Towards Designing Android Faces after Actual Humans

Using their face as their prior affective interface, android robots and other agents embody emotional facial expressions, and convey messages on their identity, gender, age, race, and attractiveness. We are examining whether androids can convey emotionally relevant information via their static facial sig-nals, just as humans do. Based on the fact that social information can be accu-rately identified from still images of nonexpressive unknown faces, a judgment paradigm was employed to discover, and compare the style of facial expres-sions of the Geminoid-DK android (modeled after an actual human) and its’ Original (the actual human). The emotional judgments were achieved through an online survey with video-stimuli and questionnaires, following a forced-choice design. Analysis of the results indicated that the emotional judgments for the Geminoid-DK highly depend on the emotional judgments initially made for the Original, suggesting that androids inherit the same style of facial expression as their originals. Our findings support the case of designing android faces after specific actual persons who portray facial features that are familiar to the users, and also relevant to the notion of the robotic task, in order to increase the chance of sustaining a more emotional interaction.