Once Upon a Story: Can a Creative Storyteller Robot Stimulate Creativity in Children?

Full text

Once Upon a Story: Can a Creative Storyteller Robot Stimulate
Creativity in Children?
Maha Elgarf
KTH Royal Institute of technology
Stockholm, Sweden
mahaeg@kth.se
Gabriel Skantze
KTH Royal Institute of technology
Stockholm, Sweden
gabriel@speech.kth.se
Christopher Peters
KTH Royal Institute of technology
Stockholm, Sweden
chpeters@kth.se
Figure 1: Children were asked to collaborate with the robot to tell a story together using the software presented to them on
the touch screen. Consent was received from the parents for publishing children’s images.
ABSTRACT
Creativity is a vital inherent human trait. In an attempt to stimulate
children’s creativity, we present the design and evaluation of an in-
teraction between a child and a social robot in a storytelling context.
Using a software interface, children were asked to collaboratively
create a story with the robot. We conducted a study with 38 children
in two conditions. In one condition, the children interacted with
a robot exhibiting creative behavior while in the other condition,
they interacted with a robot exhibiting non creative behavior. The
robot’s creativity was dened as verbal and performance creativity.
The robot’s creative and non creative behaviors were extracted
from a previously collected data set and were validated in an online
survey with 100 participants. Contrary to our initial hypothesis,
children’s creativity measures were not higher in the creative con-
dition than in the non creative condition. Our results suggest that
merely the robot’s creative behavior is insucient to stimulate
creativity in children in a child robot interaction. We further dis-
cuss other design factors that may facilitate sparking creativity in
children in similar settings in the future.
Permission to make digital or hard copies of all or part of this work for personal or
classroom use is granted without fee provided that copies are not made or distributed
for prot or commercial advantage and that copies bear this notice and the full citation
on the rst page. Copyrights for components of this work owned by others than ACM
must be honored. Abstracting with credit is permitted. To copy otherwise, or republish,
to post on servers or to redistribute to lists, requires prior specic permission and/or a
fee. Request permissions from permissions@acm.org.
IVA ’21, September 14–17, 2021, Virtual Event, Japan
©2021 Association for Computing Machinery.
ACM ISBN 978-1-4503-8619-7/21/09. . . $15.00
https://doi.org/10.1145/3472306.3478359
CCS CONCEPTS
•Computer systems organization →Robotics
;
•Human-centered
computing →Interaction design process and methods
;
User
studies.
KEYWORDS
Creativity, social agents, social robotics, human robot interaction,
storytelling
ACM Reference Format:
Maha Elgarf, Gabriel Skantze, and Christopher Peters. 2021. Once Upon a
Story: Can a Creative Storyteller Robot Stimulate Creativity in Children?.
In 21th ACM International Conference on Intelligent Virtual Agents (IVA ’21),
September 14–17, 2021, Virtual Event, Japan. ACM, New York, NY, USA,
8 pages. https://doi.org/10.1145/3472306.3478359
1 INTRODUCTION
Creativity is regarded as a crucial skill in the 21st century. Some
authors argue that it is a survival skill that has laid the founda-
tion for the current structure of our modern world [
30
]. As the
world is currently shifting from industrial to creative economies
[
33
], creativity is becoming a vital skill for an innovative working
environment. Hence, arises the signicance of developing creativity
skills since a very young age. Researchers suggest the possibility
of augmenting and nourishing creative abilities throughout life
[
36
]. Unfortunately, a creativity crisis occurs as children move from
kindergarten to elementary school [
25
] [
15
] [
40
] [
39
]. Children’s
creative abilities are known to plunge at that age because of the
structure of current educational systems that constraints children’s

IVA ’21, September 14–17, 2021, Virtual Event, Japan Elgarf, et al.
expressivity and encourages them to conform to society norms.
Scholars have attributed great attention to studying behavioral syn-
chronisation in human agent interaction. In a typical encounter
between an agent and a human, humans frequently tend to replicate
the agent’s behavior. For example, in [
38
], robots induced mimicry
in human subjects. After observing the robot putting hands on
its hips and hands behind its back, participants subconsciously
repeated the same gestures. Participants in [
21
], mimicked facial
expressions of a physical android only in the case they highly
perceived it as human-like. Furthermore, in [
9
] users exhibited a
prosodic adaptation to the speech rate of a virtual agent displayed
on a screen. Children also duplicated a social robot’s behavior when
the robot modeled creativity [
3
], a growth mindset [
28
] and curios-
ity [17].
In our work, we wanted to explore whether children will replicate
the creative behavior of a social robot in a collaborative story-
telling activity. We designed the robot’s behavior in two conditions
(creative vs. non creative). We then measured the eects of the ro-
bot’s behavior on children’ creativity skills by means of behavioral
coding analysis using recognised creativity metrics. We were also
interested in evaluating children’s perceptions of the robot with
respect to the robot’s behavior in the two dierent conditions.
