IEEE Robotics & Automation Magazine - June 2019 - 21

study activity's challenge level were positive. In general,
parents thought the infants were happy, excited, successful,
and engaged during their interactions with the robot. The
ratings of infant motion levels and imitation levels were
closer to neutral but, on average, were still above the neutral rating.
Each parent also shared his or her perceptions of the
infant's engagement and fun levels during each reward. Using
a one-way rANOVA with an a = 0.05-significance level, we
observed differences in parental perceptions of infant engagement (i.e., F(2, 18) = 3.87, p = 0.042, and h = 0.233) and
fun levels (i.e., F(2, 16) = 7.74, p = 0.005, and h = 0.375)
across reward conditions. As illustrated in Figure 7,
post hoc Tukey multiple comparison tests showed that parents found the movement and sound reward to be more
engaging and fun for their infants than the reward involving
only motion.
Discussion
The study results enabled us to evaluate our hypotheses about
how infants respond to robot-delivered contingent rewards,

whether infants imitate humanoid robots, and how parents
perceive the infant-robot interaction.
Hypothesis Testing
The results of our study support H1; the majority of infants
learned the contingency presented by the robot as we evaluated using benchmarks from past contingent learning research
(not involving robots). A statistical test of differences similarly
revealed that infants moved above the acceleration threshold
significantly more in all of the reward phases as compared to
the baseline phase, when the robot was stationary. This result
supports our use of precedents in contingency learning as one
essential component of our infant-robot interaction design
and evaluation. The knowledge that contingent rewards from
a robot can encourage infants to move more than they normally would is an essential building block for infant-robot
SAR studies.
Infant behaviors during the study provide some support
for H2. Our analysis showed that nine of the 12 infants
behaved in a way that fulfilled our definition of imitation during at least one reward phase. Changes in infant behavior over

Table 4. Parental responses to additional survey questions from the poststudy assessment.
Number of Recorded Response
Responses
Mean

Response
Standard Deviation

Lower
Anchor Point

Upper
Anchor Point

How happy or unhappy do you
think your baby felt? (SAM valence
images)

0.82

Unhappy

Happy

How calm or excited do you think
your baby felt? (SAM arousal images)

3.6

0.7

Calm

Excited

How controlled or in control do you
10
think your baby felt? (SAM dominance
images)

3.6

0.7

Controlled

In control

Do you think your baby learned
what he/she had to do?

3.8

0.79

Never

Always

Do you think your baby moved
more than he/she typically does?

3.3

1.25

Never

Always

Do you think your baby was
engaged with/focused on the robot?

0.94

Never

Always

Do you think your baby was
mimicking the robot?

3.44

0.88

Never

Always

Do you think your baby enjoyed the
activity?

4.11

0.6

Never

Always

Would you change the difficulty of
the activity in future sessions?

Much lower

Much higher

How often does your baby play with
a mobile?

1.89

1.17

Never

Always

Do you think social robotics can be
useful in improving the well-being of
babies?

4.11

1.05

Never

Always

Have you had previous experiences
with social robots?

Never

Always

Would you participate in future
infant-robot interaction studies?

4.56

0.53

Never

Always

Question

SAM: Self-Assessment Manikin.

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