Abstract:

Building upon previous work of our research group, which introduced the Qui-Bot platform and validated its gender-neutral engagement among 212 children aged 3-18 (98.7% satisfaction, 91% sustained interest) [1], this research protocol advances the framework by proposing a feminist Human-Robot Interaction (HRI) architecture. Unlike our previous progress report, which described the platform and pilot results, this paper presents a formal experimental protocol for isolating the causal effect of robot voice on gender stereotype disruption. Our earlier study confirmed that humanoid robots appeal equally to girls and boys but did not experimentally probe voice as a causal mechanism. Informed by research on the feminization of AI voice assistants [2] and studies showing that gender stereotypes consolidate by age 6 [3], this protocol addresses that gap.

Methodology:
This protocol is designed for children from early childhood through upper secondary education (ages 3-18), with initial implementation targeting early childhood in collaboration with partner schools. For this age range, the experience will be delivered in eight 60-minute group sessions guided by an educator, with video recording for later analysis. For older participants, self-reported Likert-scale measures will complement observational data. We operationalize feminist theory into three design dimensions. First, morphology employing organic curves and transparent casings to demystify internal electronics. Second, a non-gendered color palette (orange, dark blue, and white semi-transparent finishes) that rejects pink while critically reappropriating traditionally masculine-coded colors. Third, an experimental protocol wherein the humanoid robot will utilize three synthesized voice profiles (feminine, masculine, and gender-neutral) alongside three distinct response strategies—confrontational, rational/educational, and collaborative—to handle sexist comments. Design: 3 (voice) × 3 (strategy) between-subjects, targeting a minimum of 72 children (8 per condition, balanced by gender). Dependent variables: change in Implicit Association Test score for gender-science stereotypes (pre/post), hesitation time after robot correction, and post-interaction free-play choice. Statistical analysis: two-way ANOVA. A qualitative analysis will examine whether children derive enjoyment from provoking the robot, and its interaction with voice condition.

Expected results:
We hypothesize that feminine voice combined with confrontational strategy will produce the largest reduction in Implicit Association Test (IAT) scores. Gender-neutral voice is expected to produce moderate but consistent effects. Masculine voice may show no reduction or potential backfire effects.

Conclusions:
This protocol shifts focus from descriptive outcomes to causal mechanisms, providing a methodological blueprint for building equitable robotics from the ground up. Aligned with UN Sustainable Development Goals 4, 5, and 9, the framework demonstrates how feminist theory can be operationalized into testable HRI specifications. All materials are open-source and available at quibot.upc.edu.

Acknowledgement:
Supported by the Generalitat de Catalunya - Departament d'Igualtat i Feminisme under Grant IFE038/24/000087.

References:
[1] Tarrés-Puertas, M.I. et al. Sensors 2023, 23, 1675. doi.org/10.3390/s23031675
[2] Sindoni, M.G. Discourse, Context & Media 2024, 62, 100833. doi:10.1016/j.dcm.2024.100833
[3] Bian, L. et al. Science 2017, 355(6323), 389-391. doi:10.1126/science.aah6524

Keywords:

Feminist HRI, Educational Robotics, Gender Stereotypes in STEM, Child-Robot Interaction, Inclusive Design, Early Childhood STEM, Chemistry Robotics.