Q
Question
Sharpen the inquiry — and see why the answer to a simple question changes everything

Article A1 ended with a question: knowing that Socrates is mortal because all humans are mortal — are you as certain of that as you are certain that the next stone you drop will fall? Both feel like solid, obvious conclusions. But are they produced by the same kind of reasoning?

The answer, it turns out, is no. And the difference between them is one of the most consequential distinctions in the history of human thought. It separates mathematics from science, proof from evidence, certainty from probability. It is the difference that David Hume thought undermined the rational foundations of everything we believe about the physical world — a conclusion so disturbing that Kant said it woke him from a philosophical sleep that had lasted decades.

Consider the two pieces of reasoning carefully:

Reasoning A: All humans are mortal. Socrates is a human. Therefore Socrates is mortal.

Reasoning B: Every stone that has ever been dropped has fallen. Therefore, the next stone I drop will fall.

Both feel compelling. But ask yourself: in Reasoning A, could the conclusion be false if both premises are true? No — it is impossible. If it really is true that all humans are mortal, and Socrates really is human, then Socrates simply cannot be immortal. The conclusion is guaranteed.

Now ask the same question about Reasoning B: could the conclusion be false even if the premise is true? Yes — in principle. No matter how many stones have fallen, there is no logical impossibility in the next one floating upward. The premise makes the conclusion very likely. It does not make it certain.

What is the difference between reasoning that guarantees its conclusion and reasoning that merely makes it probable — and what are the consequences of that difference?

This is the question this article answers. The two kinds of reasoning have names: the first is deductive, the second is inductive. Almost every significant argument you will encounter in philosophy, in science, in law, and in ordinary life belongs to one of these two categories. Learning to tell them apart — and understanding what each can and cannot do — is the foundation of serious critical thinking.

U
Unpack
Define deductive and inductive reasoning precisely, and map the crucial differences between them

Deductive reasoning: the guarantee

A deductive argument is one in which the truth of the premises is supposed to guarantee the truth of the conclusion. If a deductive argument is well-constructed, it is impossible for the premises to be true and the conclusion to be false. Logicians call such an argument valid (we explore this term in detail in Article A3).

The key phrase is "if the premises are true." A deductive argument does not promise that its premises are actually true — only that if they are, the conclusion follows necessarily. Consider:

Deductive Argument — Standard Syllogism Deductive
P1All politicians are corrupt.
P2This candidate is a politician.
∴ CTherefore, this candidate is corrupt.
This argument is deductively valid: if P1 and P2 were true, C would have to be true. But P1 is almost certainly false — not all politicians are corrupt. A valid argument can have a false conclusion, as long as that falsity traces back to a false premise. Validity is about the logical relationship between premises and conclusion, not about whether the premises are actually true. That question — are the premises true? — determines whether the argument is sound, which is the subject of Article A3.

There is something philosophically remarkable about deduction: the conclusion is always already contained within the premises. The argument from Socrates' mortality does not tell you anything about the world that was not already implicit in what you knew. If you already accepted that all humans are mortal and that Socrates is human, you were already committed to Socrates' mortality — you just might not have noticed. Deduction makes explicit what was already implicit.

This is why deduction is the primary method of mathematics. A mathematical proof does not discover new facts about the world in the way a scientific experiment does; it unfolds the logical consequences of what was already assumed in the axioms. The Pythagorean theorem was always true, given Euclidean geometry's axioms. The proof revealed it.

Inductive reasoning: the leap

An inductive argument is one in which the premises are meant to make the conclusion probable or likely, but do not guarantee it. No matter how strong an inductive argument, the conclusion always goes beyond what the premises strictly establish. There is always, in principle, a gap.

Inductive arguments typically move from the specific to the general — from a collection of observed instances to a conclusion about all instances, including unobserved ones:

Inductive Argument — Generalisation from Experience Inductive
P1Every emerald observed so far has been green.
P2A very large and varied sample of emeralds has been observed across many centuries and continents.
∴ CTherefore, all emeralds are (probably) green.
The word "probably" does real philosophical work here. No matter how large the sample in P2, the conclusion cannot be proven — only supported. A single non-green emerald would refute it. The premises give us very good reason to believe the conclusion, but they cannot guarantee it. Notice that the argument goes beyond its evidence: P1 and P2 are about observed emeralds; C is about all emeralds, including the unobserved ones.

