This work presents a method for designing facial interfaces for sociable android robots with respect to the fundamental rules of human affect expression. Extending the work of Paul Ekman towards a robotic direction, we follow the judgment-based approach for evaluating facial expressions to test in which case an android robot like the Geminoid|DK –a duplicate of an Original person- reveals emotions convincingly; when following an empirical perspective, or when following a theoretical one. The methodology includes the processes of acquiring the empirical data, and gathering feedback on them. Our findings are based on the results derived from a number of judgments, and suggest that before programming the facial expressions of a Geminoid, the Original should pass through the proposed procedure. According to our recommendations, the facial expressions of an android should be tested by judges, even in cases that no Original is engaged in the android face creation.
Proceedings of the International Conference on Social Robotics, 2012, p. 56-65
Social robotics; Geminoid; androids; emotions; facial expressions; emotional health