Why Can Known Odds Feel Safer Than Unknown Odds?

Imagine two jars, each containing 100 red or black balls. Jar A is known to contain 50 of each colour. Jar B also contains 100 balls, but its colour mix is unknown. If drawing red wins a prize, many people choose Jar A. When the winning colour changes to black, they often still choose Jar A. Rather than believing one colour is scarce in Jar B, they appear to avoid a situation in which the probability itself cannot be specified.

This pattern comes from Daniel Ellsberg’s 1961 thought experiments and is commonly called ambiguity aversion. Risk and ambiguity differ: Jar A still has an uncertain outcome, but its probability is known; in Jar B, the probability distribution is also unknown. People may suspect missing information, worry that the organiser has an advantage, or give extra weight to the worst plausible case. Studies repeatedly observe preferences for known odds, but the evidence does not reduce these possible mechanisms to one cause.

The tendency can be useful. If limited information genuinely makes a model unreliable, caution prevents a guess from masquerading as a precise estimate. Yet it can also make a numerical option seem safer merely because it displays a percentage, even when that number comes from a small or irrelevant sample. A stated probability is not automatically better evidence, and an unknown probability is not necessarily more dangerous.

When choosing between a familiar but modest option and an unfamiliar one with sparse data, ask what is unknown: the outcome, or the probability model used to estimate it? Could more information narrow the range? If not, is the worst plausible outcome genuinely intolerable? Ambiguity aversion is an average tendency, not a fixed personality trait. Gains and losses, familiarity, framing and experience all change its strength.

https://doi.org/10.2307/1884324
https://doi.org/10.1257/jel.30.2.325