This paper describes an experiment on peer groups that had a robot as a learning partner, to examine whether the robot could encourage the participants to talk on their own initiative. The authors measured the number of proactive utterances of each participant during the sessions. The authors compared the experimental groups that had robot facilitators, which were manipulated by professional human facilitators, and the control groups, which were also led by professional human facilitators but without a robot. The result showed that the participants in the experimental sessions talked on their own initiative much more than those in the controlled sessions. Finally, the authors qualitatively examined the characteristics of the proactive utterances in the peer group and found that the utterances contained supportive responses, which encouraged the participants to voluntarily join the dialogue promoting the counseling.
Robot as a Learning Partner for Promoting Proactive Discussion in Peer Groups: A Case Study for Career Development
1. Robot
as
a
Learning
Partner
for
Promo1ng
Proac1ve
Discussion
in
Peer
Groups:
A
Case
Study
for
Career
Development
Toshio
Mochizuki,
Senshu
Univ.,
Japan
Yoshitaka
Mitate,
The
University
of
Kitakyushu,
Japan
Yoshikazu
Tateno,
Jun
Nakahara,
&
Naomi
Miyake,
The
University
of
Tokyo,
Japan
Takehiro
Wakimoto,
Aoyama
Gakuin
University,
Japan
Yuko
Miyata,
Freelance
Career
Consultant 1
3. Robots
as
learning
partner
-‐
an
emerging
agenda
for
CSCL
!
• Two
reasons
to
apply
robots
as
social
partner
in
learning
(Miyake
&
Okita,
2012;
Okita
et
al.,
2010)
– Robots
have
human-‐like
appearance
and
behavior
that
can
elicit
social
responses
that
invite
acSve
engagement
• Not
too
human-‐like
and
sSll
machine-‐like.
• This
gives
people
room
for
imaginaSon
and
creaSvity
in
social
interacSon
and
elicits
greater
empathy.
!
– A
robot
can
be
an
interface
to
collect
the
process
data
of
collaboraSve
learning.
• In
order
to
analyze
the
mechanisms
and
design
the
principles
of
producSve
learning 3
4. Past
research
of
using
robots
as
learning
partners
–
structured
seEngs
• Learning
English
with
children
(Kanda
et
al.,
2004)
– A
certain
social
and
emoSonal
support
was
required
in
order
to
enhance
collaboraSon
between
children
and
robots
!
• Use
of
Honda
ASIMO
in
a
variety
of
se[ngs
(Okita
et
al.,
2010)
– CooperaSve
interacSon
with
more
human-‐like
voice
and
gestures
was
effecSve
to
make
children
engage.
4
5. Past
research
of
using
robots
as
learning
partners
–
structured
seEngs
• A
scripted
collaboraSve
learning
based
on
Jigsaw
(Miyake,
2012)
– EffecSve
if
children
recognize
a
robot
as
“just
like
the
other
kid
who
does
not
know
the
answer,
but
sincerely
working
to
know
the
answer”
!
• Jigsaw-‐based
reciprocal
teaching
(Oshima
&
Oshima,
2013)
– Robots
can
work
as
well
as
human
facilitator.
5
6. Research
Agenda
• The
potenSal
of
robot
facilitaSon
in
ill-‐structured
se[ngs
– With
ore
self-‐regulated
student
discussions
in
order
to
learn
what
the
students
do
not
know
in
the
absence
of
educaSonal
materials.
!
• To
find
out
whether
it
creates
an
environment
to
encourage
parScipants
to
talk
on
their
own
iniSaSve.
– A
case
study
in
career
development
• Career
development
is
a
serious
learning
agenda
for
adolescents
in
every
country
– No
hierarchical
difference
between
robots
and
young
parScipants
– The
robots’
appearance
is
more
neutral
than
that
of
human
career
counselors
6
7. Research
Method
and
Design
(1)
7
• Design
– Experimental
study
design
student
student
student
student
Career
counselor
robot
Remote
control
Exp.
Control
Career
counselor student
student
8. Research
Method
and
Design
(2)
• ParScipants
– Undergraduate
students
(sophomores
and
juniors)
in
a
private
university
in
Tokyo
!
– Professional
facilitators
for
career
development
• More
than
5-‐year
experience
of
group
counseling.
– Desktop
Rovovie-‐W
robot
8
9. Research
Method
and
Design
(3)
• Design
– Experimental
study
design
9
student
student
student
student
Career
counselor
robot
Remote
control
Exp.
Control
Career
counselor student
student
10. 10
Robot
Operated
by
Remote
Human
Facilitator
Professional
Human
Facilitator
Present
11. Research
Method
and
Design
(3)
• Discussion
theme:
– “So,
please
tell
us
about
your
career
goals
for
the
future,
Mr.
B.”
– “How
do
you
plan
to
use
the
coming
summer
vacaSon
to
prepare
for
your
career
goals?”
– “Your
plan
is
(summary
of
what
Mr.
B.
says)”;
“What
do
you
think
of
his
goals,
Mr.
A?”
– “Do
you
all
have
any
suggesSons
for
preparing
for
Mr.
B’s
career
goals?
“ 11
Step Experimental
Group Control
Group
1 Pre-‐quesSonnaire
2 Ice-‐breaking
(10
min.)
3 IntroducSon
for
group
counseling
(5
min.)
4 Group
Counseling
Session
(30
min.)
5 Post
interview
(30min.)
6 Post-‐quesSonnaire
12. 12
Hello,
I’m
Robovie.
Nice
to
meet
you
all.
First
of
all,
I
would
like
to
start
this
session
by
introducing
ourselves
to
each
other.
Please
tell
your
name,
the
place
where
you
come
from,
and
your
goal
in
this
session.
