Skip to content

The idea this whole site is about

The anthropic principle

Any universe we can observe must be one that permits observers. That sentence looks empty and is not — it has made two correct predictions, and it fails in ways worth knowing about.

In one sentence

anthropic principle/anˈθrɒpɪk ˈprɪnsɪpəl/

The anthropic principle is the observation that what we can measure about the universe is constrained by the fact that observers exist to measure it — so a cosmos that permits life is the only kind anyone could ever find themselves in.

Etymology
Anthropic comes from the Ancient Greek ἄνθρωπος (anthrōpos), meaning human being — the same root as anthropology and philanthropy. It is not from anthropos meaning man in the male sense, and it does not imply the universe was made for us; the word marks who is doing the observing, not who the universe is for.
Who named it
The Australian physicist Brandon Carter named it in 1973, at a conference in Kraków marking the 500th anniversary of Copernicus's birth.

Weak, strong, and the one nobody defends

Three claims travel under one name, and most arguments about the anthropic principle are really arguments about which of them is meant. They are not equally serious.

Weak (WAP)

Our location in space and time is not random: it is constrained to a place and epoch where observers can exist.

Standing

Uncontroversial, and genuinely useful.

This is a statement about sampling, not about physics. Almost every place in the universe — deep space, stellar interiors, the first three minutes — contains no observers, so the set of vantage points we could possibly be looking from is a heavily filtered sample of the vantage points there are. Denying it means claiming you could have measured a cosmos with nobody in it.

Strong (SAP)

The universe must be such that observers arise within it at some stage of its history.

Standing

Genuinely disputed, partly because it has two readings.

Read as a claim about laws, it says something in physics compels life — which would be a substantive and currently unsupported discovery. Read as a claim about a multiverse, it says any universe containing us is necessarily one of the habitable ones, which is the weak version with the modality inflated. Carter intended something closer to the second; the first is what critics usually attack.

Final (FAP)

Intelligent information-processing must come into existence, and once it does, it will never die out.

Standing

Almost nobody defends it.

Added by John Barrow and Frank Tipler in 1986. Martin Gardner's review suggested the acronym CRAP — the Completely Ridiculous Anthropic Principle — and the label stuck harder than the idea. It is worth naming only because it is often what people have in mind when they dismiss the whole family, and the weak version deserves better company than that.

Twelve ways to come at it

The same observation reads differently depending on what you want from it. Each of these is a different discipline asking a different question of one sentence.

Cosmology

The principle in full

Carter's 1973 formulation, Hoyle's carbon resonance, Weinberg's bound on the cosmological constant, and the Boltzmann-brain problem that the reasoning creates for itself.

Read this

Philosophy of science

Is it science at all?

Whether a principle that cannot be tested against a universe we did not observe is doing explanatory work, or restating the conditions of the observation.

Read this

Probability

Self-sampling versus self-indication

Two rules for reasoning about your own position in a reference class. They disagree, they cannot both be right, and the Doomsday Argument turns on which you pick.

Read this

Selection effects

What observers filter out

The general shape of the move, from Wald's bombers to the universe: the sample you can see has already been filtered by the process that lets you see it.

Read this

The evidence

Is the universe actually fine-tuned?

Which constants are claimed to be narrow, how narrow, and how much of that narrowness survives being asked 'narrow relative to what?'

Read this

Theology

Fine-tuning and the design argument

The claim the anthropic principle is constantly confused with. Where the two actually meet, why the multiverse — not the principle — is design's real rival, and the objection that damages both.

Read this

The case against

Why many physicists refuse to use it

Sober's emptiness charge, the missing measure, Stenger on joint variation, curiosity-stopping, and the Boltzmann-brain reductio — each at full strength.

Read this

The answers

Four responses to fine-tuning

Brute fact, necessity, multiverse, design — each at its strongest and each with its price. The anthropic principle is not itself one of the four; it changes how much any of them has to explain.

Read this

The multiverse it needs

How many multiverses?

Anthropic selection needs an ensemble to select from. Tegmark's four levels are four very different proposals for what that ensemble could be.

Read this

The numbers

The six numbers

Martin Rees's shortlist of the constants that decide whether a universe can build anything. This is what 'fine-tuned' is about, concretely.

Read this

Language and culture

Why an AI company took the name

One Greek word, three owners: antiquity, cosmology in 1973, and a San Francisco laboratory in 2021. What each of them meant by it.

Read this

In your hands

Break a universe and watch

The argument is hard to feel in the abstract. Move one constant, watch the cosmos fail, and notice that nobody is left standing in it to be surprised.

