How context shapes the appropriateness of a robot’s voice

Social robots have a recognizable physical appearance,a distinct voice, and interact with users in specificcontexts. Previous research has suggested a ‘matchinghypothesis’, which seeks to rationalise how people judge arobot’s appropriateness for a task by its appearance. Otherresearch has extended this to cover combinations of robotvoices and appearances. In this paper, we examine the missingconnection between robot voice, robot appearance, anddeployment context. In so doing, we asked participants tomatch a robot image to a voice within a defined interactioncontext. We selected widely available social robots, identifiedtask contexts they are used in, and manipulated the voicesin terms of gender, naturalness, and accent. We found thatthe task context mediates the ‘matching hypothesis’. Peopleconsistently selected a robot based on a vocal feature for acertain context, and a different robot based on the same vocalfeature for another context. We suggest that robot voice designshould take advantage of current technology that enables thecreation and tuning of custom voices. They are a flexible toolto increase perception of appropriateness, which has a positiveinfluence on Human-Robot Interaction.QC 20201014