In this paper we introduce the concept of emergent representations issued from a decomposition along two axis: individual/collective and internal/external. In a collective/internal composition, emergent representations are seen as internal stable and non-reactive complex adaptive systems. Based on natural optimization algorithms, like Particle Swarm Optimization (PSO) and Ant Colony Optimization (ACO), we show how representations can emerge from Schelling's segregation model. Selection, building blocks and "inert" information requirements concerning emergence of representations are illustrated.