Robots are increasingly introduced to work in concert with people in high-intensity domains, such as manufacturing, space exploration and hazardous environments. Although there are numerous studies on human teamwork and coordination in these settings, very little prior work exists on applying these models to human-robot interaction. In this paper we propose a novel framework for applying prior art in Shared Mental Models (SMMs) to promote effective human-robot teaming. We present a computational teaming model to encode joint action in a human-robot team. We present results from human subject experiments that evaluate human-robot teaming in a virtual environment. We show that cross-training, a common practice used for improving human team shared mental models, yields statistically significant improvements in convergence of the computational teaming model (p=0.02) and in the human participants' perception that the robot performed according to their preferences (p=0.01), as compared to robot training using a standard interactive reinforcement learning approach.