We summarize the contribution of our work in the following points:
•
This work is among only a few that investigated creativity
in human agent interaction in a social collaborative setting
[3] [6].
•
Previous research has studied eects of using agents on ver-
bal creativity [
29
] [
3
]. Nevertheless, in our research we are
not only exploring the eects of robot’s verbal creativity
(i.e idea generation) but also performance creativity (i.e act-
ing out the scenario) on children’s corresponding creativity
skills.
2 RELATED WORK
2.1 Creativity
There is no concrete denition of creativity. Torrance dened it
as a series of ows that entails recognising a problem, building
hypothetical assumptions and exchanging ideas with others [
41
]
[
6
]. Whereas, creativity was dened as a process of following cues
to produce insights that shift our perceptions in order to shape
novel inventions by Cronin and Loewenstein [14].
Creative theories suggest that there are several types of creativ-
ity that include verbal creativity, gural creativity [
18
] and a type
frequently neglected by scholars known as performance creativity
[
35
]. Verbal creativity relates to the generation and communication
of ideas and thoughts (i.e stories and poetry). Figural creativity
relates to the creation of visual artifacts (i.e drawing, painting and
sculpting). Whereas, performance creativity relates to expressive
performance arts (i.e music and theater).
Four variables are usually utilized to assess verbal creativity: u-
ency, exibility, elaboration and originality [
19
] [
37
] [
39
]. Fluency is
a term used to denote the number of ideas generated in the creative
thought process. Flexibility designates the exibility of generated
ideas in addressing dierent aspects and considering the problem
from dierent perspectives. Elaboration signies the use of details
when producing ideas. And originality indicates an idea being un-
common and surprising.
Several evaluation methods have been developed in order to evalu-
ate creativity. The most widely recognised is the Torrance Test for
Creative Thinking (TTCT) [
39
] which constitutes of sections that
assess both verbal and gural creativity. Other assessment methods
for verbal creativity include the Droodle and Unusual Uses creativ-
ity tasks [23].
In our research we opt for measuring both verbal and performance
creativity in terms of uency (number of ideas generated in the
story), originality (how novel the generated ideas are) and expres-
sivity (whether or not the children are expressive in generating
ideas, for instance by making expressive sounds or acting out the
scenario).
2.2 Agents for stimulating creativity
Numerous experiments in human computer interaction have inves-
tigated the implications of using agents to boost human creativity. In
[
10
], authors explored the potential of creating virtual dynamic per-
sonas on designers’ creative thinking processes. Moreover, CUBUS
[
29
] is a system that utilizes virtual synthetic characters to promote
idea generation in children in a storytelling activity. Educational
scholars along with teachers of robotics have also suggested a posi-
tive impact of using robotics for stimulating creativity in children
[4][44][24].
Therefore, robots have long been used in schools for fostering
creativity. Robotic tool kits have been the most common tools rep-
resenting robots in classrooms. The Thymio mobile robot [
32
] has
been developed by researchers at EPFL as a low cost robotic alterna-
tive to promote creativity, learning and entertainment for children.
The authors in [
11
] have also built an educational robotic tool kit
for boosting children’s design creativity. The toolkit was positively
perceived through the course of four evaluating workshops. Fur-
thermore, robot construction kits such as LEGO Mindstorms and
Cozmo [
26
] [
13
] have been commercially available to teach children
about robotics and allow them to creatively express their ideas.
In recent years, researchers and educators have also been using
social robots for boosting creativity. As an example, PopBots [
42
] is
a robotic toolkit consisting of a social robot built as a smart phone
mounted on LEGO blocks, motors and sensors coupled with a tablet.
The developers of PopBots aimed at using it in educating children
about articial intelligence and stimulating their creative thinking
skills. Authors in [
5
], have created YOLO, a non humanoid social
robot with embedded interactive and social behaviors. The robot
is aimed at fostering children’s creativity by encouraging them to
use it as a tool or a character in storytelling activities.
Most of the previous research on robotics and creativity has ex-
plored using robots as tools to foster children’s creativity. Scarce
research has been conducted on the eects of socially interacting
with a robot on children’s creativity skills. In [
6
], the authors have
used the NAO robot as a peer in a collaborative drawing activity
with children. However, the collaboration with the robot did not
spark higher gural creativity in children who assisted to the activ-
ity with the robot than those who used a tablet instead. Similarly,
children participated in a study with a social robot where together,
they played a Droodle creativity game. Children exhibited higher

Once Upon a Story: Can a Creative Storyteller Robot Stimulate Creativity in Children? IVA ’21, September 14–17, 2021, Virtual Event, Japan
Statement Average rating Assigned condition
The princess was having fun. 2.15 NC
The prince and the princess walked in the garden. 2.2 NC
The sh and the crocodile were swimming in the lake. 2.43 NC
The princess said that she was so scared. 2.43 NC
The princess was happy that the sun was shining! 2.62 NC
The crocodile ate the chicken. 2.73 NC
The prince opened the treasure chest. 2.75 NC
A red eyed bee suddenly appeared and was going to sting the prince. 3.21 C
The princess touched the water but it was too cold. It is still not summer! 3.32 C
Both the prince and the princess changed into their swimming suits and started to swim in the
lake with the crocodile and sh.
