Attributing human-likeness to an avatar: the role of time and space in the perception of biological motion

Despite well-developed cognitive control mechanisms in most adult healthy humans, attention can still be captured by irrelevant distracting stimuli occurring in the environment. However, when it comes to artificial agents, such as humanoid robots, one might assume that its attention is “programmed” to fol-low a task, thus, being distracted by attention-capturing stimuli would not be ex-pected. We were interested in whether a behavior that reflects attentional capture in a humanoid robot would increase its perception as human-like. We imple-mented human behaviors in a virtual version of iCub robot. Twenty participants’ head movements were recorded, through an inertial sensor, during a solitaire card game, while a series of distracting videos were presented on a screen in their peripheral field of view. Eight participants were selected, and their behavioral reactions (i.e. inertial sensor coordinates, etc.) were extracted and implemented in the simulator. In Experiment 2, twenty-four new participants were asked to rate the human-likeness of the avatar movements. We examined whether move-ment parameters (i.e. angle amplitude, overall time spent on a distractor) influ-enced participants’ ratings of human-likeness, and if there was any correlation with sociodemographic factors (i.e. gender, age). Results showed a gender effect on human-likeness ratings. Thus, we computed a GLM analysis including gender as a covariate. A main effect of the time of movement was found. We conclude that humans rely more on temporal then on spatial information when evaluating properties (specifically, human-likeness) of biological motion of humanoid-shaped avatars.