The Newcomb Problem: An Unqualified Resolution

Abstract

The Newcomb problem is analysed here as a type ofcommon cause problem. Inrelation to such problems, if you take the dominatedoption your expected outcomewill be good and if you take the dominant optionyour expected outcome will be notso good. As is explained, however, these arenot conventional conditional expectedoutcomes but `conditional evidence expectedoutcomes' and while in the deliberationprocess, the evidence on which they are based isonly hypothetical evidence.Conventional conditional expected outcomes aremore sensitive to your currentepistemic state in that they are based purely onactual evidence which is available toyou during the deliberation process. So althoughthey are conditional on a certain actbeing performed, they are not based on evidencethat you would have only if that actis performed. Moreover, for any given epistemicstate during the deliberationprocess, your conventional conditional expectedoutcome for the dominant option willbe better than that for the dominated option. Theprinciple of dominance is thus inperfect harmony with the conventional conditionalexpected outcomes. In relation tothe Newcomb problem then, the evidence unequivocallysupports two-boxing as therational option. Yet what is advanced here isnot simply a two-boxing strategy. Tosee why, two stages to the problem need to berecognised. The first stage is thatwhich occurs before the information used by thepredictor in making his predictionshas been gained. The second stage is after thispoint. Provided that you are still inthe first stage, you have an opportunity toinfluence whether or not the predictorplaces the $1m in the opaque box. To maximisethe probability that it is, you need tocommit yourself to one-boxing.