3.34 C
The prince and the princess noticed something shaking between the trees. 3.55 C
The alien was scared and was shaking and shaking and shaking that the prince’s eyes turned
red.
3.6 C
The princess knew that that day was the 25th of September, it was her birthday in 2100. 3.69 C
There was a magic chicken that had a very good sense of smell. 4.01 C
The horse opened the treasure chest, there was a tiny bit of a coin that he ate to transform
himself into a lion!
4.21 C
Table 1: Originality ratings of the statements by online participants. Statements rated higher than 3 were assigned to the
creative condition (C) and statements rated lower than 3 were assigned to the non creative condition (NC).
creative behavior when interacting with the creative robot than
when interacting with the non creative robot on the verbal creativ-
ity task [3].
Consequently, we designed our interaction as a social encounter
between the child and a social robot. Moreover, we introduced the
element of collaboration between the child and the robot during
the activity. Within a storytelling context, we are further exploring
aspects of creativity that have rarely been explored in human agent
interaction namely expressive creativity and verbal creativity in a
storytelling context.
3 EXPERIMENTAL DESIGN
We designed the storytelling interaction between the robot and
the child as an interactive collaborative game. In this game, the
child is invited to tell a story together with the robot where they
collaboratively provide ideas to the story line. In order to facilitate
the storytelling game, the child and the robot use a storytelling
software displayed on a touch screen to help with the generation
of ideas.
(a) Scene 1 (b) Scene 2
Figure 2: Sample scenes from the storytelling software
Behavior
Average
rating
Assigned
condi-
tion
Expressive: The prince said: "Let’s all
dance together". (The robot dances)
3.47 C
Non expressive: The prince said: "Let’s
all dance together". (The robot does not
dance)
2.25 NC
Expressive: The princess teleported to
another kingdom: "Oh! It is so scary
here". She said (The robot exhibits a fear
facial expression)
3.39 C
Non expressive: The princess said that
she was so scared. (The robot does not
exhibit any facial expressions)
2.19 NC
Table 2: Expressivity ratings of the behaviors that were in-
cluded in the nal design of the interaction by online partic-
ipants and their corresponding condition assignment (C for
creative, NC for non creative).
3.1 Storytelling interface
We used a simplied (and translated to Swedish language) version
of our previously implemented storytelling interface [
16
] that was
developed using the Unity Game Engine
1
. We chose a fairy tale
castle as the theme for the game. In the interface, the child may
choose between a set of 4 characters and 9 objects in order to tell
the story together with the robot. The interface is displayed on a
touch screen so that the child can move the characters and objects
1https://unity.com/

IVA ’21, September 14–17, 2021, Virtual Event, Japan Elgarf, et al.
all around the scene while telling the story. The child may also
move between dierent scenes within the same fairy tale castle
theme. Screenshots from the software are displayed on gure 2.
3.2 The behavior of the robot
We used the Furhat robot [
2
] which is a a social robot with a virtual
face back projected on a physically embodied head with a pan-tilt
neck as shown on gure 3. The design of the robot allows the as-
sociated benets with a physically embodied agent as well as the
advantage of having a facially expressive virtual face. We designed
the robot’s behavior in two dierent conditions: creative vs. non cre-
ative. We were interested in modeling the robot’s creative behavior
as verbal creativity and performance creativity. For verbal creativity,
we selected both uency and originality of ideas as indicators of cre-
ativity. For performance creativity, we chose expressivity of ideas
dened as acting out the story ideas whether using verbal sounds
or expressive non verbal behavior. We developed the three criteria
with respect to the robot’s condition (creative vs. non creative):
•Fluency:
the robot in the creative condition provided 5-7
ideas during the storytelling interaction while the robot in
the non creative condition provided 2-3 ideas.
•Originality:
the robot in the creative condition provided
more original ideas than the robot in the non creative condi-
tion.
•Expressivity:
the robot in the creative condition used more
expressive behavior while telling the story than the robot in
the non creative condition.
Figure 3: The Furhat robot with the male face used
In order to validate the behavior of the robot, we conducted an
online study that consisted of two parts: one part for validating the
originality of the story ideas of the robot and the other for validat-
ing the expressivity of the robot’s behavior. 100 users participated
in the study (male=36, female=63, prefer not to say=1). For the origi-
nality validation, we extracted dierent statements relevant to the
story theme from a previously collected storytelling data set with
the same software [
16
]. We then asked participants of the online
study to rate them in terms of how creative (in terms of originality)
they think each statement is on a scale from 1 to 5 (extremely non
creative to extremely creative respectively). We then divided the
rated statements into two groups, original statements with an aver-
age rating higher than 3 that were assigned to the robot’s creative
condition and non original statements with an average rating less
than 3 that were assigned to the robot’s non creative condition. The
statements used in our interaction are provided in table 1.