Induction is the lifeblood of empirical inquiry. Every time a scientist draws a general conclusion from experimental data, they are reasoning inductively. Every time a doctor diagnoses a patient based on observed symptoms, they are reasoning inductively. Every time you assume the chair will hold your weight because it always has before, you are reasoning inductively. We cannot function without it. And yet, as Hume showed, it cannot be rationally justified in any non-circular way. This is the problem of induction, and it is one of the deepest unsolved problems in the philosophy of knowledge.

The fundamental asymmetry — a comparison

Dimension
Deductive Reasoning
Inductive Reasoning
If premises are true…
Conclusion must be true. Truth is preserved by the argument's logical form.
Conclusion is probably, but not certainly, true. The gap can never be closed.
Direction of reasoning
From general principles down to particular cases. Top-down.
From particular observations up to general conclusions. Bottom-up.
Effect of one counter-example
None — a valid deductive argument cannot be refuted by a counter-example to the conclusion (only by challenging the premises).
Decisive. A single counter-example can destroy a universal inductive conclusion.
What new information gives you
No new information about the world — only clarification of what was already implicit in the premises.
New information about the world — the conclusion extends beyond what was in the premises.
Evaluated as…
Valid or invalid; sound or unsound (A3).
Strong or weak; cogent or uncogent (A3).
Primary use
Mathematics, formal logic, philosophical argument analysis, legal reasoning from established rules.
Empirical science, medicine, history, everyday reasoning about the physical world.

A third kind? Abduction

Before leaving this conceptual map, it is worth noting that some philosophers recognise a third form of reasoning: abduction, or inference to the best explanation. Where deduction derives conclusions necessarily and induction generalises from observation, abduction asks: given these observations, what hypothesis would best explain them?

A doctor who observes symptoms and concludes that a patient has a particular illness is reasoning abductively — not merely inducing a general pattern, but hypothesising the best explanation for a specific set of facts. Sherlock Holmes's famous "deductions" are actually almost always abductions. When Holmes concludes from a man's tan and posture that he has recently returned from Afghanistan, he is not deducing (no premise guarantees the conclusion) or inducing (he has observed only one case). He is inferring the best explanation of the available evidence.

Abduction is a genuinely important form of reasoning, and we will return to it in Package I (Philosophy of Science) in the context of scientific explanation. For now, the essential distinction to master is the deductive/inductive one.

E
Examine
Investigate the philosophical consequences of the deductive/inductive distinction — and encounter Hume's devastating challenge

Francis Bacon and the inductive method

Before examining the problem with induction, it is worth understanding why induction came to seem so important in the first place. In the early seventeenth century, the English philosopher Francis Bacon argued that philosophy and natural inquiry had been dominated for too long by the deductive method — by reasoning from general principles laid down by authority rather than from careful observation of the world.

FB
Philosopher
Francis Bacon
1561–1626
Scientific method  ·  Inductive logic  ·  Philosophy of science
Bacon's major philosophical work, Novum Organum (1620), was a direct attack on Aristotelian deductive logic as the primary method of inquiry. Bacon argued that knowledge of the natural world must be grounded in systematic observation and experiment — not in reasoning downward from first principles established by ancient authorities. He proposed a methodical inductive approach: gather observations carefully, look for patterns, form tentative generalisations, and test them against further observation. This is recognisably the beginning of the scientific method as we now understand it. Bacon's insight was that deduction can only give you back what was already implicit in your premises; if you want genuinely new knowledge of the world, you need to look at the world first.
Relevant work: Novum Organum (1620) — "The New Organon." The title is a pointed response to Aristotle's Organon: Bacon is proposing a new instrument of inquiry to replace the old one.

