Possibility of Self-Recognizing Robots: From the Perspective of Research on Nonhuman Primates

Abstract

Despite roboticists' endeavours to endow their robots with cognition, robots do not exhibit as complex interactive behaviours as those of their animal counterparts. Our belief is that this failure partially originates from a lack of self-awareness of those robots. Indeed, it is difficult for a robot which cannot discriminate between his own body and its environment or understand his physical configuration to map the behaviours of other agents living in its environment onto his body and therefore exhibit social communication such as imitation (unless such a mapping is explicitely encoded in its behavior). We address here a particular aspect of self-awareness, namely self-recognition, because self-recognition reveals an awareness of stable categorical features of self. The understanding of self-recognition in nonhumans primates has been recently extensively studied and some well-defined experimental tests have been proposed. However, no model has been proposed, partly because, from a psychological point of view, it is difficult to estimate the bias of hidden factors inherent to the experimental setup in the inference of a mechanism. Robotics instead provides an alternative mean for testing theories of behavior, by implementing tentative mechanisms and analyzing the obtained behaviours along lines similar to those used for analyzing animal behaviours. In this paper, we review some work done in self-recognition in nonhuman primates and we propose a framework for exploring this process on a robot.