Newcomb’s Problem as a Reconstruction of Causal History

This paper proposes a reinterpretation of Newcomb’s Problem as an error in how the decision situation is framed. The standard argument for two-boxing treats the content of the opaque box as fixed and independent of the agent’s choice, despite its dependence on the predictor’s model of the agent. I argue that the decision should instead be understood as a reconstruction problem over causal history: a present choice does not influence the past, but it constrains which pasts are compatible with it. This perspective dissolves the apparent conflict between dominant reasoning and expected utility by explicitly modeling the decision as an inverse problem over causal histories, rather than a forward optimization over fixed states.