Bacon's intervention was historically decisive. The seventeenth century saw the birth of the experimental sciences, and inductive reasoning — reasoning from careful observation to general conclusions — became the method of empirical inquiry. Physics, chemistry, biology, medicine: all depend on it. The problem, which Bacon did not fully confront, is that Hume identified over a century later: what exactly justifies the inductive leap?

Hume and the problem of induction

David Hume, writing in the eighteenth century, posed what is still considered the most important philosophical problem about induction. His question was devastatingly simple: what reason do we have for believing that the future will resemble the past?

We use inductive reasoning constantly because we assume that patterns observed in the past will continue into the future: the sun has risen every day in recorded history, so it will rise tomorrow; water has always boiled at 100°C at sea level, so it will again; medicines that reduced pain in clinical trials will reduce pain in future patients. But what justifies this assumption?

There are only two possible types of justification, and Hume argued that neither works:

Option 1: Logical justification. We might try to argue that induction is logically guaranteed — that it is self-evidently rational to expect the future to resemble the past. But this fails immediately: we can perfectly well imagine a world in which the future does not resemble the past. There is no logical contradiction in the sun failing to rise tomorrow. Induction is not deductively valid, and no amount of past instances logically entails a future one.

Option 2: Inductive justification. We might try to argue that induction is justified because it has worked reliably in the past — we have used it before and it has served us well, so we are inductively justified in trusting it. But this is circular: we are using inductive reasoning to justify inductive reasoning. It assumes exactly what it is trying to prove.

Hume's conclusion was stark: our reliance on inductive reasoning is a matter of custom and habit, not of rational justification. We expect the future to resemble the past because that is how our minds are built, not because we have good logical grounds for it. This is not a small claim. It means that the entire edifice of empirical science — everything we think we know about the physical world through observation and experiment — rests on a foundation that cannot be rationally justified in any non-circular way.

DH
Philosopher
David Hume
1711–1776
Epistemology  ·  Causation  ·  Philosophy of science  ·  Ethics
Hume's argument about induction appears in both his Treatise of Human Nature (1739–40) and the more accessible Enquiry Concerning Human Understanding (1748). His broader epistemological project was to apply to the human mind the same rigorous observational method that Newton had applied to the physical world — with results that proved far more unsettling. Hume concluded that our beliefs about causation, about the external world, and about our own continuous selves are not rationally grounded at all: they are products of habit, custom, and the imagination. The problem of induction is the most famous consequence of this programme, but it is one thread in a much larger and more radical picture.
Relevant works: An Enquiry Concerning Human Understanding (1748), Section IV — "Sceptical Doubts Concerning the Operations of the Understanding." This is the most accessible statement of the problem. The Treatise (Book I, Part III) contains the more extended treatment.

The black swan: one counter-example and a universal claim collapses

Hume's problem is abstract. The black swan story is its most famous concrete illustration — and it has a particular resonance for Australian students.

🦢
Historical context
The Black Swan — Western Australia, 1697
For centuries, European naturalists had observed thousands of swans across Europe and the known world. Every single one was white. The inductive conclusion seemed overwhelmingly supported: "All swans are white." It had become, in effect, a working assumption of natural history — so well supported by observation that no one seriously questioned it. Then, in 1697, the Dutch explorer Willem de Vlamingh sailed into what is now the Swan River in Western Australia and observed large black waterfowl that were, unmistakably, swans. A single observation — in a land that Europeans had not previously explored — destroyed a universal inductive conclusion built on thousands of confirming instances. The "black swan" has since become the standard philosophical example for the fragility of inductive generalisations, and was given renewed prominence by Nassim Nicholas Taleb's 2007 book of the same name, which applied the concept to unpredictable high-impact events in economics and history.

The black swan story demonstrates a crucial asymmetry in inductive reasoning: while no number of confirming instances can prove a universal inductive generalisation, a single disconfirming instance can refute one. You cannot prove "all swans are white" by observing white swans — no matter how many you find. But one black swan disproves it definitively.