We
have
10
minutes
from
now....
13. Research
Method
and
Design
(3)
• Discussion
theme:
– “So,
please
tell
us
about
your
career
goals
for
the
future,
Mr.
B.”
– “How
do
you
plan
to
use
the
coming
summer
vacaSon
to
prepare
for
your
career
goals?”
– “Your
plan
is
(summary
of
what
Mr.
B.
says)”;
“What
do
you
think
of
his
goal,
Mr.
A?”
– “Do
you
all
have
any
suggesSons
for
preparing
for
Mr.
B’s
career
goal?
“ 13
Step Experimental
Group Control
Group
1 Pre-‐quesSonnaire
2 Ice-‐breaking
(10
min.)
3 IntroducSon
for
group
counseling
(5
min.)
4 Peer
Group
Counseling
Session
(30
min.)
5 Post
interview
(30min.)
6 Post-‐quesSonnaire
14. Data
collec1on
• QuesSonnaires
(pre
and
post)
– to
examine
the
results
were
significantly
different
– the
General
Self-‐Efficacy
Scales(GSES)(Sakano
&
Tojo,
1989)
– EffecSveness
of
peer
support
can
be
explained
by
individual’s
improved
self-‐efficacy
(Bandura,
1997;
Benight
&
Bandura,
2004)
• Video
of
discussion
during
the
peer
group
acSvity
– to
examine
how
students
parScipated
in
the
discourse
• Post
group
interview
(video-‐recorded
&
transcribed)
– To
examine
feelings
about
the
peer
group
experience.
14
19. Differences
of
Students’
Par1cipa1on
in
the
Peer
19
23.5%
21.3%
17.9%
21.9%
79.3%
6.4%
17.2%
10.2%
0% 20% 40% 60% 80% 100%
FP/p%
FP/a%
PP/p%
PP/a%
Experimental% Control%
(S.D.=14.8)
(S.D.=15.1)
(S.D.=4.7)
(S.D.=16.7)
(S.D.=19.7)
(S.D.=20.1)
(S.D.=6.6)
(S.D.=34.9)
***
***
*
**p<.001,
*p<.05
Human
facilitators
could
speak
fluently
and
use
their
acSve
listening
strategy
The
robots
lacked
fluency
because
the
operators
could
only
input
messages
using
the
keyboard
aser
understanding
the
parScipants’
uterances
F P
prompt
20. Differences
of
Students’
Par1cipa1on
in
the
Peer
20
23.5%
21.3%
17.9%
21.9%
79.3%
6.4%
17.2%
10.2%
0% 20% 40% 60% 80% 100%
FP/p%
FP/a%
PP/p%
PP/a%
Experimental% Control%
(S.D.=14.8)
(S.D.=15.1)
(S.D.=4.7)
(S.D.=16.7)
(S.D.=19.7)
(S.D.=20.1)
(S.D.=6.6)
(S.D.=34.9)
***
***
*
**p<.001,
*p<.05
F P
The
parScipants
spoke
to
the
robot
facilitator
much
more
than
they
did
to
the
human
facilitator.
21. Differences
of
Students’
Par1cipa1on
in
the
Peer
21
23.5%
21.3%
17.9%
21.9%
79.3%
6.4%
17.2%
10.2%
0% 20% 40% 60% 80% 100%
FP/p%
FP/a%
PP/p%
PP/a%
Experimental% Control%
(S.D.=14.8)
(S.D.=15.1)
(S.D.=4.7)
(S.D.=16.7)
(S.D.=19.7)
(S.D.=20.1)
(S.D.=6.6)
(S.D.=34.9)
*
**p<.001,
*p<.05
The
parScipants
in
the
experimental
groups
parScipated
in
the
discussion
much
more
proacSvely
than
in
the
control
groups,
even
with
fewer
prompts
from
the
robot
facilitators.
**p<.001,
*p<.05
F P
P
P
22. Characteris1cs
of
Proac1ve
USerances
in
the
Peer
Groups
• To
clarify
how
the
parScipants
proacSvely
spoke
in
the
experimental
groups
– Whether
the
peer
groups
produced
a
posiSve
mood
that
encouraged
the
members
to
disclose
what
they
thought
(Paine
et
al.,
1989)
• AcSve
listening
skills
for
proacSve
discussion
– clarifying,
paraphrasing,
and
summarizing
• given
in
the
scenario
in
both
group
– SupporSve
responses
• But
difficult
for
the
robot
operator
to
react
smoothly
by
using
supporSng
response.
22
23. Characteris1cs
of
Proac1ve
USerances
in
the
Peer
Groups
23
The
facilitator
should
try
to
elicit
supporSve
responses
from
other
members
during
the
self-‐help
group
session
(Paine
et
al.,
1989)
24. Characteris1cs
of
Proac1ve
USerances
in
the
Peer
Groups
24
There
were
significantly
many
supporSve
responses
from
the
parScipants
in
the
experimental
groups
and
from
the
facilitators
in
the
control
groups.
(χ2(1)
=
69.664,
p
<
.01),
The
facilitator
should
try
to
elicit
supporSve
responses
from
other
members
during
the
self-‐help
group
session
(Paine
et
al.,
1989)
Uh huh
25. Conclusion
• Both
experimental
and
control
peer
groups
produced
similar
results
with
regard
to
self-‐efficacy
improvement
• The
parScipants
in
the
experimental
groups
talked
on
their
own
iniSaSve
during
the
sessions.
• The
parScipants
in
the
experimental
groups
used
supporSve
responses
much
more
than
those
in
the
control
groups
did.
25
The
robot
has
the
potenSal
to
create
a
similar
effecSve
peer
discussion,
and
a
more
parScipant-‐centered
proacSve
discussion