Designing Robot Identity: The Role of Voice, Clothing, and Task on Robot Gender Perception

TL;DR

This work investigates how robot gender perception emerges from the interaction of voice, clothing, and task context. It develops a design framework with voice and clothing design principles, validated through three video-based studies that separately and jointly manipulate gender cues. The results show that gender is performative and context-dependent rather than a simple additive effect, with task context and cue alignment shaping perceptions and social attributes. The findings offer practical guidance for designing robot identities in ethical, context-sensitive ways, highlighting the importance of user-centered and participatory considerations in HRI.

Abstract

Perceptions of gender are a significant aspect of human-human interaction, and gender has wide-reaching social implications for robots deployed in contexts where they are expected to interact with humans. This work explored two flexible modalities for communicating gender in robots--voice and appearance--and we studied their individual and combined influences on a robot's perceived gender. We evaluated the perception of a robot's gender through three video-based studies. First, we conducted a study (n=65) on the gender perception of robot voices by varying speaker identity and pitch. Second, we conducted a study (n=93) on the gender perception of robot clothing designed for two different tasks. Finally, building on the results of the first two studies, we completed a large integrative video-based study (n=273) involving two human-robot interaction tasks. We found that voice and clothing can be used to reliably establish a robot's perceived gender, and that combining these two modalities can have different effects on the robot's perceived gender. Taken together, these results inform the design of robot voices and clothing as individual and interacting components in the perceptions of robot gender.

Figures (6)

• Figure 1: Perceived femininity of voices (-3 represents a masculine voice, 0 represents an ambiguously gendered voice, and +3 represents a feminine voice) as a function of average fundamental frequency ($f_0$) of the utterance. Teal lines represent five-datapoint sliding averages, the shaded region denotes $\pm$ one standard deviation, and the beige dots are individual responses.
• Figure 2: Quori (unclothed) specian2021quori, the robot we selected to use for the clothing design study and integrative video study.
• Figure 3: Appearance modifications of the Quori robot. The first two images represent the feminine and masculine versions of the robot clothing designed for the hotel receptionist task. The second two represent the feminine and masculine clothing designed for the medical professional task.
• Figure 4: Sample frames from the two tasks we selected: medical professional (top) and hotel receptionist (bottom).
• Figure 5: Summary of stimuli and results from the integration study. We found that the manipulation of the social role was significantly different between the two conditions ($p<.001$), with the expected social role of the receptionist task being lower than the social role of the medical professional task (a). Participants saw videos of the robot performing a task with a human. We also found that voice and appearance affected the perception of masculinity (b) and femininity (c) in both conditions.