This asymmetry was central to Karl Popper's philosophy of science (which we examine in Package I). Popper's response to Hume was to argue that good scientific reasoning should be structured so that it can be refuted — falsified — by a single counter-example. Science progresses not by accumulating confirmations, but by eliminating false hypotheses through falsification. The method Popper had in mind is actually deductive: "If my theory is true, then X will happen under these conditions. X did not happen. Therefore my theory is not true." This is a logically valid deductive argument. Popper's hope was to ground scientific reasoning in deduction rather than induction — but whether he succeeded is deeply contested.

S
Synthesise
Apply the distinction — practise classifying arguments as deductive or inductive, and understand what is at stake in each case

The test for distinguishing deductive from inductive arguments is this: ask yourself whether, if the premises are all true, the conclusion must be true — or whether it is merely probably true. If the conclusion follows necessarily from the premises, you have a deductive argument. If the premises make the conclusion likely but do not guarantee it, you have an inductive argument.

In practice, two additional clues help: the language of the conclusion (words like "probably," "likely," "in most cases," and "tends to" signal induction; absolute claims with no hedging signal deduction), and the direction of reasoning (deduction moves from general principles to specific cases; induction typically moves from specific observations to general claims).

Apply these tests to the arguments below before reading the verdicts.

Argument Classification — Worked Examples
1
"If an action is right, then it maximises overall happiness. Donating to effective charities maximises overall happiness. Therefore, donating to effective charities is right."
Deductive
The conclusion follows necessarily from the premises. If we grant both premises (which we may well dispute — this is the utilitarian position), we cannot coherently deny the conclusion. This is the structure of a modus ponens argument: "If P then Q; Q; therefore P" — though note the form here inverts slightly. The argument's force is purely logical.
2
"Every philosophy student I've encountered finds Hume's Treatise initially baffling. Therefore Hume's Treatise is probably a difficult text for beginners."
Inductive
This is a generalisation from a limited sample to a broader conclusion. The premise makes the conclusion probable but does not guarantee it. Perhaps every philosophy student encountered was simply unusually underprepared. The word "probably" in the conclusion is itself a signal: the speaker knows the argument does not guarantee what it claims. This is also a relatively weak inductive argument — the sample is too small and too personal to support a strong generalisation.
3
"No action is both obligatory and prohibited. Euthanasia is prohibited by this hospital's policy. Therefore, euthanasia is not obligatory under this hospital's policy."
Deductive
The first premise is a definitional or logical claim: a single action cannot be both required and forbidden by the same normative system. Given the two premises, the conclusion follows with logical necessity. This is exactly the kind of argument used in applied ethics and legal reasoning — taking a general principle (P1) and applying it to a specific case (P2) to derive a conclusion (C).
4
"Penicillin successfully treated bacterial infections in 93% of the thousands of cases studied in clinical trials across multiple countries and patient populations. Therefore, penicillin will probably be effective at treating this patient's bacterial infection."
Inductive
This is a statistical inductive argument — one of the strongest kinds of inductive reasoning available. The large, varied sample makes the conclusion highly probable. Note, however, that even a 93% success rate means the argument gives us genuine grounds for uncertainty: 7% of patients in the trials were not successfully treated. The word "probably" again does real work — it acknowledges that the conclusion is not guaranteed even by very strong evidence. Medical diagnosis and prognosis almost always involve this kind of strong-but-not-certain inductive reasoning.
5
"Triangles have three interior angles that sum to 180°. This shape is a triangle. Therefore, this shape's interior angles sum to 180°."
Deductive
A purely mathematical deductive argument. The first premise is a theorem of Euclidean geometry — it follows necessarily from Euclid's axioms. Given that it is a triangle (P2), the angle-sum property follows necessarily. Mathematical reasoning is paradigmatically deductive: mathematicians do not say "in all the triangles we've checked so far, the angles summed to 180°, so probably they always do." They prove it — which is to say, they give a deductive argument from axioms to theorem.

These examples reveal something important: the type of argument is often not obvious from the surface features of the prose. Whether an argument is deductive or inductive depends on the logical relationship it claims between premises and conclusion — not on its subject matter, its length, or how confident the author sounds. A very confident inductive argument is still inductive. A logically valid deductive argument with absurd premises is still deductive.

