IEEE Robotics & Automation Magazine - June 2019 - 64

conversations (37%). Gratifying interactions included
expressing gratitude to and/or enjoyment interacting with
other community members. Deep conversations described
enriching interactions "involved in conversation" (P18)

1
Positive, Initial
Positive, Final
Negative, Initial
Negative, Final
Neither Negative
Nor Positive

Proportion of Responses

0.8

0.6

0.4

0.2

that could include discussing older adults' "experiences,
activities, usefulness, and wisdom" (P11), "sometimes very
deep, often funny, conversational, satisfying" (P11). The
participants associated negative emotions when interacting
with members of the community who are "brutal, rude,
negative, or noisy" (P1, P3). Some participants also pointed out the lack of compassion in themselves or others as a
factor that negatively influences the interaction: "the thing
that I hate in others, I hate in myself, which is acting with
little or no compassion or understanding of where they
come from" (P10).
The themes provide a context to understand older adults'
high-valued interactions and how such relationships contribute to community social connectedness. By aligning these
interactions with the types of engagements a social robot can
stimulate, we reasoned, we should be able to understand how
social robots contribute to community-level social connections and how a social robot's interaction features should be
designed to promote high-quality community bondedness.

Low Openness

Deny

Rude
Manner

Lacking
Compassion

Pleasantries

Shared
Interests

Helping
Others

Pleasantries

Deep
Conversation

Mid Openness

High Openness

Gratifying
Interaction

What Are Older Adult's Design
Preferences for Social Robots?
0
Older adults' design preferences for a social robot were compared over time. We report older adults' preferences at the
beginning and end of the study through quantitative and
qualitative analysis. A preliminary analysis of the transcripts
from the design kit sessions revealed different opinions
related to participants' openness toward the technology. To
Figure 5. The positive and negative experiences associated with
further investigate this initial finding, participant groups
interactions in the community.
were identified based on their openness on their design kit selections
0.2
0.6
0.8
0.4
(their rate of yes responses) through a
clustering approach. Three clusters
were determined through k-means
clustering for the preinteraction
Initial Avg = 0.93 Avg = 0.56 Avg = 0.84 Avg = 0.97 Avg = 0.83
(n = 5) SD = 0.14
SD = 0.19
SD = 0.41
SD = 0.06 SD = 0.52
responses, and we termed them lowopenness, mid-openness, and highFinal Avg = 0.54 Avg = 0.85 Avg = 0.91 Avg = 0.93 Avg = 0.80
openness clusters. To track how the
(n = 6) SD = 0.26
SD = 0.16
SD = 0.07
SD = 0.22 SD = 0.06
participants in each cluster shifted
from the initial to the final interview,
Initial Avg = 0.48 Avg = 0.83 Avg = 0.50 Avg = 0.54 Avg = 0.38
we assigned older adults' final data to
(n = 4) SD = 0.27
SD = 0.12
SD = 0.48
SD = 0.19 SD = 0.14
the closest cluster using root-meansquare distance. There was some
Final Avg = 0.38 Avg = 0.92 Avg = 0.73 Avg = 0.83 Avg = 0.68
shift of participants among clusters
SD = 0.09
(n = 3) SD = 0.08
SD = 0.07
SD = 0.14 SD = 0.20
between the initial and final sessions,
but all three clusters were present in
Initial Avg = 0.21 Avg = 0.13 Avg = 0.32 Avg = 0.41 Avg = 0.34
both conditions.
(n = 4) SD = 0.12

SD = 0.06

SD = 0.20

SD = 0.20

SD = 0.21

Final Avg = 0.14
(n = 4) SD = 0.12

Avg = 0.13
SD = 0.14

Avg = 0.27
SD = 0.20

Avg = 0.34
SD = 0.26

Avg = 0.31
SD = 0.13

Mediating
Robot
Suggestions Reminders
Connections Creating and
and Alerts
With People Sharing Items

Information

Figure 6. A comparison of the rate of positive answers to items in each category between
the initial and final interviews reveals different preferences across participant openness
clusters. Avg: average; SD: standard deviation.

64

*

IEEE ROBOTICS & AUTOMATION MAGAZINE

*

JUNE 2019

Quantitative Results
The three clusters exhibited different
trends in openness between the initial
and final interviews (Figure 6). Before
the study, low-openness participants
(n = 4) were the least open of the
clusters, particularly rejecting functions related to "robot-mediated
IEEE Robotics & Automation Magazine - June 2019

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