Elusive Optimality in the Box Problem
The paper seeks to extend the investigation of the so called Box Problem, also variously known as the ``wallet game, the ``exchange paradox, the ``two envelope paradox, and the ``Ali-Baba paradox. The paradox involves two identical boxes one of which is known to have twice the (positive) sum of money contained in the other. A decision maker is supposed to be in possession of one of the boxes and must decide whether or not to trade it for the other box. Straight expected value comparisons lead to the paradoxical conclusion that trading is always preferred to status quo. In the above problem, the decision maker is allowed to find out what is the sum of money contained in the box in his/her possession before making his decision. Clearly, Bayesian analysis considering prior probability distributions of possible sums, which may be affected once the content of the box is known, may indicate strategies other than always trading to turn out to be superior. While that does turn out to be the case for \textit{some} priors, surprisingly, there are some priors for which always trading does still turn out to be optimal, The paper tries to extend the Bayesian analysis by analysing threshold strategies when the expected values are finite and characterizing conditions under which one such is optimal. It goes on to consider the kind of priors for which always trading is seemingly an optimal strategy but notes that it involves an infinite expected value. It seeks to analyse more closely this class of `anomalous distributions'. Stochastic dominance does not turn out to be a very useful method of comparison. Comparison of expectation of gain (which may be finite even though the expected value itself is infinite) of one threshold strategy over another does not yield a very fruitful method either. Simulation methods yield some interesting observations. No absolutely clear-cut results emerge.
- Imperfect information as a source of non-symmetry in the two envelope problem
- On the two-box paradox
- Randomized switching in the two-envelope problem
- A resolution of the ‘exchange paradox’
- The Box Problem: To Switch or Not to Switch
- An optimal threshold strategy in the two-envelope problem with partial information
- Diagrammatic Representation and Inference
- Anna Karenina and the two envelopes problem
This page was built for publication: Elusive Optimality in the Box Problem
Report a bug (only for logged in users!)Click here to report a bug for this page (MaRDI item Q2773225)
