Reviews: Multi-View Decision Processes: The Helper-AI Problem

The paper introduces multi-view decision processes (MVDP) as a cooperative game between two agents, where one agent may have an imperfect model of the transition probabilities of the other agent. Generally, the work is motivated well, the problem is interesting and clearly relevant to AI and the NIPS community. The theoretical results are nicely complemented by experiments. Presentation is mostly smooth: the main goals are consistent with the narrative throughout the paper, but the some details from the experiments seem less clear (seem points below.) Some points are confusing and need further clarification: 1.