There is an increasing consensus that existing mobility models, such as the well-known random walk or random waypoint models, are insufficient to represent real node mobility. In this paper, we discuss the need for a better characterization of natural mobility. Our contributions rely on recent advances of real-life network analysis and modelling, and in particular on the observation that natural networks behave on a scale-free basis. We devise then a novel mobility modelling approach that focuses on the behavioral aspect of individuals and the interactions between them. This fulfils a gap between individual and group mobility models. Our first results show a strong relevance of the scale-free distribution in mobility modelling, and open further directions in modelling the costs associated to building a network structure in general.