Eliciting curiosity — the intrinsic desire to explore and to find new information — enhances learning, promotes information-seeking and improves memory. Understanding how to elicit and maintain curiosity in students could help us design educational robots that stimulate interest, engagement and a quest for knowledge.
“We’re conducting a number of studies that examine human–robot interaction, in particular if a robot peer expresses curiosity does its inquisitiveness spark curiosity in students,” said Edith Law, a Professor at the Cheriton School of Computer Science and a member of its Human-Computer Interaction lab.
To explore the role of curiosity in learning, Professor Law and her research team conducted an experiment using a cooperative rock-classification game that a participant and an educational robot play together. Its goal was to see if curiosity could be fostered in participants through the verbal behaviour of a social robot.
“We created a matching game called LinkIt! to help teach how to categorize rocks,” said Jessy Ceha, a PhD student at the Cheriton School of Computer Science who is supervised by Dr. Law. “In front of a participant is a row of rocks, each with a card so the participant can see the rock’s type or features. The robot across from the participant also has a row of rocks with cards. The human player can see the robot’s rocks but not its cards.”
Between the players is a stack of cards placed face down that describes a feature or type of rock. For example, the card could show sedimentary rock — a type of rock formed by the accumulation of mineral or organic particles, or igneous rock — a type of rock that forms when magma cools and solidifies. The human player turns the top card over and tries to identify which rock on the robot’s side has the feature described in the card. Then the robot does the same while looking at the rocks on the human’s side. When rocks with the feature are identified correctly and linked up, they are put aside. The game continues until all rocks have been correctly matched.
L to R: Nalin Chhibber, a recent master’s graduate who was supervised by Dr. Law, helps PhD student Jessy Ceha set up cards and rocks to play LinkIt!, a cooperative rock-matching game, with NAO, a humanoid educational robot.
“The rock classification game is just the context in which we’re exploring human–robot interaction,” said Nalin Chhibber, a recent master’s graduate from the School of Computer Science. “Our study varied the extent to which the robot expresses verbal curiosity.”
“To
examine
this,
our
study
had
three
test
conditions,”
Jessy
added.
“One
was
called
neutral
and
in
that
condition
the
robot
did
not
express
any
verbal
curiosity.
The
second
condition
was
called
curious
and
there
the
robot
expressed
curiosity
through
on-topic
question-asking
behaviour
—
the
robot
was
actively
searching
for
new
knowledge
by
asking
questions
about
rocks
and
wanting
to
know
more
about
them.
The
third
condition
we
called
curious
plus
reveal,
and
here
the
robot
expressed
curiosity
through
on-topic
question
asking,
just
as
in
the
curious
condition,
but
it
also
revealed
to
the
participant
why
it
was
curious.”
“In
the
neutral
condition
the
robot
could
say,
‘Some
rocks
have
a
really
glassy
surface.
They
probably
cooled
too
quickly
to
form
any
crystals,’”
Nalin
said.
“In
the
curious
condition
the
robot
could
say,
‘I
am
curious
why
some
rocks
have
a
glassy
surface.
Could
they
have
cooled
too
quickly
to
form
any
crystals?’
And
in
the
curious
plus
reveal
condition
the
robot
could
say,
‘A
glassy
surface
is
new
to
me.
I
am
curious
why
some
rocks
have
that.
Could
they
have
cooled
too
quickly
to
form
any
crystals?’”
“By revealing its internal model — the robot giving some insight into what it could be thinking — might affect how curious participants are,” Jessy said.
Thirty participants at the University of Waterloo volunteered to take part in the study and they were randomly assigned to one of the three conditions. Ten participants played the rock-matching game with the robot in the neutral condition, ten with the robot in the curious condition, and ten with the robot in the curious plus reveal condition. Regardless of which group a participant was in, LinkIt! was played one-on-one — a single human player with the educational robot in one of the three conditions.
“We found that participants who interacted with the robot in the two curious conditions — curious and curious plus reveal — perceived the robot to be significantly more curious than participants who interacted with the neutral robot,” Jessy said. “This shows that we designed a robot that behaved in a way that was considered curious by participants. Most said that it was the on-topic question-asking that indicated to them the robot was curious.”
“We found that the robot’s expression of curiosity led to an increase in emotional curiosity — a significant increase in the state of feeling curious,” she added. “We also saw an increase in the behavioural expression of curiosity — a significant increase in on-topic question-asking and more verbal curiosity — in participants.”
Professor Law explains that they are interested in curiosity’s applications in many areas, not just in education.
“We want students to become curious about a subject because their interest leads to learning and the desire to find out more,” Professor Law said. “But there are many other applications of curiosity. If you have a new product and you want people to use it regularly — for example, a health-tracking app you want people to use so they report data to their doctor — understanding how to elicit and maintain curiosity are critical in keeping people interested. By designing curiosity into apps and devices the people who use them become and stay engaged.”
To learn more about the research on which this feature article is based, please see —
Jessy Ceha, Nalin Chhibber, Joslin Goh, Corina McDonald, Pierre-Yves Oudeyer, Dana Kulić, Edith Law. Expression of Curiosity in Social Robots: Design, Perception, and Effects on Behaviour. CHI '19: Proceedings of the 2019 CHI Conference on Human Factors in Computing Systems, May 2019. https://doi.org/10.1145/3290605.3300636
This research was funded by the Natural Sciences and Engineering Research Council of Canada, Discovery Grant RGPIN-2015-0454.