Evolutionary model of the concepts of ‘color chip’ and ‘color category’ in Evolutionary models of color categorization based on discrimination

Komarova et al. (2007) showed that a color naming system could be evolved by combining perceptual discrimination, basic pragmatic constraints on communication, and simple learning rules. Their system relied on a fixed number of color chips and of color categories. Here, I construct an evolutionary game theory model where the equivalent concept of the former and the number of the latter are emergent properties of the learning algorithm. The solution involves not a static but a continuously evolving “perceptual space,” where the continuous change is arguably a more realistic evolutionary model and echo psychological models in that it is based on a form of memory, possessing a dynamic learning feature. It is shown that such dynamic structure does lead to a stable structure of both color concepts and categories. Simulation results are aimed at replicating the major result of Komarova et al. (2007) both for individual learning and in a population of learners. The concept of icon chips is introduced whose number and stability in an icon structure are an emergent property of multiple exposures to stimuli (colors) drawn from a continuous color space. The icon chip represents a form of a memory that changes in response to novel stimuli (color) and to the distribution of those stimuli. The results replicate those of Komarova et al. (2007) using constructs they only assumed could be evolved and demonstrate that a stable macro-structures can be maintained with continuously evolving microstructures, arguably a realistic feature evolutionary learning models.