Bounded Reasoning and Higher-Order Uncertainty