Learning interpretable communication is essential for multi-agent and human-agent teams (HATs). In multi-agent reinforcement learning for partially-observable environments, agents may convey information to others via learned communication, allowing the team to complete its task. However, the utility of discrete and sparse communication in human-agent team experiments has not yet been investigated. In this work, we analyze the efficacy of sparse-discrete methods for producing emergent communication that enables high agent-only and human-agent team performance. 

The ability to infer another human’s beliefs, desires, intentions and behaviour from observed actions and use those inferences to predict future actions has been described as the Theory of Mind (ToM).  In conjunction with researchers at CMU we are working to develop AI agents capable of making such ToM inferences.  This research started with modelling single human's mental state in a search and rescue task, and later extended to modelling the team state (i.e. shared mental model) of three human rescuers.

- Li, H., Fan, Y., Zheng, K., Lewis, M., & Sycara, K. (2023). Personalized Decision Supports based on Theory of Mind Modeling and Explainable Reinforcement Learning. In 2023 IEEE International Conference on Systems, Man, and Cybernetics (SMC 2023)

- Li, H., Oguntola, I., Hughes, D., Lewis, M., & Sycara, K. (2022, August). Theory of Mind Modeling in Search and Rescue Teams. In 2022 31st IEEE International Conference on Robot and Human Interactive Communication (RO-MAN) (pp. 483-489). IEEE.

As domestic service robots become more common and widespread, they must be programmed to act appropriately by efficiently accomplishing tasks while aligning their actions with relevant norms. We are taking the first step to equip domestic robots with normative reasoning competence which is understanding the norms that people apply to the behaviour of robots in specific social contexts.