Design of human likeness in HRI from uncanny valley to minimal design

Abstract

Human likeness of social agents is crucial for human partners to interact with the agents intuitively because it makes the partners unconsciously respond to the agents in the same manner as what they show to other people. Although many studies suggest that an agent's human likeness plays an important role in human-robot interaction, it remains unclear how to design humanlike form that evokes interpersonal behavior from human partners. One approach is to make a copy of an existing person. Although this extreme helps us explore how we recognize another person, the Uncanny Valley effect must be taken into account. Basic questions, including why we experience the uncanny valley and how we overcome it should be addressed to give new insights into an underlying mechanism in our perception of human likeness. Another approach is to extract crucial elements that represent human appearance and behavior, as addressed in design of computer-animated human characters. The exploration of minimal requirement to evoke interpersonal behavior from human partners provides more effective and simpler way to design social agents that facilitate communication with human.