For validating expressivity, we designed a set of four behaviors for
the robot in two conditions; one that is deemed to be more expres-
sive than the other. We asked the online users to rate videos of each
of the presented robot’s behaviors in terms of expressivity on a
scale from 1 to 5 (extremely non expressive to extremely expressive
respectively). We then discarded two behaviors where the users
did not nd a signicant dierence between the expressive and
non expressive behaviors in terms of expressivity. Whereas, we
included another two sets of behaviors where the users rated one
of the coupled behaviors as more expressive than the other. Details
about the included expressive behaviors rated in the online study
are presented in table 2. The robot provided at least one expressive
idea in the creative condition and at least one non expressive idea
in the non creative condition. The robot’s behavior also entailed
general behavior that was constant across the study conditions.
For example, encouraging the child to generate ideas by asking
questions (i.e “What do you think would happen next?” or “Why?”),
praising the child whenever they come up with a good idea by ut-
tering praise expressions ( i.e “That’s great”) and engaging with the
child by laughing or uttering interactive expressions (i.e “Ouch”).
4 EXPERIMENTAL EVALUATION
4.1 Hypotheses
We designed our interaction and were interested in exploring the
eects of the robot’s creative behavior (verbal and performance
creativity) on children’s verbal and performance creativity measures
in a storytelling context. As demonstrated by previous research,
children tend to model the behavior of the robot they are interacting
with. In previous experiments, children modeled robot’s curiosity
[
17
], growth mindset [
28
] and creativity in a Droodle creativity
game [
3
]. In our work, we are investigating whether the same
concept applies to modeling the creative behavior of the robot in a
storytelling context.
We therefore hypothesised the following:
•Hypothesis 1 (H1):
children who interacted with the robot
in the creative condition will exhibit higher verbal creativity
measures than children who interacted with the robot in the
non creative condition.
•Hypothesis 2 (H2):
children who interacted with the robot
in the creative condition will exhibit higher performance
creativity than children who interacted with the robot in the
non creative condition.
We were also interested in evaluating the impacts of the robot’s
creative behavior on children’s perceptions of likeability of the ro-
bot. There is no evidence in research about the correlation between
creative behavior of a robot and perceptions of likeability. Never-
theless, previous literature suggests a positive correlation between
expressivity and perceptions of likeability of the robot by the users
[
34
] [
1
] [
20
]. Our version of the creative robot is highly expressive
(performance creativity) within the context of storytelling and thus,
we hypothesised the following:

Once Upon a Story: Can a Creative Storyteller Robot Stimulate Creativity in Children? IVA ’21, September 14–17, 2021, Virtual Event, Japan
Study condition
N Gender Age
Creativity
scores
Creative 16
male=7,
female=9
M=7.5,
SD=1.9
M=7.56,
SD=2.99
Non creative 16
male=8,
female=8
M=8.1,
SD=1.17
M=8.38,
SD=3.12
Table 3: Balanced group assignments based on gender, age
and creativity level scores
•Hypothesis 3 (H3):
children will perceive the robot as more
likeable in the creative condition than in the non creative
condition.
4.2 Participants
The experiment was conducted at the local Museum of Technology
in Stockholm, Sweden where we recruited 38 children to participate
in our study. We excluded 6 users for either withdrawing from the
activity or for not talking to the robot. Therefore, we analysed the
data for 32 participants (male=15, female=17). 16 children were as-
signed to the creative condition and 16 to the non creative condition.
Their age ranged from 5 to 10 years old (M=7.84, SD=1.61).
4.3 Pre-test
We wanted to eliminate biases from innate children’s creativity lev-
els. Therefore, we decided to divide our sample into two balanced
groups in terms of creativity levels, age and gender to be assigned
to both creative and non creative conditions. We balanced creativ-
ity levels by balancing the creativity scores means and standard
deviations of the two conditions. We measured creativity scores by
administering a pre-test at the start of the interaction. We used two
tasks to evaluate creativity scores as per the strategy proposed in
[23]:
•The Droodle creativity task:
to assess originality of ideas.
•The Unusual Uses creativity task:
to assess both uency
and originality of ideas
The pre-test was evaluated according to the procedures explained
in [
23
] at the end of each day of the study to manage the group
assignment the following day. Statistics about the balanced two
groups assigned to the two dierent conditions are shown on table
3.
