Deep Interpretable Models of Theory of Mind For Human-Agent Teaming