Social robots are often designed to present with a human-like identity that include masculine or feminine genders, to facilitate natural interactions, yet this approach poses significant risks: it can encourage emotional and cognitive over-reliance on robotic partners, and reinforce stereotypes by associating social categories, such as gender, with specific behaviors, traits, and social roles. Use of a non-gendered, mechanical identity for social robots may be an approach that can reduce the perpetuation of gender stereotypes and over-reliance on robots, but research is needed to understand the impact on interactions and perceptions of robots presenting a mechanical identity. To address this need, we conducted an exploratory study of participants (N=12) collaborating on two problem-solving tasks with a social robot that expressed a mechanical, non-gendered identity, including mechanical pronouns (it/its). Based on post-task interviews and surveys, we present exploratory insights that non-gendered, mechanical identities may be successfully implemented during collaborative human-robot interactions. Most participants described the robot's mechanical identity as acceptable and believable, and had similar perceptions of competence, warmth and comfort levels after a stereotypically feminine task (a makeup selection task) and a more neutral task (solving the tower of Hanoi problem). Robot competence and warmth seem to be supported by feeling the robot was helpful, supportive and trustworthy, and feelings of discomfort toward the robot diminished over time. Some participants found the animacy and warmth of the robot made the use of mechanical pronouns feel too impersonal, and some felt the robot needed sufficient context and justification for its actions to maintain perceived competence. These findings have preliminary implications for the practical and ethical design of robot identities in social collaborative settings, suggesting that non-gendered, mechanical identities might possibly be successfully implemented without necessarily compromising the overall quality of human-robot interactions.
Grosso et al. (Sat,) studied this question.