Exploring Key Challenges in Child-Robot Interaction Using Haru4Kids: Engagement, Language Understanding, and Privacy

Abstract

Abstract

This chapter examines three critical challenges in long-term child-robot interaction in the home, once the novelty effect has faded away: engagement, language understanding, and privacy concerns. The study used the Haru4Kids (H4K) platform, a child-oriented family robot simulator that features a rotating iPad-based interface offering interactive activities. This platform facilitated sustainable child-robot interactions in family environments, as it offered an interactive platform while also allowing us to assess user engagement and behavior throughout each interaction. Over two weeks, seven families in Southern Spain cohabitated with H4K. The study provides comprehensive insights into user engagement by integrating user logs, annotated images, and verbal interaction analysis. The image-based engagement was assessed via an innovative Engagement Level Metric, which we used to estimate which activities the children found most engaging. A natural language processing analysis revealed that mixed-initiative dialogs enhanced user agency over time, shifting interactions from system-driven to user-driven. Privacy concerns varied between children and parents, with children showing more hesitancy toward third-party data sharing. This chapter offers valuable design recommendations for future child-robot interaction platforms, emphasizing personalization, transparent data practices, and diverse activity offerings. For researchers and developers, it underscores the importance of addressing dynamic and multi-modal engagement and privacy concerns in realistic, unsupervised settings.