I do believe you get in an interesting confusion when talking about the blue tentacle example. When you’re asked to answer why, you’re not prompted for scenarios which are likely beforehand, you’re prompted for argmax(P(Ai|T)) (T is the situation of waking up that day with a blue tentacle, argmax taken to choose from i’s), which equals—by theoreme of Bayes—argmax(P(T|Ai)*P(Ai)/P(T))=argmax(P(T|Ai)*P(Ai))=argmax(P(Ai&T)). And this—despite the fact that ALL P(Ai&T) are very small and the P(T), equaling their sum, is also too small to worry about it in usual life—is an allowed mathematical task in itself. The task of asking “who is taller, John or Mary” does not entail either John or Mary is tall—in fact, they can be two dwarves. Same logic applies here.
I do believe you get in an interesting confusion when talking about the blue tentacle example. When you’re asked to answer why, you’re not prompted for scenarios which are likely beforehand, you’re prompted for argmax(P(Ai|T)) (T is the situation of waking up that day with a blue tentacle, argmax taken to choose from i’s), which equals—by theoreme of Bayes—argmax(P(T|Ai)*P(Ai)/P(T))=argmax(P(T|Ai)*P(Ai))=argmax(P(Ai&T)). And this—despite the fact that ALL P(Ai&T) are very small and the P(T), equaling their sum, is also too small to worry about it in usual life—is an allowed mathematical task in itself. The task of asking “who is taller, John or Mary” does not entail either John or Mary is tall—in fact, they can be two dwarves. Same logic applies here.