(\exists x) \text{Likes}(x, bob) &&(1.), \text{ and } \exists I: ann/x \end{aligned}$$
Existential Instantiation (∃-elimination)
Existential instantiation
By existential instantiation, given the premise (∃x)ϕ(x), we may derive ϕ(x/t) where t is a particular object of the domain:
ϕ(x/t)(∃x)ϕ(x)(EI)
Note: we cannot select an arbitrary object in (EI) but rather it must be an object t for which ϕ(x/t) is true. Usually, we have no knowledge of what the actual object is, we only know that it does exist. Since it exists, we can give it a new name, not used anywhere before, say t, and continue the derivation.
Existential instantiation can be seen as ∃E.
By identity elimination (substitution of identicals), if t and t′ are terms, from t=t′ and ϕ(x/t), we may derive ϕ(x/t′) (we substitute t for t′ since they are identical):
We can prove that the equality relation is symmetric, i.e. prove that:
⊨∀x∀y(x=y→y=x)1.a=b→b=asubcomputation1.11.21.3a=ba=ab=aassumeII(1.1),(1.2)(ϕ(x):x=a;x/a), and IE:x/bb=a2.3.∀x(x=b→b=x)∀x∀y(x=y−>y=x)from (1.) and ∀I:a/xfrom (2.) and ∀I:b/y