The Problem of Induction: Why We Cannot Prove the Sun Will Rise

The Sun Has Risen Every Day for 4.5 Billion Years

That is roughly 1.6 trillion sunrises. Every single one confirmed the pattern. Yet no number of past observations can logically guarantee the next one. This is the problem of induction, and it has haunted philosophy and science since David Hume articulated it in 1739. The problem is deceptively simple. Its consequences run deep.

Hume's Original Argument

In A Treatise of Human Nature, Hume distinguished two types of knowledge. Relations of ideas — like mathematics — are certain because they are tautological. Matters of fact — like whether the sun will rise — depend on experience. And experience, Hume argued, can never justify conclusions about the future.

The reasoning proceeds in three steps:

All knowledge of cause and effect comes from experience
Experience only tells us what has happened, not what will happen
Any argument that the future will resemble the past assumes the very principle it tries to prove (that nature is uniform)

The argument is circular. To justify induction, you must use induction. Hume did not deny that induction works in practice. He denied that it can be justified by reason alone. This distinction matters enormously.

The Uniformity of Nature Assumption

Every inductive inference assumes that unobserved cases will resemble observed ones. Scientists assume physical laws do not change overnight. But this assumption cannot itself be derived from observation without begging the question. Hume called this inference a matter of "custom or habit" — psychologically irresistible but logically unsupported.

Why It Matters for Science

Science relies on induction at every level. A chemist performs an experiment 50 times and generalizes. A physicist observes that gravity behaves consistently and writes a universal law. An epidemiologist sees a correlation and infers causation. Every scientific prediction is an inductive leap.

Scientific Activity	Inductive Element	Hume's Challenge
Empirical generalization	"All observed X have property Y, therefore all X have Y"	Unobserved X may lack Y
Causal inference	"A has always preceded B, so A causes B"	Correlation does not establish necessary connection
Prediction	"This law held yesterday, so it holds tomorrow"	Past regularity cannot guarantee future regularity
Theory confirmation	"Evidence E supports theory T"	Infinitely many theories are compatible with E

If Hume is right, no scientific theory is ever proven. Every law of physics is, strictly speaking, a well-supported guess. This does not make science useless — it makes certainty unavailable.

Attempted Solutions Across Three Centuries

Philosophers have tried to solve or dissolve the problem of induction for nearly 300 years. None has achieved consensus.

Kant's Synthetic A Priori

Immanuel Kant argued that causation is not learned from experience but imposed on experience by the structure of the human mind. We cannot perceive the world except through causal categories. This sidesteps Hume's argument by relocating causation from the external world to the knowing subject. Critics respond that Kant's categories might be cognitive illusions rather than truths about reality.

Popper's Falsificationism

Karl Popper accepted Hume's conclusion and built a philosophy of science around it. Science does not proceed by induction, Popper argued. It proceeds by conjecture and refutation. Scientists propose bold hypotheses and try to falsify them. Theories that survive rigorous testing are "corroborated" but never confirmed.

Approach	Philosopher	Core Claim	Main Criticism
Skepticism	Hume	Induction has no rational justification	Leads to radical skepticism about all empirical knowledge
Transcendental idealism	Kant	Causation is a necessary condition of experience	May confuse how we think with how the world is
Falsificationism	Popper	Science uses deduction, not induction	Falsification itself relies on inductive assumptions
Bayesianism	Various	Probability theory formalizes rational belief updating	Requires prior probabilities that may be subjective
Pragmatism	Peirce, Reichenbach	Induction is justified because it works better than alternatives	"It works" is itself an inductive claim

The Bayesian Response

Bayesian epistemologists use probability theory to model how evidence should update beliefs. Starting with a prior probability for a hypothesis, each confirming observation raises the probability via Bayes' theorem. The approach does not claim certainty. It claims rationality — there is a mathematically optimal way to adjust confidence in light of evidence.

Critics note that Bayesianism requires choosing initial priors, and different priors yield different conclusions. The problem of induction resurfaces as the problem of priors.

Goodman's New Riddle: The Problem Gets Worse

In 1955, Nelson Goodman deepened the problem. He invented the predicate "grue" — an object is grue if it is observed before a certain date and is green, or is not observed before that date and is blue. All observed emeralds are both green and grue. Why do we project "green" into the future and not "grue"?

Both predicates are equally supported by all available evidence
Logic alone cannot distinguish between them
Our preference for "green" seems to rest on habit, not reason
Goodman called this the "new riddle of induction" — not just whether induction is justified, but which inductive inferences to make

Goodman's riddle shows that even if we accept induction, we face a further problem of selecting among infinitely many possible generalizations, all equally supported by existing data.

Living With Uncertainty

Working scientists rarely lose sleep over the problem of induction. The laws of physics have held reliably across every observation ever made. Engineering works. Medicine improves. Bridges stay up. The practical success of induction is overwhelming.

But Hume's point stands as a logical matter. No amount of practical success constitutes a logical proof. The gap between "has always worked" and "must always work" is precisely the gap that the problem of induction identifies — and after nearly three centuries, it remains open.