IEEE Robotics & Automation Magazine - June 2019 - 13

contingency. Of these learners, two responded less frequently
to the movement and lights reward than other rewards, while
nine of the 12 infants imitated the NAO robot during at least
one reward condition phase. These imitators displayed different learning rates and sometimes changed their behavior to
imitate less frequently during later reward conditions. The
infants were generally alert and nonfussy when interacting
with the robot. The parents of the participants found the
robot's reward featuring light and sound to be the most engaging for their children. To the best of our knowledge, this article
discusses the first interaction of socially assistive robots with
infants. As this new research area develops, our results are likely to inform continued work involving targeted robot-assisted
infant motion interventions.
Motivation
Socially assistive robotics (SAR) has demonstrated the potential to benefit a broad range of populations, from children
with developmental delays [1] to older adults facing health
challenges [2]. The research presented in this article introduces SAR in a new domain: that of teaching and reinforcing
infant motion. As an initial foray into socially assistive infant-
robot interaction, this research intends to gain novel insights
into such interactions and inform targeted robot interventions for infants.
Motor exploration and the practice of motions are essential components of infant development. Learning to perform
actions, e.g., grasping a favorite toy or kicking with knee
extension, has a substantial impact on infant cognitive and
motor development. Some infants, especially those at risk for
developmental delays, move an insufficient amount or practice nonoptimal movement patterns. This tendency can lead
to inadequate development of age-appropriate strength, proprioception, and coordination. A recent estimate determined
that approximately 9% of infants born in the United States are
at risk and could benefit from early targeted interventions [3].
Although early, frequent, and targeted interventions for atrisk infants may potentially improve these infants' neurodevelopmental function, the state of the art in motor
interventions for at-risk infants is infrequent, low-intensity
movement therapy [4]. Accordingly, this article introduces
SAR in the form of a humanoid, infant-sized robot intended
to teach, promote, and reinforce desirable types of infant
motion. The robot assists infants through noncontact interventions, as shown in Figure 1. Such SAR-supported therapy
is more scalable than the state of the art and could promote
infant motor practice more regularly in homes and other
accessible everyday environments.
Past infant behavior research points to contingency learning and imitation as two infant behaviors that could be leveraged to support robot-based motor interventions. Specifically,
infants as young as three months old may become motivated
to perform particular actions to gain a reward, such as the
motion and sounds of an overhead mobile [5]. Infants also
demonstrate the capacity to imitate, e.g., by producing motor
actions modeled by a parent [6]. As a first step toward

enabling SAR use for scalable infant motor interventions, we
attempted to understand whether behaviors such as contingency learning and imitation also apply to infant-robot interactions. If infants imitate
robot motion or robot rewards motivate infants to
The goal of SAR research is
move in particular ways,
then SAR holds promise
to use nonphysical human-
for targeted infant-robot
interventions. Before workrobot interactions to aid
ing with more vulnerable
at-risk infants, we conpeople with various needs.
ducted a study to learn
how infant contingency
learning and imitation behaviors observed in child-toy and child-adult interactions are
manifested in interactions between a typically developing infant and a robot.
In this article, we focus on answering the following
questions:
1) Do robot actions serve as a contingent reward for motivating infant leg motion?
2) Do infants imitate robot motions as defined by our proposed exploratory definition?
3) How do parents perceive child-robot interactions?
Answering these questions forms the necessary foundation
toward developing socially assistive infant-robot interactions
that improve neurodevelopmental outcomes.
Related Work
To evaluate the potential of SAR for infant motor interventions, we first consulted related literature on infant responses
to contingent rewards and infant imitation tendencies. We
reviewed work on infant interaction with adults and toys as
well as past studies of infant motor delays and infant-robot
interaction. In the following sections, we outline the background literature that motivated the design of our study,
explain the need for interventions in this space, and demonstrat the potential of SAR.

Figure 1. The infant-robot intervention scenario. An infant-sized,
socially assistive robot sits facing the infant and interacts with
the infant using noncontact modalities such as motion, light,
and sound.

JUNE 2019

IEEE ROBOTICS & AUTOMATION MAGAZINE

IEEE Robotics & Automation Magazine - June 2019

Table of Contents for the Digital Edition of IEEE Robotics & Automation Magazine - June 2019