A case for diverse social robot identity performance in education
by Lux Miranda, Ginevra Castellano, and Katie Winkle
Department of Information Technology
Uppsala University, Sweden
Abstract - Educational outcomes for students belonging to disadvantaged social identities are unavoidably influenced by overlapping systems of inequity which arise along lines such as gender, ethnicity, and age. Robot platforms like Furhat require designers to select features which are interpreted by users as these same kinds of social identity. Prior work has posited that social robots might be intentionally designed to leverage these social identities in a “norm-breaking” fashion with the aim of disrupting social stereotypes in STEM education. However, research in HRI has been largely limited to the examination of gender only. We present a 2x2, between-subjects study in which 161 participants aged 9-12 are shown a robot-delivered lecture presented by a group of three separate robot personas with varying gender and ethnicity performances. We find that participants place greater trust in the persona groups with high gender diversity. Incorporating ethnic diversity seems to have little impact on our quantitative interaction metrics, however we do find evidence to suggest diversity in robots’ language capabilities may be important for trustworthiness. In all, the study contributes nuance to the discussions on the implications of (norm-breaking) social identity performance when using robots to pursue more equitable STEM education.