Towards Trustworthy Human-AI Teaming under Uncertainty