Multi-agent Logics for Reasoning About Higher-Order Upper and Lower Probabilities