Experiment #83 · Thought experiment

The Two Envelopes Paradox

Expected-value reasoning misbehaves

Folklore; modern: Maurice Kraitchik (1953), John Broome (1995) · 1953 · Decision theory

First published: M. Kraitchik, *Mathematical Recreations* (1953); modern: J. Broome, "The Two-Envelope Paradox", *Analysis* 55 (1995).

You hold an envelope containing $X. The other envelope contains either $2X or $X/2. Switching has expected value 1.25X. So you should switch — but the same logic applies after switching.

Two envelopes are presented; one contains twice the amount of the other. You select one and find $X inside. The other contains either $2X (probability 1/2) or $X/2 (probability 1/2), giving expected value 1.25X — so you should switch. After switching, the same reasoning recommends switching back. The paradox reveals subtleties in expected-value reasoning under ill-defined priors; standard resolutions appeal to improper priors over unbounded distributions or to conditioning issues. The case is a clean illustration of how naive Bayesian reasoning can produce contradictions.

Formulation

Envelopes A and B; one holds $Y, other $2Y. Choose A, observe $X. P(B = 2X) = P(B = X/2) = 1/2 (naive). E(B) = (2X + X/2)/2 = 1.25X > X. Hence switch. Same applies after switching.

Dimensions Engaged

Observer

Observer · Agency under uncertainty: when expected-value reasoning self-destructs, what should an agent do?

Information

Tests how prior assumptions implicit in expected-value calculations can produce contradictions when ill-specified.

Responses — How Schools Engage

Affirms / takes the bait 2

A clean demonstration that without operational specification of the prior, "expected value" lacks definite content. Decision theory is operational physics for choice.

Even decision theory — the supposed bedrock of rational choice — can dissolve into contradiction under sufficiently abstract setups. A vindication of skeptical caution.

Reframes the question 2

The paradox is a technical artifact of improper priors over unbounded distributions; with any proper prior, expected-value reasoning is consistent. A useful warning against careless Bayesianism.

In any actual application, the prior distribution is constrained by context; the paradox arises from a context-free idealisation.

Holds it inconclusive 1

Multiple resolutions (Broome's conditioning analysis; Albers' bounded-utility resolution); the case continues to motivate careful decision theory.

Related Experiments

Experiments engaged by an overlapping set of schools — likely to surface the same fault lines.

Further reading

  • Broome (1995), op. cit.
  • Albers, Kooi, Schaafsma, "Trying to resolve the two-envelope problem", *Synthese* 145 (2005)

Related Historical Debates

Debates that share dimensions and/or aligned schools with this experiment.

Personas Most Aligned With This Experiment

Ranked by total declared-influence weight in the schools that respond to this experiment.

Works Most Aligned With This Experiment

Ranked by total declared-influence weight in the schools that respond to this experiment.

Related Films

Films engaging the same dimensions as this experiment.

Related Contemporary Dilemmas

Dilemmas that engage the same dimensions as this experiment.

← The Lottery Paradox The Surprise Examination Paradox →