Humanlikeness and Aesthetic Customization's Effect on Trust, Performance, and Affect

Abstract

Human-machine interactions have become a staple of people's daily lives through the use of mobile devices, robotics, and a myriad of smart technologies. Previous research has established that anthropomorphism can significantly affect subjective perceptions of, and interactions with, machines. Furthermore, the ability to customize digital tools has been shown to affect user preferences, video game enjoyment, and the efficacy of digital mental health interventions. This study examined whether the customization of a machine teammate could influence the performance of the human-machine team and generate an affective response on the part of the human teammate. To evaluate this premise, we developed a bomb-defusing task simulation using the Unity game engine wherein participants were randomly assigned to one of two (humanlike or machinelike) robot avatars or were given the ability to customize one. The customizable robot avatar allows the participant to select either a humanlike or machinelike robot and customize the color of the wheels and casing. The customization is aesthetic in nature and has no effect on the functionality of the robot. The game design incorporates a high-risk environment and uncertainty with respect to the bomb-defusing distance and required button presses to encourage cautious guidance of the robot. We predicted that the ability to customize the robot will increase performance and subjective measures of trust, affect, attachment, identification, immersion, and control. We also predicted that the humanlikeness of the robot would increase performance and our subjective measures. Finally, we expected to see a significant effect of customization and humanlikeness such that the customization and humanlikeness have an additive effect on performance and our subjective measures. The results of all analyses were nonsignificant. These results may help inform the design of such systems and address fears that customization could lead to over-empathizing with a machine teammate in a way that would reduce use in high-risk environments.