CLASP
The Centre for Linguistic Theory and Studies in Probability

Gaze, Personality, and the Uncanny Valley: Implicit Cues of Uncanny feelings and Interaction Strategies to Overcome them

In this talk, I will give an overview of two studies that I have recently published in Frontiers in Robotics and AI and Computers in Human Behavior with colleagues of Uppsala University and the University of Potsdam. Both the studies revolve around a geography-themed Rapid Dialogue Game (RDG), the map game, in which a human and a robot are tasked with identifying as many countries as possible on the world map in a given time of 10 minutes. In the Frontiers in Robotics and AI paper, we presented a study in which we explored how people’s perceptions of a Furhat robot and engagement with it and the map game could be assessed through implicit and continuous measurement techniques, such as gaze, over repeated interactions. We involved participants in three interaction sessions with multiple days of zero exposure in between. Each interaction session consisted of 10 minutes of play with the map game and two short social chats with Furhat before and after the game. Furhat’s facial texture was varied so as to make the robot’s face look humanlike, mechanical, or a morph between the humanlike and the mechanical. We measured participants’ gaze patterns with a wearable eye-tracker and gauged their perception of the robot and engagement with it and the joint task using questionnaires. The study results disclosed that aversion of gaze in a social chat is an indicator of a robot’s uncanniness and that the more people gaze at the robot in a joint task, the worse they perform. Furthermore, the analyses of gaze patterns in repeated interactions showed that people’s mutual gaze in a social chat develops congruently with their perceptions of the robot over time. These are key findings for the HRI community as they entail that gaze behavior can be used as an implicit measure of people’s perception of robots in a social chat and of their engagement and task performance in a joint task.

In the Computers in Human Behavior paper, we presented a study in which participants played the map game with a Furhat robot displaying one of two personalities, which corresponded to two different interaction strategies. The robot was either optimistic and encouraging or impatient and provocative. We performed the study in a science museum and recruited participants among the visitors. The study’s goal was to understand whether a robot’s interaction strategy, in this case, the robot’s personality, could weaken initial uncanny feelings due to the robot’s appearance (humanlike or a morph between a humanlike and a mechanical appearance). Our findings suggest that giving participants the time to interact with a robot that is rated high on agreeableness, emotional stability, and conscientiousness can indeed lower initial perceptions of uncanniness. This study has important implications for the design of dialogue systems as the robot’s utterances that contributed to the perception of the robot’s personalities were completely authored and evaluated by crowd-workers on Amazon Mechanical Turk (AMT).

• Perugia, G., Paetzel-Prüsmann, M., Madelene Alanenpää, & Castellano, G. (2021). I Can See it in Your Eyes: Gaze as an Implicit Cue of Uncanniness and Task Performance in Repeated Interactions with Robots. Frontiers in Robotics and AI. DOI: 10.3389/frobt.2021.645956

• Paetzel-Prüsmann, M., Perugia, G., & Castellano, G. (2021). The Influence of Robot Personality on the Development of Uncanny Feelings. Computers in Human Behavior, 106756. DOI: 10.1016/j.chb.2021.106756