You are on page 1of 9

Philosophy 350: Logic Homework 3 Ian Cramer

PHIL 350: Problem Set 3


Spring 2023
Due May 25, 11:59pm.

Problem 1
(20 points)

Complete these parts of the proof of the Soundness Theorem:

1. Show that all axioms of the form ϕ → (ψ → ϕ) hold on all models. [10pts]

Solution. Proof
Let M be a model and assume for contradiction M ̸|= ϕ → (ψ → ϕ). Then M |= ϕ and
M ̸|= ψ → ϕ. This means M |= ψ and M ̸|= ϕ, but this is a contradiction because we know
M |= ϕ. Thus M |= ϕ → (ψ → ϕ), and by the arbitrary nature of M we can conclude that
axioms of the form ϕ → (ψ → ϕ) hold on all models.

2. Show that all axioms of the form (¬ϕ → ¬ψ) → (ψ → ϕ) hold in all models. [10pts]

Solution. Proof
Can you prove that the notion of contrapositive is true by contrapositive?
Let M be a model and assume for contradiction M ̸|= (¬ϕ → ¬ψ) → (ψ → ϕ). Then

M |= ¬ϕ → ¬ψ (1)
and
M ̸|= ψ → ϕ (2)

From line 1 above we get that either M |= ¬ψ or M ̸|= ¬ϕ.


Suppose we are the first case is true, so M |= ¬ψ. From line 2 above we get that M |= ψ
and M |= ¬ϕ, but this is a contradiction that M |= ¬ψ. Thus the first case cannot hold.
Then we have the second case must be true, that M ̸|= ¬ϕ. Again, Ffrom line 2 above we
get that M |= ψ and M |= ¬ϕ, but this is a contradiction that M ̸|= ¬ϕ. Thus the second
case cannot hold.
So, in either case we get a contradiction, and therefore, by the arbitrary nature of M, axioms
of the form (¬ϕ → ¬ψ) → (ψ → ϕ) hold in all models.

1
Philosophy 350: Logic Homework 3 Ian Cramer

Problem 2

Here is a theory in the language whose non-logical vocabulary is as follows:

The 1-place predicate S:


(Pronounce “Sx” by “x is a Shark”.)
The 1-place predicate J:
(Pronounce “Jx” by “x is a Jet”.)
The 2-place predicate L:
(Pronounce “Lxy” by “x likes y”.)
The 2-place predicate H:
(Pronounce “Hxy” by “x hates y”.)

Here are axioms of the theory T (call the theory: West Side Story): the universal closures of:

(1) Sx ↔ ¬Jx (4) Lxy ↔ ¬Hxy


(2) Sx ∧ Jy → Hxy (5) Sx ∧ Sy → Lxy
(3) Sx ∧ Jy → Hyx (6) Jx ∧ Jy → Lxy

(a) Show that L forms an equivalence relation on the domain of any model of the theory T . [25
pts]

Solution. Proof
Reflexivity: ∀a, either Sa or Ja, so by elements 5 and 6 of the theory we have that Laa
Symmetry: Let a, b such that Lab. Then there are four cases:
1. Sa and Sb so by element 5 of the theory with x = b and y = a, we get that Lba
2. Ja and Jb so by element 6 of the theory with x = b and y = a, we get that Lba
3. Sa and Jb. Then by element 2 of the theory Sa ∧ Jb → Hab and by the contrapositive
of element 4 of the theory Hab → ¬Lab, a contradiction. Thus this case is impossible.
4. Ja and Sb. Then by element 3 of the theory Sb ∧ Ja → Hab and by the contrapositive
of element 4 of the theory Hab → ¬Lab, a contradiction. Thus this case is impossible.
Therefore in all possible cases L is symmetric.
Transitivity: Let a, b, c such that Lab and Lbc. Then there are eight cases:
1. Sa and Sb and Sc. So by element 5 of the theory with x = a and y = c we get that Lac
2. Ja and Jb and Jc. So by element 6 of the theory with x = a and y = c we get that Lac
3. Sa and Sb and Jc. Then by element 2 of the theory Sb ∧ Jc → Hbc and by the
contrapositive of element 4 of the theory Hbc → ¬Lbc, a contradiction. Thus this case
is impossible.

