Following the development of cognitive sciences, a lot of works in robotics and Artificial Intelligence (Walter, 1950, Khatib, 1986, Brooks and Stein, 1994, Arkin, 1998) have shown the interest of a direct ‘sense-act’ approach where the intelligence requires a body as opposed to the classical Artificial Intelligence using the ‘sense-think-act’ paradigm and involving a strong dissociation between the body and mind. In this paradigm, ‘thinking’ involves modeling and planning capabilities and could be performed without a body 1 . For instance, in classical robotics, walking can be seen as producing a particular sequence of actions for the different joints of the two legs (from the ankles to the feet) in order to maintain the equilibrium and to reach a desired location. This approach implies a precise planning and a real-time control of the different joints to avoid falling. At the opposite, the solution of a passive walker (McGeer, 1990) takes benefits from the physical and dynamical properties of human biped anatomy enabling walking without any control circuit. Hence, what is the real role of our brain? As discussed by Pfeifer, we consider the Body Shapes the Way We Think (Pfeifer et al. , 2007, Gallagher, 2005). Yet, is it possible to generalize this approach to higher-level cognitive tasks involving social interactions?