Quantifying the Human Likeness of a Humanoid Robot
In research of human-robot interactions, human likeness (HL) of robots is frequently used as an individual, vague parameter to describe how a robot is perceived by a human. However, such a simplification of HL is often not sufficient given the complexity and multidimensionality of human-robot interaction. Therefore, HL must be seen as a variable influenced by a network of parameter fields. The first goal of this paper is to introduce such a network which systematically characterizes all relevant aspects of HL. The network is subdivided into ten parameter fields, five describing static aspects of appearance and five describing dynamic aspects of behavior. The second goal of this paper is to propose a methodology to quantify the impact of single or multiple parameters out of these fields on perceived HL. Prior to quantification, the minimal perceivable difference, i.e. the threshold of perception, is determined for the parameters of interest in a first experiment. Thereafter, these parameters are modified in whole-number multiple of the threshold of perception to investigate their influence on perceived HL in a second experiment. This methodology was illustrated on the parameters speed and sequencing (onset of joint movements) of the parameter field movement as well as on the parameter sound. Results revealed that the perceived HL is more sensitive to changes in sequencing than to changes in speed. The sound of the motors during the movement also reduced perceived HL. The presented methodology should guide further, systematic explorations of the proposed network of HL parameters in order to determine and optimize acceptance of humanoid robots.