Knightian uncertainty makes sense to me at least in a restricted scenario where something is optimizing directly against your belief state.
Say you’re Omega in the transparent Newcomb’s Problem, and the pesky human you’re simulating implements the strategy of always doing the opposite of what your revealed prediction indicates (i.e. taking both boxes when both boxes are full, and taking only box B if it is empty).
Then even Omega has Knightian uncertainty about what is going to happen, its belief probably flip-flops between “human takes both boxes” and “human takes only box B”.
Merely anti-inductive scenarios are not enough, though—you need this explicit optimization against belief state.
Furthermore, it could be that there is often optimization against stated beliefs, especially in political contexts, or one wants to avoid optimization against recursive self-belief, which could be a reason for people to avoid explicit predictions about the success of their own plans.
In political contexts, this predicts that conflict theorists more often appeal to Knightian uncertainty than mistake theorists.
Knightian uncertainty makes sense to me at least in a restricted scenario where something is optimizing directly against your belief state.
Say you’re Omega in the transparent Newcomb’s Problem, and the pesky human you’re simulating implements the strategy of always doing the opposite of what your revealed prediction indicates (i.e. taking both boxes when both boxes are full, and taking only box B if it is empty).
Then even Omega has Knightian uncertainty about what is going to happen, its belief probably flip-flops between “human takes both boxes” and “human takes only box B”.
Merely anti-inductive scenarios are not enough, though—you need this explicit optimization against belief state.
Furthermore, it could be that there is often optimization against stated beliefs, especially in political contexts, or one wants to avoid optimization against recursive self-belief, which could be a reason for people to avoid explicit predictions about the success of their own plans.
In political contexts, this predicts that conflict theorists more often appeal to Knightian uncertainty than mistake theorists.