Information Seeking Behaviour in Collaborative and Competitive Child-Robot Interactions

Children have an impressive capacity for active, strategic information seeking in noisy and dynamic environments—an evolutionary adaptive skill that is increasingly relevant with the rise of digital and robotic educational technologies. This study investigated how collaborative versus competitive reward structures influence children's information-seeking behaviour during word learning with a social robot partner. A total of 140 children aged 6–10 interacted with a humanoid robot in a gamified learning task, where they learned and recognized novel object labels under either a collaborative or competitive framework. Results showed robust object-label mapping and flexible adaptation in information sampling aligned to the reward structure; children preferentially sampled the most informative object, with optimal sampling rates increasing with age. Strategic sampling was evident across conditions, but performance was modulated by children's understanding of the reward rules, which was lower in the competitive scenario due to increased task complexity. The findings extend ecological active learning frameworks to embodied, socially interactive settings, underscoring children's ability to tune their learning strategies to both social context and environmental affordances. These insights highlight the need for educational technologies that transparently scaffold reward structures to maximize goal-directed learning.