A Model Child? Behavior Models for Simulated Infant-Robot Interaction

Download

• Paper
• Github

Abstract

Simulated child-robot interaction offers ways to test robot behaviors before real-world trials with vulnerable populations. At the same time, this type of simulation requires realistic models of child behavior. We combined cognitive science research on infant attention with real-world child-robot interaction data to develop two behavior tree-based models of child behavior (i.e., robot-interested and robotuninterested). We evaluated these models through a video-based study (N = 60). Participants rated the proposed models as more familiar, humanlike, and natural than a control (random behavior) model. This work can support related work on child-robot interaction, as well as broader efforts on using technology to support infant development.

Citation

Helmi, Ameer and Koenig, Kristen and Fitter, Naomi T. “A Model Child? Behavior Models for Simulated Infant-Robot Interaction” Proceedings of the 2022 International Conference on Social Robotics (ICSR)

@inproceedings{helmi2022model,
  title={A Model Child? Behavior Models for Simulated Infant-Robot Interaction},
  author={Helmi, Ameer and Koenig, Kristen and Fitter, Naomi T.},
  booktitle={Proceedings of the International Conference on Social Robotics (ICSR)},
  pages={3--12},
  year={2022},
  organization={Springer}
}