4.4 Procedures
We conducted the study in an empty room at the local technology
museum over the course of 3 days (a long weekend). The study was
advertised on the museum’s website where people registered for
participation at specic timing slots. We also recruited participants
by randomly speaking to visitors at the museum. The experiment
was conducted in Swedish language. An experimenter who spoke
the language welcomed the children, guided them through the ex-
periment and tele-operated the robot. Another experimenter was
handling logistics by collecting the questionnaires, assigning partic-
ipant IDs and managing the recording of the video and audio data.
Children’s parents were allowed to stay in the room and witness
the experiment from afar without helping the children. Prior to the
interaction, we collected demographic data about the children from
their parents and administered consent forms for the parents to
sign. The parents gave their stance about the participation of the
child in the experiment, the collection of audio and video data and
the dissemination of the collected data for research purposes. The
experiment lasted around 15-20 minutes per child. This study was
conducted amidst the Covid-19 circumstances and thus disinfection
measures have been taken into consideration.
The interaction started by the rst experimenter welcoming the
child and then guiding him/her through solving the pre-test with-
out providing explicit help. Some children were young and not very
comfortable with writing and thus, the experimenter wrote the in-
formation that they said on their behalf on the pre-test paper. Next,
the experimenter briey explained to the child the storytelling game
that they will participate in with the robot. The experimenter em-
phasized that the child is collaboratively telling the story with the
robot. Therefore, the robot might tell him/her some ideas aligned
with the story plot and the child should share his/her own ideas
with the robot as well. The experimenter told the child that he/she
was free to end the storytelling game whenever he/she wanted. We
controlled the duration of the storytelling part of the interaction to
last for a maximum of 10 minutes to facilitate the comparison of
creativity measures between the dierent children in comparable
timing windows. Some children ended the storytelling interaction
before the 10 minutes limit. Then, the children solved a short ques-
tionnaire about the robot and the game. The experimenter thanked
the children at the end of the interaction by oering them candies.
4.5 Materials
The robot was placed on a table in the middle of the room. The
child sat in front of the robot with the touch screen situated on
the table between both of them. We collected both video and audio
data. We mounted two video cameras: one in front of the child to
capture a frontal view and one next to the child to capture a lateral
view. A microphone was placed underneath the screen in a way
that allowed a high quality of voice recording as well as minimal
distraction for the child by having multiple devices around him/her.
4.6 Measures
In order to assess our hypotheses, we were interested in evaluating
verbal and performance creativity measures. In alignment with the
categories we used for generating the robot’s behavior, we wanted
to measure the same aspects for the resulting children’s behavior
in the storytelling interaction. Therefore we were interested in
measuring the following variables for the children throughout the
storytelling game:
Fluency:
this variable assessed how many ideas each child pro-
vided in the storytelling game. We measured an idea as a clause
with a subject and a verb. A dot denoting the end of the sentence
denotes the end of the idea. Using a conjunction (i.e and, or, but,
then) denoted a new idea only if the child used a new subject. Chil-
dren varied on their performance in terms of uency of ideas that
ranged from providing as low as 2 ideas to 40 ideas per child.

IVA ’21, September 14–17, 2021, Virtual Event, Japan Elgarf, et al.
Originality:
we assessed this variable as the element of using
surprising uncommon ideas by the children. We evaluated the orig-
inality of ideas on a scale from 1-3. 1 denoted basic ideas using
existing elements in the scene (i.e “The prince saw a horse” ). 2
denoted elaborated ideas or ideas with a creative power using ex-
isting elements in the scene (i.e “The crocodile took the bomb in
his mouth and threw it up in the air”). 3 denoted the highest level
of creativity where a child would use an unusual and surprising
idea including but not limited to the use of newly created elements
that do not exist in the scene already (i.e“The princess was having
fun because the prince tickled her”). The nal originality score was
calculated as the weighted average of the originality values of the
ideas generated by the child throughout the story.
Expressivity:
inline with the robot’s expressive behavior, we
wanted to assess expressivity levels of the children during the
storytelling activity. We assessed the expressivity merely in terms
of verbal expressive utterances. We marked the occurrence of an
expressive behavior whenever the child used acting or expressive
sounds while telling the story. For example, some children acted as
the story characters by changing their tone of voice accordingly.
Others used expressive sounds such as “pew pew” to denote that
one character has killed the other.
These variables were measured by means of behavioral coding
analysis. A native language speaker rst transcribed the data in
the videos and generated English subtitles for them to make them
easier for coding by English speakers. We primarily used frontal
videos for the coding and in a few cases, we used the lateral video
instead if the frontal video was unavailable or incomplete. We
developed a coding scheme based on the approach demonstrated
in [
27
]. The coding scheme consisted of measures that evaluated
the child’s as well as the robot’s verbal behavior. Measures related
to the child’s verbal behavior assessed the uency, originality and
expressivity of the children’s ideas. Nevertheless, measures related
to the robot’s verbal behavior validated that the utterances of the
robot conveyed the intended behavior (creative vs. non creative).
