The PInSoRo dataset: Supporting the data-driven study of child-child and child-robot social dynamics

Area of Science:

• Developmental Psychology
• Social Psychology
• Human-Robot Interaction (HRI)

Background:

• Understanding fine-grained social dynamics in children is crucial for psychology and increasingly for Human-Robot Interaction (HRI).
• Robots in healthcare and education require socially contingent behaviors for long-term engagement.
• Generating sustained, engaging social behaviors in robots remains an open research challenge.

Purpose of the Study:

• To introduce a novel, open dataset (PInSoRo) for data-driven research on child social dynamics.
• To facilitate machine learning approaches for analyzing and synthesizing complex social interactions.
• To support the development of robots capable of meaningful, long-term social engagement with children.

Main Methods:

• Collected over 45 hours of hand-coded recordings from child-child and child-robot free-play interactions.
• Utilized an engaging, methodologically sound, yet underspecified free-play paradigm to capture natural behaviors.
• Dataset includes multi-modal data: calibrated video, 3D facial recordings, skeletal information, audio, and game interactions.

Main Results:

• The PInSoRo dataset captures a rich variety of behavioral patterns in natural child social interactions.
• Provides detailed annotations of social constructs alongside comprehensive multi-modal recordings.
• Enables data-driven research into the complexities of social dynamics between children and with robots.

Conclusions:

• The PInSoRo dataset is a valuable resource for advancing research in developmental psychology, social psychology, and HRI.
• Facilitates the development of AI systems that can better understand and engage in social interactions with children.
• Advances the goal of creating robots with appropriate and meaningful long-term social behaviors.