← CHI 2026 map

CHI '26 · Honorable mention · full-paper review · confidence medium-high

Take the Dog to the Park: Quadruped Robot for Joint Attention Training with Autistic Children in Naturalistic Settings

Yuyang Fang , Yi Hu , Jiayu Teng , Yu Cai , Jiayang Liu , Feifan Xia , Yilin Tang , Liuqing Chen

DOI PDF Program page Page JSON

This is a credible and timely CHI paper because it moves a familiar autism-robot intervention into a more ecologically meaningful setting and backs that move with an exploratory study. The contribution is strongest as a design-and-feasibility advance, not as definitive efficacy evidence.

Axes Lens

Rare contribution shape, typical evidence profile. The point here is not a score. It is to show what kind of claim the paper makes, and whether the evidence pattern is unusual or baseline in this 268 -review set.

Review Summary

This paper’s main value is that it takes a well-established CHI/child-robot intervention pattern—robot-supported joint-attention training for autistic children—and pushes it into a setting that is meaningfully less artificial than the usual lab or classroom deployment. The abstract explicitly frames the field’s prior limitation as being “confined to stationary robots in indoor settings,” and the paper responds with a quadruped robot dog used across indoor and outdoor environments. That is a real departure in interaction setting and intervention framing, not just a cosmetic hardware swap. The novelty is therefore best understood as an interaction-technique and deployment innovation: the robot is positioned as a peer-like partner that directs attention to distributed targets and initiates JA trials through embodied co-exploration. The validation, however, is appropriately modest. The evidence packet supports a four-week pre-post exploratory study with six autistic children, plus a cue fidelity/clarity validation step and discussion of transfer to daily social communication. That is enough to support feasibility and initial promise, but not enough to justify strong causal claims or broad generalization. The limitations are important and well grounded: the sample is small, the outdoor environment is relatively controlled, Wizard-of-Oz operation was used for safety and fidelity, and camera occlusions forced the analysis to focus on discrete attentional events rather than continuous engagement. So the paper is strongest as an honorable-mention-style contribution because it opens a new design space for mobile, naturalistic robot intervention and provides early empirical support, while remaining careful about what the study can and cannot prove.

What Changed

Canon before

Prior CHI robot-assisted joint-attention work is described here as largely stationary and indoor, with limited support for generalization to naturalistic environments.

Departure from common sense

The paper departs from the common CHI pattern of stationary indoor robot interventions by moving joint-attention training into naturalistic indoor/outdoor settings with a mobile quadruped robot dog, aiming to support embodied co-exploration and broader generalization.

Actual novelty

The paper’s novelty is the integration of a mobile quadruped robot into joint-attention intervention in outdoor settings, with practice organized around embodied child-robot co-exploration and distributed targets rather than a purely static indoor setup.

Evidence

Evidence supports a concrete departure from stationary indoor robot interventions, a specific new intervention protocol using a quadruped robot dog in naturalistic settings, and an exploratory four-week pre-post study with six autistic children. The paper also explicitly reports limitations around sample size, Wizard-of-Oz control, outdoor variability, and occlusion-constrained behavioral analysis.

Limits

Method limits

The study is exploratory and pre-post in nature, with six participants, so causal inference and statistical generalization are limited. Behavioral analysis was constrained by camera occlusions and focused on discrete attentional events rather than continuous engagement metrics.

Deployment limits

Deployment is constrained by the need for Wizard-of-Oz control, safety and reliability concerns in dynamic outdoor environments, and the practical difficulty of recruiting and retaining families for repeated sessions.

Boundary conditions

Findings are bounded to a small sample of autistic children in relatively controlled outdoor pathways and to a protocol that can be operated with human supervision; transfer claims should be read as initial indications rather than broad proof.

Manual check: No additional manual check required for schema completion; the evidence packet is sufficient for a cautious expert review, though the underlying full text should still be consulted for any downstream publication decision.

Position in field

This work sits at the intersection of child-robot interaction, autism intervention, and mobile/field-deployed HCI, extending robot-supported joint-attention training beyond stationary indoor systems into naturalistic settings.

Abstract

Robot-supported interventions for joint attention (JA) in autistic children have shown encouraging outcomes, yet most remain confined to stationary robots in indoor settings, limiting opportunities for skill generalization and broader developmental benefits. We introduce an intervention that employs a quadruped robot dog as a peer-like partner for JA training across both indoor and outdoor environments. In this intervention, the robot dog directs children's attention to distributed targets in the environment and initiates JA trials. A four-week pre-post exploratory study with six autistic children demonstrated improvements in JA performance and indications of transfer to daily social communication. Spontaneous behaviors such as motor imitation (crawling) and novel social interactions with the robot also emerged, suggesting potential for broader developmental gains. These findings provide initial evidence for the efficacy of mobile robot-supported JA interventions in naturalistic contexts and offer implications for future design.