We used the ELAN
2
[
43
] software in order to code the videos. ELAN
is an annotation and transcription tool used for behavioral coding
analysis purposes. It was developed by the the Max Planck Institute
for Psycho-linguistics. A primary coder coded all the participants’
videos. A secondary coder coded 25% of the data chosen as a random
sample to enable the evaluation of the agreement as suggested by
standard practice [
12
]. We then used the EasyDIAg toolbox [
22
]
developed for the calculation of inter-rater agreement measures for
ELAN coded data. The tool generates various agreement indicators
that include Cohen’s Kappa, the most signicant statistical measure
for evaluation of agreement in observational research [
7
]. The
Cohen Kappa denoted high agreement between the two coders for
our selected videos with an average of 0.85.
We were also interested in measuring children’s perceptions of the
robot. In the questionnaire that was answered by the children at the
end of the interaction, we assessed three variables: likeability of the
storytelling game, likeability of the robot and perceived intelligence
of the robot. The questionnaire that was presented to the children
was an adapted simplied version of the Godspeed questionnaire [
8
]
2https://archive.mpi.nl/tla/elan
(a) Fluency of ideas (b) Originality of ideas
(c) Expressivity of ideas
Figure 4: Objective measures analysis per condition. We did
not nd a signicant eect of condition on uency, original-
ity and expressivity of ideas.
answered by means of a 5 point Likert scale using a Smily-o-meter
[31] so that it is easier for the children to understand.
5 RESULTS
We evaluated three variables: uency, originality and expressivity
of ideas to assess children’s creative abilities throughout the inter-
action. Our sample followed a non normal distribution, hence, we
used the Wilcoxon signed-rank non parametric test for our statisti-
cal analysis. Our independent variable was the condition (creative
vs. non creative) and the response variables were the uency, orig-
inality and expressivity of the children’s ideas. As demonstrated
on gure 4, analysis of the results did not present statistically sig-
nicant dierence for uency (M=17.28, SD=11.28, p=0.46) between
the creative (M=18.94, SD=12.33) and the non creative conditions
(M=15.63, SD=10.24). Likewise, the dierence between originality
measures were not statistically signicant (M=1.44, SD=0.27, p=0.67)
between the creative (M=1.42, SD=0.28) and non creative conditions
(M=1.47, SD=0.25). Analysis of expressivity measures showed the
same non statistical signicance eect (M=1.72, SD=3.22, p=0.63)
between the creative (M=2.13, SD=3.77) and non creative conditions
(M=1.31, SD=2.6).
We also assessed three variables: likeability of the storytelling game,
likeability of the robot and perceived intelligence of the robot to
evaluate children’s perceptions of the interaction and the robot. We
conducted our data analysis using a one way MANOVA parametric
test given that our sample followed a normal distribution. We used
the condition (creative vs. non creative) as our independent variable
and likeability of the storytelling game, likeability of the robot and
perceived intelligence of the robot as our response variables. There
was no signicant main eect of condition on likeability of the
storytelling game (M=4.06, SD=0.76, p=0.17) , likeability of the robot

Once Upon a Story: Can a Creative Storyteller Robot Stimulate Creativity in Children? IVA ’21, September 14–17, 2021, Virtual Event, Japan
(M=4.4, SD=0.71, p=0.47) and the perceived intelligence of the robot
(M=3.81, SD=0.93, p=0.26).
6 DISCUSSION
As opposed to our hypotheses and to results of previous research
[
3
], we did not nd a signicant main eect of the robot’s creative
behavior on children’s creativity skills. We interpret our results as
an indicator that the robot’s creative behavior is not the only factor
inuencing children’s creativity skills in a child robot interaction.
An element that might have aected the results is that we limited
the duration of the storytelling activity to 10 minutes. 10 minutes
was a short duration since children spent some time getting ac-
quainted with the robot and the activity and thus, might have not
had the time to be highly expressive in this short duration. We
also distinguished another limiting factor: the constrained single
theme of the activity that might have hindered children’s creative
thinking process. In [
3
], authors conducted a study where children
played a turn taking Droodle creativity game with a social robot.
The experiment was also structured in two conditions (creative vs.
non creative). The authors noted that the eect that the robot’s
behavior had on children’s creativity measures varied from one
Droodle to the other. This suggests that in our interaction, varying
the theme of the storytelling activity might produce dierent re-
sults.
