A brief description of the axiom of choice, with some philosophical considerations

1. On the Axiom of Choice
Flora Dellini
Marco Natale
Francesco Urso

2. Preliminary
›For every set S, a set U is a subset of S
if, for every item in U, this item belongs to
S too.
›For every set S, we define the set of the
parts of S, P(S), as the set of all the
possible subsets of S.
Example:
𝑆 = 1,2,3
𝑃 𝑆 = {∅, 1 , 2 , 3 , 1,2 , 1,3 , 2,3 , {1,2,3}}

3. The Axiom of Choice (AC)
› The Axiom of Choice is a statement about the
existence of a certain kind of functions.
› A choice function is a function which selects an
item from a subset of a given set.
› AC claims that, for every group of subsets of S,
there exists a function of choice which selects a
particular item from every given subset.
∀𝑆, ∀𝑈 ⊆ 𝑃 𝑆 ∅, ∃𝑓: 𝑈 → 𝑆
such that ∀𝑋 ∈ 𝑈, 𝑓(𝑋) ∈ 𝑋

4. AC: Examples with finite sets
When a set is finite, everything is trivial. The
existence of f is not disputable: we can actually
show and build it!
𝑆 = 1,2,3,4,5
𝑈 = { 1 , 1,2,3 , 2,3 , 3,4,5 } ⊆ 𝑃(𝑆)
𝑓: 𝑈 ⊆ 𝑃(𝑆){∅} → 𝑆
𝑓 1 = 1 ∈ 1
𝑓 1,2,3 = 3 ∈ 1,2,3
𝑓 2,3 = 2 ∈ 2,3
𝑓 3,4,5 = 5 ∈ 3,4,5

5. AC: Examples of Infinite Sets
Hilbert’s Shoes and Socks
› Let’s suppose there is a shop with infinite pair of
shoes, we want to built an infinite set containing
a shoe for each pair.
› A Turing machine (i.e. a personal computer) can
choose between right or left shoe because it
can distinguish them. For example we can build
the set of all right shoes.
› Can we do the same with
infinite pair of socks?

6. AC: Examples of Infinite Sets
Hilbert’s Shoes and Socks
› A machine can not do it while a man could do.
› The reason is: right and left shoes can be
distinguished due to this feature. A man can do
this, a machine can do it too.
› It is not possible to choose right or left socks,
because there are no such things as right or left
socks!
› As a consequence, in principle, a compuer can
not build the wanted set.

7. AC: Examples of Infinite Sets
Hilbert’s Shoes and Socks
›Could a man build a set from a single
sock for each pair?
›Actually, it is impossible, because we
would die before we can name all the
items in the set.
›But is it possible in principle? How can we
distinguish between the socks of a pair?
›We do this by choosing “this one”, without
any criterion but our free will.

8. AC: Examples of Infinite Sets
Hilbert’s Shoes and Socks
›As human beings, it is in our everyday
experience that we can distinguish
between two socks by calling «this
one» or «that one».
›Is it admissible that in maths it is also
possible to act in such a way?
›The answer is far from trivial!

9. The Way of Formalism
› The mathematical concept of Set raised up in
XIX century by Georg Cantor (1845-1918).
› Cantor’s idea of Set was a “collection” of
“objects” which satisfy certain properties (e.g.
“the set of all odd numbers”, “the set of all the
right shoes”).
› Paradoxes arise from a “too free” use of the
concept of “property”.

10. The Way of Formalism
› Russell Paradoxes: the set of all the sets which
don’t contain themselves.
› The problem is in the semantic.
› D. Hilbert’s “Formalism” school proposed the
reduction of mathematics to a pure “formal
game”. In this way, Mathematics would have
been stripped of all its “human components”.

11. The Way of Formalism
› Zermelo and Fraenkel, following Hilbert’s intuition,
began to develop the Formal Set Theory (also
known as ZF).
› In according to the formalistic concept of
mathematics, the semantic in ZF is “eliminated”
reducing the concept of proprieties to pure
syntactic formulas, computable in principle by a
machine.
› The nature of Sets is so implicitly defined by
syntactic formulas.

12. The Way of Formalism
›The validity of a formula must be
determined through “propositional
calculus”, an absolutely formal procedure
which can be, theoretically, implemented
on an ideal machine.
›E.g. instead of saying “there is the empty
set”, we shall write the following formula:
∃𝑦 ∀𝑥 (𝑥 ∉ 𝑦)

13. The Axioms of Zermelo-Fraenkel
1. Axioms describing implicitly the concept of Set
(Regularity and Extensionality Axioms).
2. Axiom of Existence. The unique axiom of
existence in ZF is the Axiom of Infinity, which
asserts that there exists an infinite set.
3. Axioms of Individuation, which allow us to
“individuate” (i.e. build) new sets starting from
ones already known (Axiom of Power Set, Axiom
of Union and Axiom of Replacement).

14. Is AC compatible with ZF?
› Being compatible means that, if we add AC to ZF,
we cannot deduce a theorem and its negation.
› ZF claims to describe all mathematical universe:
“…with regard to ZF it’s hard to conceive
of any other model”. P. Cohen.
› Because we would like to proceed in maths as we
do with socks, that is by choosing items as we
want to, the compatibility of AC with ZF is highly
desirable.

15. ZF does not disprove AC
“Inside only ZF, it’s not possible to prove
that AC is false” (Gödel, 1938).
Main steps of proof:
1. Gödel added another axiom to ZF
(“every set is constructible”), obtaining
the stronger theory ZFL.
2. Gödel proved that ZFL is consistent. In
this stronger theory he proved that
every set can be well-ordered, that is a
demonstration of AC. So, ZFL → AC.

16. ZF does not disprove AC
3. If, by contradiction, AC is false in ZF, it has to
be false also in ZFL. But we have just seen
that AC is true in ZFL! So AC could not be
disproved in ZF.
ZF does not disprove AC!
This does NOT mean that AC is true in ZF
!

17. ZF does not imply AC
In 1960 Cohen has completed
Gödel’s demonstration about
independence of AC from ZF.
So ZF does not imply AC.

18. Independence of AC from ZF
As a consequence, we can add or
remove AC from ZF as we like.
So, its presence is actually a
preference of the mathematician who
can want a “richer” or “poorer”
theory.

19. Theorems we lose without AC
• Every non empty Vectorial Space has a
base
– i.e. imagine that, in the classic Euclidean
3D space, you don’t have the 𝑖, 𝑗, and 𝑘
vectors you use to build every other
vector.
• Every field has an algebraic closure
– i.e. imagine that you could not define the
complex numbers

20. What does it mean?
“Simply” that there are sets whose volume is not invariant under
translation and rotation.
Strange! Isn’t it?
This is why some mathematicians worry
about the Axiom of Choice
Decomposition of a ball into four pieces which, properly
rotated and traslated, yield two balls
Counterintuitive effects of AC:
Banach-Tarski Paradox

21. Counterintuitive effects of AC:
Zermelo’s Lemma
This statement is equivalent to the Axiom of
Choice:
Every set S can be well ordered
As a consequence, we could «well order»
ℝ - i.e. defining an order in ℝ such that
every subset of ℝ has a minimum. This
order relation is strongly counterintuitive, as
it implies that sets like (0,1), without a
minimum – 0 ∉ (0,1) do not exist.