Probabilistic aggregation is a self-organised behaviour studied in swarm robotics. It aims at gathering a population of robots in the same place, in order to favour the execution of other more complex collective behaviours or tasks. However, probabilistic aggregation is extremely sensitive to experimental conditions, and thus requires specific parameter tuning for different conditions such as population size or density. To tackle this challenge, in this paper, we present a novel embodied evolution approach for swarm robotics based on social dynamics. This idea hinges on the cultural evolution metaphor, which postulates that good ideas spread widely in a population. Thus, we propose that good parameter settings can spread following a social dynamics process. Testing this idea on probabilistic aggregation and using the minimal naming game to emulate social dynamics, we observe a significant improvement in the scalability of the aggregation process.