Ellsberg Paradox¶

Show that people prefer betting on a known-composition urn over an ambiguous one of equal expected value on both colors at once — a pattern no single subjective probability can rationalize — proving ambiguity is a separately priced dimension of uncertainty distinct from risk.

Version: v3 · 2026-09-28 · History · 3 corrections
Domain-specific #: 325
Domain group: Social Sciences
Origin domain: Economics & Finance
Subdomain: Decision Theory Under Uncertainty → Economics & Finance

Core Idea¶

The Ellsberg paradox is the finding that people prefer bets on urns with known probabilities over urns with unknown composition of equal expected value — and do so on both the red bet and the black bet, a pattern no single subjective probability over the ambiguous urn can rationalize (it would require P(red) < 0.5 and P(black) < 0.5 at once). It establishes that ambiguity — uncertainty over which distribution governs — is a separately priced dimension distinct from risk, and violates Savage's sure-thing principle.

How would you explain it like I'm…

The Mystery Jar Puzzle

Imagine two jars of red and black balls. In one jar you know there are exactly half red and half black. In the other jar you don't know the mix at all. Most people choose the known jar whether they are hoping for red or for black. That is puzzling, because if you think the mystery jar has few reds, it must have lots of blacks, so it can't be worse for both colors.

Known Odds Beat Unknown Odds

Jar A has 50 red and 50 black balls. Jar B has 100 balls, but nobody tells you how many are red. You win $100 if you draw the color you named. Most people pick Jar A both when they bet on red and when they bet on black. No single guess about Jar B can explain both choices, because fewer reds would mean more blacks. This shows that people dislike not knowing the odds, even when the average prize is the same.

Ambiguity Aversion Paradox

The Ellsberg Paradox, found by Daniel Ellsberg in 1961, compares a known urn with 50 red and 50 black balls to an ambiguous urn with 100 balls in an unknown ratio. People prefer the known urn when betting on red and also when betting on black. Preferring it for red would require believing P(red) < 0.5 in the unknown urn. Preferring it for black would require believing P(black) < 0.5. Since the two probabilities must add to 1, both beliefs cannot hold at once. So the choices can't come from any single probability belief. The finding separates risk, where the odds are known, from ambiguity, where the odds themselves are unknown, and shows that people are ambiguity-averse.

The Ellsberg Paradox is the empirical finding that decision-makers prefer bets with known probabilities over bets with unknown probability composition, even at equal expected value. In the canonical design, Urn A has 50 red and 50 black balls, while Urn B has 100 balls in an unspecified ratio. Subjects prefer Urn A whether they bet on red or on black. Any single subjective probability that rationalizes both choices would need P(red) < 0.5 and P(black) < 0.5 in Urn B, which contradicts additivity. The pattern therefore violates Savage's sure-thing principle. It cannot come from any single probability measure over Urn B. It shows that ambiguity, meaning uncertainty about which distribution applies, is priced separately from risk. People display ambiguity aversion and accept lower expected value to avoid it. The finding launched models such as Choquet expected utility, maxmin expected utility over multiple priors, smooth ambiguity and variational preferences.

Scope of Application¶

Requires a chooser whose belief-states distinguish "I know the probability" from "I don't."

Decision theory beyond expected utility — the home: the entry test every generalisation must pass.
Asset pricing — ambiguity premia help explain the equity premium and home bias.
Insurance and incomplete contracts — loading beyond actuarial loss when the distribution is uncertain.
Deep-uncertainty policy — precaution for climate, pandemics, and AI risk under Knightian uncertainty.
Behavioural finance — the field ambiguity premium extending the lab finding to real choices.

Clarity¶

The paradox separates two dimensions the Savage-Bayesian frame collapses: risk (known probabilities) and ambiguity (the distribution itself unspecified). Forcing the known-urn preference on both bets proves no single probability rationalizes the choices, so belief must be richer than one measure. It converts a vague sense that the standard model is missing something into a definite missing primitive — ambiguity aversion — plus a diagnostic to locate it.

Manages Complexity¶

The two-urn pattern compresses two sprawls at once: a proliferation of expected-utility generalisations becomes a single pass/fail gate (does the model accommodate the Ellsberg choices?), and a scatter of field anomalies — equity premia, home bias, catastrophe loading — becomes one priced primitive. The analyst tracks one new dimension and one parameter, reading the qualitative pattern off a single risk-versus-ambiguity fork rather than re-deriving each anomaly.

Abstract Reasoning¶

Keyed to the risk-versus-ambiguity fork, the paradox licenses a boundary-drawing move (classify a prospect before pricing it), a diagnostic impossibility move (the two-bet pattern proves belief is richer than one probability), a predictive move (ambiguity aversion demands a premium wherever the distribution is unpinned), and a model-admission gate (a generalisation is admitted only if it reproduces the pattern).

Knowledge Transfer¶

Within decision theory the paradox transfers as mechanism in two capacities — a reusable diagnostic instrument (the two-urn test) and a located primitive (ambiguity aversion) that carries across asset pricing, insurance, and deep-uncertainty policy. The two-urn design itself applies only where belief-states distinguish knowing from not knowing a probability, so it does not travel to thermostats or populations. What recurs cross-domain is the parent Ambiguity Aversion; the urns and colour bets are the apparatus, not the portable content.

Relationships to Other Abstractions¶

Current abstraction Ellsberg Paradox Domain-specific

Parents (2) — outgoing typed relations

Ellsberg Paradox is a kind of Paradox Prime

The Ellsberg paradox is a paradox whose apparently reasonable preferences produce a contradiction under every single-prior representation.
Ellsberg Paradox is a decomposition of Ambiguity Aversion Domain-specific

Removing the two-urn and Savage-axiom apparatus from the Ellsberg paradox leaves the known-over-unknown probability preference of ambiguity aversion.

Hierarchy paths (3) — routes to 3 parentless roots

Ellsberg Paradox → is a kind of Paradox

Show alternative paths (2)

Neighborhood in Abstraction Space¶

Ellsberg Paradox sits in a sparse region of the domain-specific corpus (68^th percentile for distinctiveness): few abstractions share its structure, so a faithful description tends to retrieve it precisely.

Family — Game-Theoretic Models & Paradoxes (38 abstractions)

Nearest neighbors

Computed from structural-signature embeddings · 2026-10-25