This also explains why the deductive/inductive distinction matters for argument evaluation. Accusing an inductive argument of failing to guarantee its conclusion is not a good objection — it was never trying to. The right criticism of a weak inductive argument is that its premises do not make the conclusion sufficiently probable, or that the sample is too small or biased. The right criticism of an invalid deductive argument is that the conclusion does not follow from the premises — that it is possible to accept the premises and still coherently deny the conclusion. You need to know which kind of argument you are evaluating before you can evaluate it fairly.

T
Transfer
See how the deductive/inductive distinction shapes mathematics, science, and our understanding of knowledge itself

Why mathematics can prove, and science can only show

One of the most striking consequences of the deductive/inductive distinction is what it tells us about the different epistemic statuses of mathematical and scientific knowledge — that is, about what kind of certainty each kind of knowledge can achieve.

Mathematical knowledge is deductive. When a mathematician proves a theorem, they construct a valid deductive argument from axioms — premises accepted as the foundation of the system — to the theorem as conclusion. Once proven, a mathematical theorem is certain in the strongest sense possible: given the axioms, the theorem must be true. This is why 2 + 2 = 4 is not a matter of opinion, and why no amount of new empirical evidence could ever refute it. It is not an empirical claim about the world; it is a logical consequence of the definitions and axioms of arithmetic.

Scientific knowledge is primarily inductive. A scientific law — Newton's law of gravitation, the germ theory of disease, the theory of evolution by natural selection — is an inductive generalisation from observed evidence. This means scientific knowledge is always, in principle, provisional. No amount of confirming evidence makes a scientific theory certain. The black swan can always appear. Einstein's general relativity did not refute Newton's gravity because Newton was lying; it revealed that Newton's theory was an excellent inductive approximation for most conditions, which broke down at relativistic speeds and scales.

This is not a weakness of science — it is what makes science progressive. A theory that can be overturned by new evidence is one that is genuinely attempting to describe the world, rather than merely elaborating logical consequences of its own assumptions. But it does mean that when you say "science has proven X," you are being imprecise in a philosophically important way. Science has provided very strong inductive evidence for X. The difference matters.

The Theory of Knowledge connection

For students studying IB Theory of Knowledge, the deductive/inductive distinction maps directly onto several of TOK's central questions. How do different Areas of Knowledge establish their claims? What kinds of certainty are available in mathematics versus the natural sciences? What is the relationship between evidence and knowledge?

TOK asks students to reflect on the "methods" of different disciplines. The core insight this article provides is that those methods differ in their logical structure — and that difference determines what kind of certainty each discipline can offer. Mathematics proves its conclusions deductively and achieves a kind of necessity that science, for all its power, cannot match. Science reasons inductively from evidence and achieves a kind of breadth and applicability that pure mathematics, for all its rigour, cannot reach on its own. Both are genuine forms of knowledge; they are simply different kinds.

Connecting to the rest of this package

In Article A3, you will take the concepts from A1 and A2 and combine them into a complete evaluative toolkit. An argument is not simply deductive or inductive — it is also valid or invalid (if deductive), or strong or weak (if inductive), and sound or unsound (depending on whether the premises are actually true). These are the terms that allow you to say not just what kind of argument you have, but whether it is a good one. That is the final step in equipping yourself to evaluate any philosophical argument you encounter.

And beyond this package, the problem of induction — Hume's challenge — will surface again in Package B (Epistemology), where it connects to broader questions about the nature and limits of human knowledge. Hume's conclusion that inductive reasoning cannot be rationally justified is not just a result about logic; it is one of the most radical claims in the history of epistemology. When you arrive at Package B and ask "What can we know?", you will find that Hume's shadow has arrived before you.

The question to carry with you into Article A3
You now know that deductive arguments can be valid, and that inductive arguments can be strong. But "valid" and "strong" are not the same as "good." A deductive argument can be valid and still have a false conclusion. An inductive argument can be very strong and still be wrong. What, then, makes an argument not just valid or strong — but genuinely good?
That question is answered by the concepts of soundness and cogency. Article A3 will complete your evaluative toolkit — giving you precise language for every judgment you need to make about an argument's quality.