"One major challenge in robotics research is to develop robots able to operate and help humans in everyday life. Personal robots are an emerging technology with a significant potential across broad applications (eldercare, healthcare, education, etc). However, for acting in a human world, acting for, with, or in contact of humans, we need to provide the robot with the abilities not only to act but also to interact. For humans, this capacity is named social intelligence and makes us exhibit fundamental communication skills such as looking at someone who speaks to you, adapting action speed to the person we interact with, acting in synchrony, sharing knowledge about tasks state, etc. A personal robot would need to be able to deal with such social skills. We want to study this question in the particular context of joint task realization, where the robot and the human need to cooperate to achieve a task, with particular attention to figuring out when and how to use communication (verbal and non verbal) for coordination of action and maintaining common ground. This proposal’s objectives are: (a) to develop a robot knowledge database able to handle social understanding of the context (b) to study and develop robots’ (communication) social skills (c) to develop a framework to tackle human-robot joint task performance (d) to implement and demonstrate the developed framework Our idea is to give the robot the ability to (a) “have in mind” its partner’s beliefs and intentions (b) to enable it to support joint actions that include flexible and tightly intermeshed communication for establishing/re-establishing common-ground. These abilities will be integrated in a (c) framework that will tackle task realization and needed communication. This framework will be used (d) to demonstrate the system on real robots. The goal is to develop a conceptual framework that could be implemented both in the outgoing (MIT, USA) and in the return host (LAAS-CNRS, France)"