2
Philosophy 350: Logic Homework 3 Ian Cramer

4. Sa and Jb and Sc. Then by element 2 of the theory Sa ∧ Jb → Hab and by the
contrapositive of element 4 of the theory Hab → ¬Lab, a contradiction. Thus this case
is impossible.
5. Sa and Jb and Jc. Then by element 2 of the theory Sa ∧ Jb → Hab and by the
contrapositive of element 4 of the theory Hab → ¬Lab, a contradiction. Thus this case
is impossible.
6. Ja and Sb and Sc. Then by element 3 of the theory Sb ∧ Ja → Hab and by the
contrapositive of element 4 of the theory Hab → ¬Lab, a contradiction. Thus this case
is impossible.
7. Ja and Sb and Jc. Then by element 3 of the theory Sb ∧ Ja → Hab and by the
contrapositive of element 4 of the theory Hab → ¬Lab, a contradiction. Thus this case
is impossible.
8. Ja and Jb and Sc. Then by element 3 of the theory Sc ∧ Jb → Hbc and by the
contrapositive of element 4 of the theory Hbc → ¬Lbc, a contradiction. Thus this case
is impossible.

Therefore in all possible cases L is transitive.

(b) Show that for any natural numbers n and m, there is a model of T in which the first-order
domain has size n + m, and the extension of S has size n and the extension of J has size m.
[10pts]

Solution. Proof
Consider the model M with domain M = {1, ..., n + m} such that M |= Si for all i ≤ n and
M |= Ji for all i > n.
Then LM = {(a, b) | a, b ≤ n ∨ a, b > n} and H M = {(a, b) | a ≤ n < b ∨ b ≤ n < a}

(c) Show that every model with a finite first-order domain is homomorphic to a model in which
there are only two objects. [10pts]

Solution. Proof
Consider the model A with domain A = {1, 2} where, for x ∈ A, A |= Sx if x = 1 and
A |= Jx if x = 2.
Then any model with a finite first order domain is homomorphic to this model because,
consider a model B with a finite first order domain B. Then there exists a homomorphism
f : B → ⊣ Such that
(
1 B |= Sb
f (b) =
2 B |= Jb

This is a homomorphism because

• Clearly for b ∈ B if S B b then f (b) = 1 and S A f (b) and if J B b then f (b) = 2 andJ A f (b),
by construction.

3
Philosophy 350: Logic Homework 3 Ian Cramer

• ⟨a, b⟩ ∈ H B → ⟨f (a), f (b)⟩ ∈ H A because ⟨a, b⟩ ∈ H B → (S B a∧J B b)∨(J B a∧S B b) which


means by the preservation of S and J shown above (S A f (a)∧J A f (b))∨(J A f (a)∧S A f (b))
which implies ⟨f (a), f (b)⟩ ∈ H A .
• ⟨a, b⟩ ∈ LB → ⟨f (a), f (b)⟩ ∈ LA because ⟨a, b⟩ ∈ LB → (S B a ∧ J B b) ∨ (J B a ∧ S B b) which
means by the preservation of S and J shown above (S A f (a)∧J A f (b))∨(J A f (a)∧S A f (b))
which implies ⟨f (a), f (b)⟩ ∈ LA .

As there are no constants or functions this completes the proof that f is a homomorphism.

(d) Does this hold for models with infinite first-order domains? Give a counterexample if “no”,
and explain how to modify the proof if “yes”. [5pts]

Solution. Proof
Yes. The previous proof should hold without modification even for models with infinite first-
order domains.

(e) Does (c) hold if we replace “is homomorphic to” with “can be embedded into”? Give a
counterexample if “no”, and explain how to modify the proof if “yes”. [5pts]

Solution. Proof
No. Consider the model B with domain B = {1, 2, 3, 4} where for x ∈ B we have S B x if x is
odd and J B x if x is even. Assume for contradiction there exists an embedding f : B → A.
Then clearly f (1B ) = 1A and f (2B ) = 2A . But now, no matter where you map 3B or 4B you
will have a collision, two elements in B will be mapped to the same element in A because
every element in A has already been mapped to and we still have two elements left in B to
create a mapping for. Therefore (c) does not hold if we replace “is homomorphic to” with
“can be embedded into”.

4
Philosophy 350: Logic Homework 3 Ian Cramer

Problem 3

(This is called Skolem’s paradox.) Standard set theory is called Zermelo-Frankel set theory with
Choice; its axioms are usually denoted ZF C. You don’t need to know these axioms. Assume there
is an infinite model of ZF C. Here are some facts:

• Every model of ZF C contains a set N that is countably infinite. Basically, N is a copy of


the natural numbers.

