21735
The Problems with “Look, Don't Touch”: Designing Robots to Maximally Benefit Young Children with ASD
Despite the fact that the first five years of a child’s life are foundational for their social, emotional, behavioral, cognitive and communicative development, a paucity of robots exist that can withstand the physical interactions and play through which very young children naturally learn about the world. A common approach has been to employ expensive off-the-shelf robots with numerous degrees of freedom and sensors, many of which may be functionally unnecessary to achieve the end goal for which they are employed. This is deleterious to productive human-robot interaction (HRI) in two ways. First, child-robot interactions are restricted because children are either physically separated from the robot or are instructed not to touch the robot. Second, robots with many moving parts and behaviors can add unnecessary complexity, particularly for highly-targeted interventions with children who have behavioral, social and cognitive difficulties. In this work, we present 3 low-cost ($100- $300USD), field-tested robots and their fundamental design features that may help inform development of other robust and minimalistic robots suitable for use with very young populations of children with autism spectrum disorder (ASD).
Objectives:
To identify typical physical interactions that lead to hardware failures, describe the evolution of key design features and resulting design modifications to each of three robots used with individual groups of young children. Further, to evaluate the efficacy of three low-cost, robot prototypes that may help inform the development of other robust and minimalistic robots suitable for use with young populations of children with ASD.
Figure 1. Robots: CHARLIE, L-E and Sphero
Methods:
We analyzed child-robot interactions in each of three studies conducted with different subpopulations of children with ASD. Participant mean age for each study was 4.83 (CHARLIE, n_participants=8, n_sessions=96), 7.91 (L-E, n_participants=15, n_sessions=15) and 2.50 (Sphero, n_participants=12, n_sessions=12) years. Further, we summarized characteristics of interactions that lead to mechanical or hardware failure and identified specific design features that promoted maximal opportunities for unrestricted interactions and productive child-robot interaction. Finally, we evaluated the viability of low-cost robots with robust design features using empirical results obtained for each study.
Results:
Three generalizable, fundamental design features that directly contributed to promoting each study’s aims resulted: (1) Simplified interface, with minimal DOF, and behaviors, (2) Securable base paired with break-away and/or compliant parts and, (3) Self-contained and impact-resistant. While low-cost and simple in design, these robot prototypes assisted in achieving significant improvements in speech (Mean Length Spontaneous Utterances), receptive language, interpersonal skills and play (Vineland Adaptive Behavior Scales – Parent/Caregiver Form). Additionally, one robot (Sphero) was used in unstructured scenarios to elicit and measure statistically significant differences in how typically developing children and a group of children with ASD play with a robot.
Conclusions:
Optimal HRI for young children requires robust robots. Simple, low cost design features can be easily integrated or added to robots employed to assist very young children with ASD while still contributing to significant improvements in several measures of communication and social skill proficiency.