IEEE Robotics & Automation Magazine - June 2019 - 62

people at home as a helpful companion. It is a tabletop
robot standing 11 in tall, with three degrees of freedom, a
voice user interface, and a touchscreen face. It offers a range
of skills, e.g., music, general question and answer, news,
weather, photos, and contents that highlight its personality,
such as dancing, jokes, and answers to personal questions,
such as "Hey Jibo, what is your favorite movie?" Its companion-like features include being attentive to the user, taking naps and going to sleep at night, and proactively
greeting people based on face detection. For three weeks,
Jibos were set up in the common space area on each of the
four floors, where participants and other community members could freely interact with the robot on Monday
through Friday from 7 a.m. to 4 p.m. (Figure 2). Multiple
research methods were incorporated to provide a holistic
approach to understanding the social connections in the
community. This article will focus on the three methods
utilized in the study: 1) interviews, 2) design kit workshops,
and 3) common space area observations.

(a)

(b)

(c)
Figure 2. (a) The larger common space area where the social
robot was placed during the day. (b) The social robot setup
with feedback forms for participants to note their thoughts.
(c) Two participants interacting with the social robot in the
common space.

IEEE ROBOTICS & AUTOMATION MAGAZINE

JUNE 2019

Observations in Common Space Area
Researchers observed the community members' interactions
with the social robot three times a day on each floor: in the
morning, at midday, and in the afternoon. In addition, the
number of people in the common areas of the residential
floors at those times was recorded.
Interviews
Participants engaged in two interviews over the course of the
study, one at the beginning and another at the end. The semistructured interview protocol was kept concise to help prevent tiring or misunderstanding among participants [15].
Each session was piloted by three researchers. The questions
focused on how residents start interactions with each other,
the easy and hard parts of relating with one another, and
people's characterization of their interfacing with each other.
The end-of-study dialogue also included questions on the
effect the social robot may or may not have had on the community. At the conclusion of each interview, participants
completed a one-question survey describing their relationship with the community (the Inclusion of Community in
Self Scale [16]).
Design Kit
As part of the participatory design activity, residents
described their preferences for social robot functions through
a design kit [14]. For this study, five categories were tested
with participants:
1) the robot's ability to mediate connections with people (e.g.,
sending a text message or e-mail to someone)
2) the robot's capacity to create and share things (e.g., a playlist)
3) the ability of the robot to make suggestions (e.g., for eating
healthily or reading a good book)
4) the capability to provide reminders and alerts (e.g., events
or medication)
5) the robot's ability to offer information (e.g., news or weather).
Each category contained between four and 13 function
cards that participants categorized as "yes," "maybe," or "no"
based on whether they desired the function or not. The sessions were audio- and video-recorded.
Data Analysis
Audio and video recordings of the interviews and design kit
sessions were transcribed and reviewed for accuracy. The first
analysis focused on using an inductive approach to identify
themes surrounding older adults' preferences for social robot
technology [17]. To determine the interrater reliability among
researchers, Cohen's kappa [18] was calculated for the initial
and final interviews and design kits for 66% of the coded
transcripts (six initial interviews, six final interviews, six initial design kits, and 12 final design kits), coded independently
by a team member after an initial coding by other team members. Cohen's kappa was 0.86 for the initial interviews, 0.74 for
the final interviews, 0.63 for the initial design kits, and 0.67
for the final design kits-within the range for the substantial
agreement considered acceptable for interrater reliability [18].

IEEE Robotics & Automation Magazine - June 2019

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