Uncanny valley for interactive social agents: an experimental study

Abstract

The uncanny valley hypothesis states that users might experience eerie when interacting with almost but not fully human-like artificial characters. The advancements in artificial intelligence, robotics, and computer graphics have led to life-like virtual humans and humanoid robots. It is necessary to revisit the hypothesis to check if they positively or negatively affect the current population, who are much more accustomed to the latest technologies. In this paper, we study and present a unique evaluation of the uncanny valley hypothesis by allowing participants to interact live with four different humanoid robots (of varying levels of humanlikeness). To evaluate the affinity of each robot, each participant needs to fill a survey questionnaire. Apart from this, we also use deep learning methods to quantify the participants’ emotional states using multi-modalcues, including visual, audio, and text, by recording the participant-robot interaction. The multi-modal analysis and surveys provide interesting results and insights into the uncanny valley hypothesis.