• Every model of ZF C contains a set, P (N ), the power-set of N . This is the set of subsets of
N , but that’s not important for this problem.

• Every model of ZF C satisfies the sentence which ”says”: There is no injection from P (N ) to
N.

The Löwenheim-Skolem theorem (downhill) says that every first-order theory with an infinite model
in a countable language has a countably infinite model. The language of ZF C is countable. There-
fore there is an infinite model. What is the paradox? And how does one escape it? [5pts]

Solution. Proof
Start with a big model of ZFC which includes things that are uncountable. Go through the
existential sentences that are made true by this model. The domain of our new model will be
the witnesses of these sentences. There is no sentence held true by the original model stating the
model is countable because it is not. Construct a new model based on a countable selection of these
sentences (though there was only countably many to start with). The new model will not include
any sentence stating there exists a bijection with the natural numbers.
The paradox is that the model is in truth countable, but such a statement is not held true by the
model. This seems like a paradox because it seems like we have an object that is both countable and
uncountable. This in fact is not an actual paradox, it can be escaped by disambiguating between
the model and the meta theory.

5
Philosophy 350: Logic Homework 3 Ian Cramer

Problem 4

Show that for each set Nn = {m ∈ N | m < n}, there is no injection f : Nn → Nn with f (Nn ) ⊊ Nn .
[5 pts]
HINT: You can do this by induction on n, or you can use the fact that N is well-ordered. Try using
the latter argument, setting X to be the set of n such that Nn has no injection f as described.
Assume X ̸= N, so N − X is not empty. Now use well-ordering to get a contradiction.

Solution. Proof
Let X = {n ∈ N | there is no injection f : Nn → Nn with f (Nn ) ⊊ Nn }. Clearly X ⊆ N, so consider
Y = N \ X. Assume for contradiction X ̸= N, so Y is non empty. By the well ordering principle,
Y has a least element, call it k. As k ∈ Y , there exists an injection f : Nk → Nk with f (Nk ) ⊊ Nk .
Let a ∈ Nk \ f (Nk ), that is a is not mapped to by the injection f . Assume a = k − 1. This is
without loss of generality because if a ̸= k − 1, then we can construct f ′ such that
(
′ f (x) if f (x) ̸= k − 1
f (x) =
a if f (x) = k − 1

That is whatever element maps to k − 1 we set to map to a and change nothing else. This way we
can be sure k − 1 ∈ Nk \ f (Nk )
Now consider f ◦ f : f (Nk ) → f (Nk ). Obviously f (a) ∈ f (Nk ). But a ∈/ f (Nk ), so f (a) ̸= a. Since
f is injective we know that no other element in Nk maps to f (a) besides a, thus f (a) ∈ / f (f (Nk )
since a wasn’t in the domain f (Nk ) to map to f (a). But this means there exists a set f (Nk ) ⊊ Nk
such that there is an injection f ◦ f : f (Nk ) → f (Nk ) where f (f (Nk ) ⊊ f (Nk ) and f (Nk ) ⊊ Nk .
This implies there exists i < k such that there is an injection f : Ni → Ni with f (Ni ) ⊊ Ni . This
contradicts that k is the least such element that this is true. Thus Y must have no least element
and is consequently empty. Therefore for any set Nn = {m ∈ N | m < n}, there is no injection
f : Nn → Nn with f (Nn ) ⊊ Nn .

6
Philosophy 350: Logic Homework 3 Ian Cramer

Problem 5

Recall that a set X is said to be Dedekind finite if there is no injection f : X → X with f (X) ⊊ X.
But we can also say that X is (regular) finite just if there is a natural number n and a bijection
g : X → {m | m < n}.
Show that if X is Dedekind infinite, then there is an injection f : N → X. [5 pts]

Solution. Proof
First define the notation f n (x) as a function composed with itself n times, that is

f n = f ◦ f ◦ ... ◦ f
| {z }
n times

Let X be a set that is Dedekind infinite, that is there exists an injection g : X → X such that
g(X) ⊊ X. Let x ∈ X \ g(X), that is x ∈ X but x ∈ / g(X). Then g(x) ̸= x because g(x) ∈ g(X)
but x ∈/ g(X). Since g is an injection, only x is mapped to g(x). Thus g(g(x)) ̸= g(x) because then
g would not have been injective. So, if we consider g ◦ g : g(X) → g(X) we notice that x ∈/ g(x), so
g(x) ∈/ g(g(X)) because x was the only thing g could map to g(x) and x is no longer in the domain
of g ◦ g.
Then we can define the function f : N → X as

f (0) = x
f (n + 1) = g n+1 (x)