Read this

Fifty years in five moments

The idea is older than the name, and the name was a joke about Copernicus.

  1. 1957

    Dicke answers Dirac

    Paul Dirac had noticed that several enormous dimensionless numbers in physics are suspiciously close to each other, and proposed that the gravitational constant must be changing over time to keep them so. Robert Dicke pointed out that the coincidence holds only during the era when stars burn — which is the only era in which physicists exist. No new physics required. This is the anthropic move, sixteen years before it had a name.

  2. 1973

    Carter names it, at Copernicus's party

    Brandon Carter introduced the term at a Kraków symposium for the 500th anniversary of Copernicus's birth. The venue was the argument. The Copernican principle says our position is not special; Carter's point was that it cannot be entirely typical either, because typical places in the universe are lethal. He was proposing a correction to Copernicus at Copernicus's own birthday.

  3. 1986

    Barrow and Tipler write the book

    The Anthropic Cosmological Principle ran to over 700 pages, collected the fine-tuning coincidences into one place, and introduced the final anthropic principle that damaged the reputation of everything around it. The catalogue of coincidences is still the standard reference; the closing chapters are not.

  4. 1987

    Weinberg bets on it

    Steven Weinberg asked not why the cosmological constant is small but how large it could be while still allowing galaxies — and therefore observers. The bound he calculated said the value should be small but not zero, at a time when most physicists expected exactly zero. Eleven years later, two supernova surveys found a small positive value.

  5. 2002

    Bostrom makes it a rulebook

    Nick Bostrom's Anthropic Bias moved the argument from physics into formal epistemology: what exactly is the reference class of observers, and what probability rule applies to your own position inside it? The Self-Sampling and Self-Indication Assumptions come from this line of work, and so does most of the modern disagreement.

The two times it paid

Anthropic reasoning earns its keep when it makes a risky prediction and not otherwise. It has done so twice, and both times the experiment could have come back empty.

Hoyle's carbon resonance, 1953

Stars build carbon by fusing three helium nuclei, and the reaction is far too slow to produce the carbon we are made of — unless carbon-12 has an excited state at around 7.65 MeV that lets the third nucleus arrive in time. No such state was known. Fred Hoyle reasoned backwards from his own existence, told William Fowler's group at Caltech where to look, and they found it close to where he said.

What would have killed it

The measurement finding no state near 7.65 MeV. Hoyle predicted a specific unobserved nuclear energy level at a stated value; that is a number an experiment can refuse to produce.

Weinberg's bound, 1987

Quantum field theory's natural scale for the energy of empty space exceeds the observed value by roughly 120 orders of magnitude — arguably the worst quantitative failure in physics. Weinberg asked how large the cosmological constant could be before expansion outran the formation of galaxies. The answer was small, but distinctly not zero, and it put the likeliest value near the top of the habitable range.

What would have killed it

A measured cosmological constant of exactly zero, which is what most of the field expected at the time and what a new symmetry would have delivered.

The objections, at their strongest

These are not fringe complaints. The first is the reason many working physicists will not use the reasoning at all.

It is empty

"We could only observe a life-permitting universe" is true, and from a true statement about what we could observe, nothing follows about what is likely. Elliott Sober put it sharply: the observation is guaranteed by our existence, so it cannot discriminate between hypotheses that both allow our existence. A fact you were certain to find tells you nothing when you find it.

The reply

Correct as stated, and it is why the principle is worthless as a general-purpose explanation. It is not correct about the two cases above, where the reasoning generated a number that had not been measured yet. The honest position is that anthropic arguments are only as good as the risky prediction they produce.

There is no probability distribution

To call a constant improbable you need a measure over the values it might have taken, and nobody has one. Several standard fine-tuning figures quietly assume a uniform distribution over a range chosen after the fact. Change the range or reparameterise the constant and the improbability moves.

The reply

This objection stands, and it is the best reason to distrust any specific fine-tuning number. What survives it is the qualitative claim: these are not cases where a small change produces a slightly different universe, but cases where it produces no observers of any kind.

It stops inquiry

Every time physics has found a real mechanism — inflation explaining flatness, symmetry breaking explaining certain asymmetries — the anthropic share of the puzzle shrank. The worry is not that the reasoning is false but that it is available too early, and most tempting exactly when a harder answer is still out there.

The reply

A fair warning rather than a refutation. It argues for treating anthropic explanation as a last resort, which is roughly how Weinberg himself framed it, and not for pretending the selection effect is not real.

It implies you are a Boltzmann brain

In many cosmological models, observers assembled by random thermal fluctuation vastly outnumber observers produced the ordinary way. Apply anthropic reasoning consistently — assume you are typical — and you should conclude you are almost certainly a momentary fluctuation with false memories.