In the same study where the children played the Droodle creativity
game with the social robot [
3
], the authors analysed children’s be-
havior and suggested that they perceived the activity with the robot
as rather competitive. Children expressed a desire to beat the robot
by generating more ideas. According to the authors’ interpretation,
the competitive nature of the task might have hindered children’s
expression of creativity. They also proposed designing the interac-
tions with the robot that aim at boosting children’s creative abilities
in a more collaborative way in the future. Nevertheless, in[
6
], chil-
dren complained about the NAO robot ruining their drawings when
they collaborated with it in a drawing activity. In our work, we
observed a similar phenomena. We also designed our activity as a
collaborative interactive game between the child and the robot but
we noticed that some of the children did not like the interruption
of the robot when it started proposing ideas. Some of them showed
signs of frustration by telling the robot “no” each time it suggested
an idea. After introducing a new character to the story plot, one
child said “let’s hope the robot does not kill it!” in an annoyed tone
that suggested she did not like the interference of the robot with
her story. Furthermore, when the robot started providing numer-
ous and successive ideas in the creative condition, some children
stopped generating ideas. They just responded by “yes” or “okay”
instead and started moving the characters according to the robot’s
suggested ideas. In a previous study that we conducted with the
same storytelling software [
16
], the interaction was designed such
that the child was exclusively telling a story to the robot. The robot
merely acted as a listener and provided feedback throughout the
story. In that previous study, we observed that children were more
elaborate than in our current setup where the storytelling activity is
designed as a collaborative game with the robot. This sounds rather
intuitive since children would have more freedom of expression by
telling a story to the robot as a listener. In our current activity, they
might have been constrained by trying to complement the robot’s
suggestions and construct a story around them. On the other hand,
we also observed that although some children were silent at the
start of the activity, they also got motivated after the robot started
proposing ideas and started providing their own suggestions too.
We did not nd a signicant main eect of the creativity condition
on children’s perceptions of likeability and intelligence of the robot
in contrast with what was expected. Nevertheless, we observed that
the average likeability of the game and the robot was rather high
(4.06 and 4.4 respectively on a scale from 1 to 5 where 1 is “did not
like it at all” and 5 “totally liked it)”. We therefore anticipate future
potential for the robot and the storytelling activity in child robot
interaction.
7 CONCLUSION AND FUTURE WORK
We conducted a study with children and a social robot in two con-
ditions: creative vs. non creative. Our aim was to assess whether
children will model the robot’s creative behavior through the inter-
action. In contrast with our expectations, results of the study did
not support the hypothesis that children who interacted with the
creative robot will express higher creativity measures than children
who interacted with the non creative robot. We therefore conclude,
that children’s creativity skills in a child robot interaction are af-
fected by various other factors than the robot’s creative behavior
(i.e the duration of the interaction and the nature of the activity).
In the future, we propose to implement the behavior of the robot
as adaptive to the children’s behavior. The robot would detect chil-
dren’s frustration at its interference and would only suggest ideas
when children are lost or silent for a long time. We also suggest
some minor tweaks to the design of the activity such as prolonging
the limit of the duration of the interaction and providing more
themes for the storytelling activity for the child to choose from.
8 ACKNOWLEDGEMENTS
We would like to thank the reviewers for their valuable feedback
that helped us improve the paper to its current nal version. This
work was supported by the European Commission Horizon 2020 Re-
search and Innovation Program under Grant Agreement No. 765955.
We would also like to acknowledge the support oered by Linda
Sandberg and Siri Olofgörs at the Museum of Technology in Stock-
holm that enabled us to conduct the study in a smooth and safe
environment and to recruit participants for it.
REFERENCES
[1]
Sigurdur Orn Adalgeirsson and Cynthia Breazeal. 2010. MeBot: A robotic platform
for socially embodied telepresence. In 2010 5th ACM/IEEE International Conference
on Human-Robot Interaction (HRI). IEEE, 15–22.
[2]
Samer Al Moubayed, Jonas Beskow, Gabriel Skantze, and Björn Granström. 2012.
Furhat: a back-projected human-like robot head for multiparty human-machine
interaction. In Cognitive behavioural systems. Springer, 114–130.
[3]
Sanah Ali, Tyler Moroso, and Cynthia Breazeal. 2019. Can children learn
creativity from a social robot? In Proceedings of the 2019 on Creativity and
Cognition. 359–368.
[4]
Dimitris Alimisis. 2013. Educational robotics: Open questions and new challenges.
Themes in Science and Technology Education 6, 1 (2013), 63–71.
[5]
Patrícia Alves-Oliveira, Patrícia Arriaga, Matthew A Cronin, and Ana Paiva. 2020.
Creativity encounters between children and robots. In Proceedings of the 2020
ACM/IEEE International Conference on Human-Robot Interaction. 379–388.
[6]
Patrıcia Alves-Oliveira, Silvia Tulli, Philipp Wilken, Ramona Merhej, Joao Gan-
dum, and Ana Paiva. 2019. Sparking creativity with robots: A design perspective.

IVA ’21, September 14–17, 2021, Virtual Event, Japan Elgarf, et al.
In 14th Annual ACM/IEEE International Conference on Human Robot Interaction
(HRI).
[7]
Roger Bakeman and Vicenç Quera. 2011. Sequential analysis and observational
methods for the behavioral sciences. Cambridge University Press.
