Experimental Study for the Effect of Atypical Speed of Humanlike Agent’s Facial Movements on Facial Expression Recognition

Abstract

Facial expressions of agents (e.g. robots or computer-animated avatars) play a crucial role to implicitly convey their emotional states to users, and such emotional states are expected to be conveyed appropriately. However, “uncanny valley” effect implied that dynamic aspect of humanlike agents’ face should have high impact on human evaluation towards agents. In this study, we conducted the experiments where participants observed facial expressions of humanlike CG avatars—“Disgust” and “Happiness”—and speed was modulated to generate atypical (fast and slow expression) or typical facial dynamics. The results indicated two effects of atypical speed of the agent’s facial movement: 1) Fast expression of “Happiness” can make human feel strange but that of “Disgust” did not, according to evaluation to agent’s feature 2) Human strongly feels strange and repulsive when facial expressions are shown slowly. And, as is partly shown, human extracts social intentions of the agents from their facial expression, when the facial expressions are shown slowly.