Developing the knowledge of number digits in a child-like robot

Abstract

Number knowledge can be boosted initially by embodied strategies such as the use of fingers. This Article explores the perceptual process of grounding number symbols in artificial agents, particularly the iCub robot—a child-like humanoid with fully functional, five-fingered hands. It studies the application of convolutional neural network models in the context of cognitive developmental robotics, where the training information is likely to be gradually acquired while operating, rather than being abundant and fully available as in many machine learning scenarios. The experimental analyses show increased efficiency of the training and similarities with studies in developmental psychology. Indeed, the proprioceptive information from the robot hands can improve accuracy in the recognition of spoken digits by supporting a quicker creation of a uniform number line. In conclusion, these findings reveal a novel way for the humanization of artificial training strategies, where the embodiment can make the robot’s learning more efficient and understandable for humans. Number processing is linked to bodily systems, especially finger movements. The authors apply convolutional neural network models in the context of cognitive developmental robotics. They show that proprioceptive information in the child-like robot iCub improves accuracy and recognition of spoken digits.