The reply

Nobody believes that, and the usual move is to treat it as a constraint: a model predicting mostly Boltzmann brains is taken to be ruled out on those grounds. It is also a standing warning about how much work the word 'typical' is doing.

It is the design argument in disguise

Fine-tuning talk and design talk arrive together so often that the two get treated as one position, with the anthropic principle as the polite version.

The reply

They are different claims running in opposite directions. The design argument infers a tuner from apparent tuning. The anthropic principle infers a constraint on what we could have observed, and is neutral about what set the constants. It is most often deployed against the design argument, which is a use of the principle, not the principle itself.

See it happen

The abstract version is hard to feel. The concrete version is a slider.

The Universe Tuner

Nine constants, one cosmos. Move any of them and watch which door closes first — no stars, no chemistry, no time.

Can you reason like an observer?

Eight questions on selection effects, typicality, reference classes, and what anthropic arguments can actually prove.

You shouldn't exist

The same filtering argument pointed at you instead of at the cosmos: stack every improbability behind your own existence into one number.

The one-paragraph version

The glossary entry, for when you want the definition and not the argument.

Common questions

Short answers. Each links to the longer one.

The anthropic principle is named for the word anthropos, which is the Greek word for what?
Human being. Anthropic derives from the Ancient Greek ἄνθρωπος (anthrōpos), meaning human or human being — the same root as anthropology, anthropomorphic and philanthropy. Brandon Carter chose it in 1973 to mark that the observer is part of the observation, not to suggest the universe was built for people.
What does the anthropic principle actually say?
That the properties we can observe in the universe are constrained by the requirement that observers exist to observe them. We could never have measured a cosmos incapable of producing us, because there would be no one present to take the measurement.
What is the difference between the weak and strong anthropic principle?
The weak version says our position in space and time is constrained to somewhere observers can exist — a claim about sampling. The strong version says the universe must be such that observers arise in it at some point — a claim about the laws themselves, or about an ensemble of universes. The weak version is almost universally accepted; the strong version is not.
What are the implications of the weak anthropic principle?
Mainly that some apparent coincidences need no explanation beyond the filter. Dicke's 1957 answer to Dirac is the model case: a numerical coincidence that holds only during the stellar era needs no changing constants, because the stellar era is the only time anyone is around to notice it. It also means any cosmological statistic drawn from our own position has to be corrected for the fact that we are in it.
Does the anthropic principle explain fine-tuning?
No. It explains why we observe a life-permitting universe rather than a barren one, given that both kinds exist. It does not explain why any life-permitting universe exists, and it does not by itself establish that other universes do. Selection can only select from what is already there.
Who came up with the anthropic principle?
Brandon Carter named and formalised it in 1973 at a Kraków conference for the 500th anniversary of Copernicus's birth. The reasoning predates the name: Robert Dicke used it in 1957, and Alfred Russel Wallace made a recognisably similar argument in 1903.
Is the anthropic principle scientific?
Contested. It has produced two risky, numerical, subsequently confirmed predictions — Hoyle's carbon resonance and Weinberg's bound on the cosmological constant. Used that way it behaves like ordinary science with an unusual premise. Used as a general explanation for anything surprising, it predicts nothing and the charge of emptiness lands.
Is the anthropic principle the same as the argument from design?
No, and they run in opposite directions. The design argument infers a designer from apparent tuning. The anthropic principle infers a constraint on what we could have observed and stays neutral about what set the constants. It is most often used as a reply to the design argument.
Does the anthropic principle prove the multiverse?
No, but the strongest anthropic arguments require one. A selection effect needs a pool to select from: if the constants could only ever have taken one set of values, there is nothing to select and no puzzle. Weinberg's bound is a genuine prediction only if an ensemble of values exists.
What is an example of the anthropic principle?
Fred Hoyle's 1953 prediction. Carbon exists, carbon can only form in stars through a three-body reaction that is far too slow, therefore carbon-12 must have a specific excited state nobody had found. He gave the energy, Caltech looked, and it was there.
What is the observation selection effect?
The general form of the same move: any dataset is filtered by the process that produced it. Abraham Wald's bombers are the standard illustration — armour the places where returning planes were not hit, because the planes hit there did not return to be measured. The anthropic principle applies this at the scale of the universe, with existence as the filter.
Why is the AI company called Anthropic?
It took the word, not the principle. The company was founded in 2021 and named for keeping humans central to the technology; the cosmological principle has held the term since 1973 and the Greek root since antiquity. This site is not affiliated with the company.

Reviewed 22 September 2026