Development of Counting Ability: An Evolutionary Computation Point of View

Examination of numerical cognition encompasses multiple facets (eg, discrete vs. continuous properties, subitizing, estimation, counting, etc.). Many models have been suggested to explain these features. By looking into the basic ability to perceive size, against the complex one of counting, we hypothesize that counting system evolved on the basis of a primitive size perception system rather than the two systems evolved separately. In this chapter, we present a novel way of using evolutionary computation techniques to evolve artificial neural networks (ANNs) first to perceive size and then to count, and compare their counting skills to a different group of ANNs who evolved to count from scratch. The results revealed better counting skills when evolving first to perceive size (or other classification task) and then to count over those who evolved just to count. In addition, ANNs who evolved with continuous stimuli presented better counting skills than those evolved with discrete stimuli.