Towards automatically identifying contingent responses: differentiating toddlers based on human-robot interactions
This study characterizes baseline patterns of proxemic behavior in typically developing toddlers during a structured interaction with a social robot. Using overhead video and a convolutional neural network, we automatically identified the locations of the child, robot, and other individuals in the room and extracted distance-based measures throughout the interaction. These behavioral measures were analyzed alongside standardized developmental assessment scores to quantify variability within this normative sample. The combined data demonstrated substantial variability in children's proxemic behaviors, highlighting the range of ways young children engage with a novel robotic partner. These findings provide foundational descriptive data for early-childhood human-robot interaction and establish a methodological framework that future hypothesis-driven studies, including those involving clinically diagnosed populations, can build upon. No diagnostic or screening interpretations should be drawn from this work.
Authors
- Jed T. Elison (ORCID: https://orcid.org/0000-0002-6246-1000)
- Maria L. Gini (ORCID: https://orcid.org/0000-0001-8841-1055)
- Marie D. Manner (ORCID: https://orcid.org/0009-0008-9504-9902)
Institutions
- University of Minnesota (US)
Publication Details
- Journal
- ACM Transactions on Human-Robot Interaction
- Published
- 2026-11-20
- DOI
- https://doi.org/10.1145/3861331
- Primary Topic
- Social Robot Interaction and HRI
- Type
- article
- Field-Weighted Citation Impact
- 0.00
Funders
- National Science Foundation
- University of Minnesota