Isn't Induction Circular? You're Assuming What You Want to Prove

The complaint is one of the most common objections students raise when they first meet induction. You want to prove P(n) for every natural number n. The method says: "Assume P(k) is true. Using that, show P(k+1) is true." But isn't P(k) exactly the kind of statement you are trying to prove? How is this not circular reasoning — quietly assuming the conclusion to derive the conclusion?

The objection is a good one to take seriously, because seeing through it is how you understand why induction actually works. The short answer is that "assume P(k)" is not the same as "believe P(k) is true for all k without proof." It is a conditional supposition used to build a specific implication. The induction machine then chains those implications into a proof that runs on its own.

What you are really proving in the inductive step

The inductive step does not prove P(k). It does not claim P(k). It proves a conditional statement:

P(k) \Rightarrow P(k+1)

The phrase "assume P(k)" is how you start the proof of this conditional. Every time you prove a statement of the form "if A, then B" in any area of mathematics, you begin by assuming A and deriving B. That assumption does not claim A is actually true in the universe — it is the hypothesis of a conditional, used locally to derive a conclusion that is itself a conditional fact.

Why this is not circular: a circular argument uses the conclusion as a premise. Here the conclusion of the inductive step is "P(k) \Rightarrow P(k+1)" — a conditional. The premise is "P(k) — assumed." The conclusion being proved is the whole implication, not P(k+1) by itself and certainly not "P(n) for all n." The chain that eventually establishes "P(n) for all n" is built outside the inductive step, by combining many such conditionals with a separately proven base case.

The two ingredients, and how they combine

Induction has two independent ingredients, each proved on its own terms:

Base case. You prove P(1) unconditionally — a direct computation with no assumption. This is where the chain is anchored in reality.
Inductive step. You prove the conditional P(k) \Rightarrow P(k+1) for every k \geq 1. This is a conditional fact; it does not claim any specific P(k) is true.

Combining these two gives the conclusion P(n) for every n \geq 1. Here is the structure of the combination, made explicit:

P(1) is true. (By the base case.)
P(1) \Rightarrow P(2) is true. (By the inductive step with k = 1.)
Therefore P(2) is true. (Modus ponens.)
P(2) \Rightarrow P(3) is true. (By the inductive step with k = 2.)
Therefore P(3) is true.
...and so on, forever.

At no point did you assume P(3) in order to prove P(3). You derived P(3) by combining two already-proved facts: P(2) (from the previous step) and P(2) \Rightarrow P(3) (from the inductive step). Nothing is circular. The argument runs forward, one rung at a time.

The domino analogy (refined)

The classic domino picture is apt once you state it precisely:

The base case is like pushing over the first domino. That action really happens — it is an event, not an assumption.
The inductive step is like verifying that for every k, "if domino k falls, then domino k+1 falls." This is a fact about the physics of the setup — the spacing and arrangement — and can be checked by inspecting the dominoes. It does not claim that any particular domino has fallen.

When you combine "the first domino really fell" with "every falling domino topples the next," you get "every domino falls" — but you got there by running the dominoes, not by assuming the outcome. The inductive proof is the mathematical version of running the dominoes.

Where the confusion comes from

The word "assume" is overloaded in everyday English. It sometimes means "take as unjustified fact" (as in "don't assume I'm lying"). In mathematical writing, inside the phrase "to prove A \Rightarrow B, assume A and derive B," the word means something more technical: "introduce A as a temporary hypothesis for the duration of this sub-proof." The assumption is not a belief; it is a scaffold.

You can see the same pattern in direct proofs outside of induction. To prove "if x is even, then x^2 is even," you write "assume x is even." You are not claiming that x is actually even in some absolute sense. You are setting up a conditional reasoning scope in which x has that property, to derive what follows. The "assume" in the inductive step works the same way, only with k as a free natural number.

A formal re-statement of the inductive step

To make the structure airtight, you can rephrase the inductive step in a way that removes the ambiguity:

Inductive step (formal). Let k be an arbitrary natural number. Prove: if P(k) holds, then P(k+1) holds.

"Let k be an arbitrary natural number" — no claim that any specific P-value is true. "If P(k) holds" — a conditional entered temporarily. "Then P(k+1) holds" — the conclusion of the conditional.

The thing being proved is universally quantified over k: for every k, the conditional P(k) \Rightarrow P(k+1) is true. That universal quantification is the crucial piece. You do not prove the conditional for one specific k (that would be a weak statement); you prove it as a uniform rule that works for all k at once.

A working analogy: the payback rule

Imagine a club where the rules say:

Rule 1 (base). Every member who joined in 2025 starts with one free coffee.
Rule 2 (step). Every time a member has a free coffee, they must share it with the person who joined right after them.

Rule 2 is conditional. It does not claim anyone currently has a free coffee. It says what must happen if someone does. Combine Rule 2 with Rule 1 (Rule 1 guarantees the first member has a free coffee), and you conclude that every later member eventually gets a coffee.

The rule by itself says nothing about coffees being actually consumed. Only when you apply it repeatedly, starting from the base case, do coffees flow. Induction works the same way: the inductive step is a rule, not a fact about specific n, and the rule starts producing truths only when the base case fires it up.

The one-line answer

The inductive step proves a conditional, not the conclusion. You assume P(k) temporarily as the hypothesis of a conditional; you derive P(k+1); you therefore establish "P(k) \Rightarrow P(k+1)" as a rule. That rule, combined with the unconditional base case P(1), generates P(n) for every n by iterated modus ponens. You never assumed P(n) anywhere without already having derived it from earlier, independently-proved facts. The "circularity" is an illusion created by the word assume, which in mathematical proof means "hypothesise locally" — not "take as given globally."