2

Set theory proof: Let $\mathbf{A}$ be the set such that $\{0,1,2,...,n\} \subset \mathbf{A} \implies n+1 \in \mathbf{A}$. Our goal is to show that $\mathbf{A} = \mathbb{N}$. To do this, we construct the set $\mathbf{B}$ such that it contains $n$ whenever $\{0,1,2,...,n\} \subset \mathbf{A}$. Now we induct on $\mathbf{B}$. Clearly it contains $0$. Suppose it contains $n$, then $\{0,1,2,...,n\} \subset \mathbf{A}$. By the hypothesis of complete induction, $n+1 \in \mathbf{A}$, hence $\{0,1,2,...,n+1\} \subset \mathbf{A}$, hence $n+1 \in \mathbf{B}$. So $\mathbf{B} = \mathbb{N}$, which implies that $\mathbf{A} = \mathbb{N}$, and we're done.

Logic proof: Let $\mathbf{Q}(n)$ $\equiv$ $\ \mathbf{P}(m)$ for all $0 \leq m \lt n$. Inducting on $n$, we're done.

Now comes the question that's been puzzling me: Naively, I would've thought that the set theory proof is just the logic proof written in the language of sets. But this doesn't seem to be the case, in the set theory proof, we constructed the set $\mathbf{B}$ which is in some sense "weaker" than $\mathbf{A}$. However, the logic proof hinged on the statement $\ \mathbf{Q}(n)$, which is "stronger" than $\mathbf{P}(m)$. This means (I think) the set theory proof isn't a direct translation of the logic one. Is there a deeper reason why this is the case? What's the difference between the language of sets and mathematical logic using quantifiers?

Maxis Jaisi
  • 1,553
  • What are $\bf Q$ and $\bf P$? – Asaf Karagila Jul 24 '16 at 08:11
  • Another way of doing induction is by contradiction. If $A \neq \Bbb N$, then there is a lowest $k$ such that $k \notin A$. Thus ${0,\ldots,k-1}\subseteq A$, which implies $k \in A$, and we're done. – Arthur Jul 24 '16 at 08:38
  • @AsafKaragila, $\mathbf{P}(n)$ refers to an arbitrary statement that's true for $0 \leq m \lt n$, perhaps I shouldn't have used boldface. – Maxis Jaisi Jul 24 '16 at 08:39
  • Ah, okay. Now I get it. But what I don't get now, is why "Inducting on $n$, we're done" is any different than the "set theory proof". Here you just don't fill in the details. – Asaf Karagila Jul 24 '16 at 08:41
  • $\mathbf{Q}(n) \implies \mathbf{P}(n)$, but $\mathbf{A} \implies \mathbf{B}$. In the set theory version we induct on the weaker set, but in the other proof we induct on the stronger statement. Are they similar? – Maxis Jaisi Jul 24 '16 at 08:48
  • 1
    @MaxisJaisi: Your comment about $A \to B$ makes no sense. $A,B$ are sets, and don't imply anything. However $n \in B \to n \in A$, so the direction is same as with $P,Q$. – user21820 Jul 24 '16 at 10:17
  • "Let $A$ be the set such that ${0,1,2,...,n}\subset A⟹n+1\in A$. Our goal is to show that $A=\mathbb N$." — That doesn't follow from the premise. All that follows is $\mathbb N\subset A$ (and even that only if you additionally assume $0\in A$). – celtschk Jul 24 '16 at 10:30

1 Answers1

2

In my answer I'm assuming that you also have $0 \in A$ and that $A$ is a set of natural numbers. Otherwise as celtschk said in a comment your proof is invalid since it could be that $A$ contains things besides natural numbers. The logical flaws in your proof are: $\def\nn{\mathbb{N}}$

  • You claimed that $0 \in B$. This would require $0 \in A$ by definition, but you didn't give that as a premise.

  • Your induction on $B$ only gives $\nn \subseteq B$ and you have no control over what other objects are in $B$.

With that, the set-theoretic proof does actually correspond to the logic proof. After all:

$Q(n) \equiv \forall m \in \nn_{<n}\ ( P(m) )$, for every $n \in \nn$.

$B = \{ n : \forall m \in \nn_{\le n}\ ( m \in A ) \}$.

The only difference is that your $B$ uses "$\le n$" while your $Q$ uses "$<n$". For natural numbers both turn out to be essentially the same, but later if you ever want to learn about transfinite induction (which is in a nutshell induction on sequences longer than $\nn$), then the latter on is the 'structurally correct' one to use.