It remains to show this function is injective. Let n1 , n2 ∈ N such that f (n1 ) = f (n2 ). Then

f (n1 ) = f (n2 )
⇒ g n1 (x) = g n2 (x)

Assume for contradiction that n1 ̸= n2 . Then without loss of generality suppose n1 < n2 . Then
g n2 (X) ⊊ g n1 (X), so g n2 (X) ⊆ g n1 +1 (X). But g n1 (x) ∈
/ g n1 +1 (X). So, since g n2 (x) ∈ g n2 (X) ⊆
g n1 +1 n
(X) and g (x) ∈
1 /g n 1 +1 (X), we get that g (x) ̸= g n1 (x), a contradiction. Therefore n1 = n2
n2

and f is injective.

7
Philosophy 350: Logic Homework 3 Ian Cramer

Problem 6

Show that under the assumption that every set can be well-ordered, a set is finite if and only if it
is Dedekind finite. [5 pts]
HINT: Use the previous two problems.

Solution. Proof
First we show that if a set is finite it is Dedikind finite. Let X be a finite set. Then by problem 4
there is no injection onto a proper subset of X. Thus X is Dedekind finite.
Now we want to show that if a set is Dedikind finite it is finite. We can do this by proving the
contrapositive, that if a set is infinite then it is Dedikind infinite. Let X be an infinite set. Then
there exists and injection f : N → X. Now consider g : X → X such that
(
x x∈/ f (N)
g(x) = −1
f (2f (x))

First we show that g(x) is injective. Let x1 , x2 ∈ X such that g(x1 ) = g(x2 ). Then there are four
cases:

• Suppose x1 , x2 ∈
/ f (N). Then x1 = g(x1 ) = g(x2 ) = x2 .

• x1 ∈ f (N) and x2 ∈ / f (N). Then g(x2 ) = x2 ∈ / f (N), but g(x1 ) = f (2f −1 (x1 )) ∈ f (N).
As g(x1 ) ∈ f (N) but g(x2 ) ∈
/ f (N) we get that g(x1 ) ̸= g(x2 ). Thus this case leads to a
contradiction making it impossible.

• x1 ∈
/ f (N) and x2 ∈ f (N). Then g(x1 ) = x1 ∈ / f (N), but g(x2 ) = f (2f −1 (x2 )) ∈ f (N). As
g(x2 ) ∈ f (N) but g(x1 ) ∈
/ f (N) we have g(x1 ) ̸= g(x2 ). Thus this case leads to a contradiction
making it impossible.

• x1 , x2 ∈ f (N). Then let n1 = f −1 (x1 ) and n2 = f −1 (x2 ). Assume for contradiction that
x1 ̸= x2 . Since f is an injection, f −1 is an injection, so n1 = f −1 (x1 ) ̸= f −1 (x2 ) = n2 .
Then 2n1 ̸= 2n2 and f (2n1 ) ̸= f (2n2 ) because f is an injection. But then g(x1 ) = f (2n1 ) ̸=
f (2n2 ) = g(x2 ), a contradiction. Thus x1 = x2 .

Therefore, in all possible cases, if g(x1 ) = g(x2 ) then x1 = x2 so g is an injection.


Now we show that g(X) ⊊ X. Consider f (1) ∈ X. Assume for contradiction there exists x ∈ X such
that g(x) = f (1). First suppose x ∈/ f (N). Then f (1) = g(x) = x, but f (1) ∈ f (N) and x ∈ / f (N)
−1
so x ̸= f (1), a contradiction. Thus we conclude x ∈ f (N). Then f (1) = g(x) = f (2f (x)). Since
f is injective we can conclude that f (1) = f (2f −1 (x)) → 1 = 2f −1 (x). But then f −1 (x) = 12 ∈
/ N,
a contradiction because f −1 : X → N. Either way we have a contradiction. Thus there exists
f (1) ∈ X such that there does not exist x ∈ X such that g(x) = f (1). Therefore g(X) ⊊ X and X
is Dedekind infinite.
Therefore, by contrapositive, if a set is Dedekind finite it is finite.

8
Philosophy 350: Logic Homework 3 Ian Cramer

Collaborators

I worked with the following other students on my homework:

1. Maddy Roffey

You might also like