An action is later in logical time than the prior considerations that determine it. And the prior considerations shouldn’t always be consequentialist with respect to the resulting action (they could be unrelated background facts, or consequentialist with respect to some other decision, perhaps made by a different agent, which is important for acausal coordination). I think this is conceptually cleaner than only considering consequentialist arguments.
In general, you can’t insist on waiting for a prior consideration to settle before making a decision, since it can take too long (or it can even directly wait for your own action before giving its answer, making it impossible to out-wait it). So there needs to be a timeout. Proof length bounded provability ia one way of doing it, where you wrap all prior considerations in provability-within-length-k boxes. But the issue here is that you might want to care about the proof being used, rather than be OK with any old sufficiently short proof. This is similar to how you might want to care which prior considerations to use, rather than go with all possible consequentialist considerations with respect to your action.
An action is later in logical time than the prior considerations that determine it. And the prior considerations shouldn’t always be consequentialist with respect to the resulting action (they could be unrelated background facts, or consequentialist with respect to some other decision, perhaps made by a different agent, which is important for acausal coordination). I think this is conceptually cleaner than only considering consequentialist arguments.
In general, you can’t insist on waiting for a prior consideration to settle before making a decision, since it can take too long (or it can even directly wait for your own action before giving its answer, making it impossible to out-wait it). So there needs to be a timeout. Proof length bounded provability ia one way of doing it, where you wrap all prior considerations in provability-within-length-k boxes. But the issue here is that you might want to care about the proof being used, rather than be OK with any old sufficiently short proof. This is similar to how you might want to care which prior considerations to use, rather than go with all possible consequentialist considerations with respect to your action.