Anyway let's see both proofs of strong induction (from ordinary induction) in full:


$\forall n \in \nn\ ( \forall m \in \nn_{<n}\ ( P(m) ) \to P(n) )$.   [premise]

Let $Q(n) = \forall m \in \nn_{<n}\ ( P(m) )$, for every $n \in \nn$.

Then $Q(0)$.   [Trivially because $\nn_{<0}$ is empty.]

Given any $n \in \nn$ such that $Q(n)$:

$\forall m \in \nn_{<n}\ ( P(m) )$.   [(1)]

$P(n)$   [(2); by the premise.]

  Given any $k \in \nn_{<n+1}$:

    $k < n$ or $k = n$   [By properties of $\nn$.]

    If $k < n$:

      $P(k)$.   [By (1).]

    If $k = n$:

      $P(k)$.   [By (2).]

    Therefore $P(k)$.

  Therefore $\forall k \in \nn_{<n+1}\ ( P(k) )$.

  Thus $Q(n+1)$.

Therefore by induction $\forall n \in \nn\ ( Q(n) )$.

Given any $n \in \nn$:

$Q(n+1)$.

  Thus $\forall m \in \nn_{<n+1}\ ( P(m) )$.

  Thus $P(n)$.   [Since $n<n+1$ by properties of $\nn$.]

Therefore $\forall n \in \nn\ ( P(n) )$.


$\forall n \in \nn\ ( \forall m \in \nn_{<n}\ ( m \in A ) \to n \in A )$.   [premise]

Let $B = \{ n : \forall m \in \nn_{<n}\ ( m \in A ) \}$.

Then $0 \in B$.   [Trivially because $\nn_{<0}$ is empty.]

Given any $n \in \nn$ such that $n \in B$:

$\forall m \in \nn_{<n}\ ( m \in A )$.   [(1)]

$n \in A$   [(2); by the premise.]

  Given any $k \in \nn_{<n+1}$:

    $k < n$ or $k = n$   [By properties of $\nn$.]

    If $k < n$:

      $k \in A$.   [By (1).]

    If $k = n$:

      $k \in A$.   [By (2).]

    Therefore $k \in A$.

  Therefore $\forall k \in \nn_{<n+1}\ ( k \in A )$.

  Thus $n+1 \in B$.

Therefore by induction $\forall n \in \nn\ ( n \in B )$.

Given any $n \in \nn$:

$n+1 \in B$.

  Thus $\forall m \in \nn_{<n+1}\ ( m \in A )$.

  Thus $n \in A$.   [Since $n<n+1$ by properties of $\nn$.]

Therefore $\forall n \in \nn\ ( n \in A )$.

Thus $\nn \subseteq A$.

[If you further want $A = \nn$ you need an extra premise $A \subseteq \nn$.]

user21820
  • 57,693
  • 9
  • 98
  • 256
  • In my proof I constructed $\mathbf{B}$ so that $n \in \mathbf{A} \implies n \in \mathbf{B}$. Does this not work? – Maxis Jaisi Jul 24 '16 at 12:35
  • @MaxisJaisi: No you didn't at all construct $B$ as you claimed in your comment. By your own definition; how do you know without induction that if $n \in A$ then $n \in B$? – user21820 Jul 24 '16 at 12:40
  • Could you please include an edit in your answer with a full proof using set theory? It seems like my proof is full of holes (which also means that my understanding is shaky). – Maxis Jaisi Jul 24 '16 at 12:50
  • @MaxisJaisi: Ok I've added both the logic and set-theory proofs of strong induction. By the way, I'm using a Fitch-style natural deduction (see my profile page) which will enable you to see clearly the logical structure of the entire proof once you are familiar with it. – user21820 Jul 24 '16 at 13:28
  • Thank you so much for your help. If you don't mind me asking, is there a difference between the language of logic, and that of sets? – Maxis Jaisi Jul 24 '16 at 15:03
  • @MaxisJaisi: Set theory was invented to capture how mathematicians have intuitively been reasoning about collections. First-order logic has predicates that can do that in only a rudimentary way since predicates are not objects in the domain of discourse, so quantifiers do not range over them. In contrast, sets in a set theory are objects in the domain of discourse, which is also called the set-theoretic universe. There is of course a trade-off; you can add a new predicate to stand in for any first-order formula, but because of Russell's paradox not every set can do that. – user21820 Jul 24 '16 at 17:02