[8]
Christoph Bartneck, Dana Kulić, Elizabeth Croft, and Susana Zoghbi. 2009. Mea-
surement instruments for the anthropomorphism, animacy, likeability, perceived
intelligence, and perceived safety of robots. International journal of social robotics
1, 1 (2009), 71–81.
[9]
Linda Bell, Joakim Gustafson, and Mattias Heldner. 2003. Prosodic adaptation in
human-computer interaction. In Proceedings of ICPHS, Vol. 3. Citeseer, 833–836.
[10]
Nathalie Bonnardel and Nicolas Pichot. 2020. Enhancing collaborative creativity
with virtual dynamic personas. Applied ergonomics 82 (2020), 102949.
[11]
Silvia SC Botelho, Lilian G Braz, and Ricardo N Rodrigues. 2012. Exploring
creativity and sociability with an accessible educational robotic kit. In Proc. 3rd
Int. Conf. on Robotics in Education (RiE 2012), Prague, Czech Republic. Citeseer,
55–60.
[12]
Jill MacLaren Chorney, C Meghan McMurtry, Christine T Chambers, and Roger
Bakeman. 2015. Developing and modifying behavioral coding schemes in pedi-
atric psychology: a practical guide. Journal of pediatric psychology 40, 1 (2015),
154–164.
[13] Cozmo. 2016. Anki. https://www.digitaldreamlabs.com/pages/cozmo.
[14]
Matthew A Cronin and Jerey Loewenstein. 2020. The craft of creativity. Stanford
University Press.
[15]
Mihaly Csikszentmihalyi. 1997. Flow and the psychology of discovery and
invention. HarperPerennial, New York 39 (1997).
[16]
Maha Elgarf, Natalia Calvo-Barajas, Ana Paiva, Ginevra Castellano, and Christo-
pher Peters. 2021. Reward Seeking or Loss Aversion? Impact of Regulatory Focus
Theory on Emotional Induction in Children and Their Behavior Towards a Social
Robot. In Proceedings of the 2021 CHI Conference on Human Factors in Computing
Systems. 1–11.
[17]
Goren Gordon, Cynthia Breazeal, and Susan Engel. 2015. Can children catch
curiosity from a social robot?. In Proceedings of the Tenth Annual ACM/IEEE
International Conference on Human-Robot Interaction. 91–98.
[18]
Joy P Guilford. 1957. Creative abilities in the arts. Psychological review 64, 2
(1957), 110.
[19] Joy Paul Guilford. 1967. The nature of human intelligence. (1967).
[20]
Adriana Hamacher, Nadia Bianchi-Berthouze, Anthony G Pipe, and Kerstin Eder.
2016. Believing in BERT: Using expressive communication to enhance trust and
counteract operational error in physical Human-robot interaction. In 2016 25th
IEEE international symposium on robot and human interactive communication
(RO-MAN). IEEE, 493–500.
[21]
Galit Hofree, Paul Ruvolo, Marian Stewart Bartlett, and Piotr Winkielman. 2014.
Bridging the mechanical and the human mind: spontaneous mimicry of a physi-
cally present android. PloS one 9, 7 (2014), e99934.
[22]
Henning Holle and Robert Rein. 2015. EasyDIAg: A tool for easy determination
of interrater agreement. Behavior research methods 47, 3 (2015), 837–847.
[23]
Peter H Kahn Jr, Batya Friedman, Rachel L Severson, and Erika N Feldman.
2005. Creativity tasks and coding system–used in the plasma display window
study. University of Washington, Seattle, WA: The Information School, University
of Washington (2005).
[24]
Ahmad Khanlari. 2013. Eects of robotics on 21st century skills. European
Scientic Journal 9, 27 (2013).
[25]
Kyung Hee Kim. 2011. The creativity crisis: The decrease in creative thinking
scores on the Torrance Tests of Creative Thinking. Creativity research journal 23,
4 (2011), 285–295.
[26]
LEGO Mindstorms. 1998. The LEGO Group. https://www.lego.com/en-us/themes/
mindstorms.
[27]
Yfke P Ongena and Wil Dijkstra. 2006. Methods of behavior coding of survey
interviews. Journal of Ocial Statistics 22, 3 (2006), 419.
[28]
Hae Won Park, Rinat Rosenberg-Kima, Maor Rosenberg, Goren Gordon, and
Cynthia Breazeal. 2017. Growing growth mindset with a social robot peer.
In Proceedings of the 2017 ACM/IEEE International Conference on Human-Robot
Interaction. 137–145.
[29]
André Pires, Patrícia Alves-Oliveira, Patrícia Arriaga, and Carlos Martinho. 2017.
Cubus: Autonomous embodied characters to stimulate creative idea generation
in groups of children. In International Conference on Intelligent Virtual Agents.
Springer, 360–373.
[30]
Gerard J Puccio. 2017. From the dawn of humanity to the 21st century: Creativity
as an enduring survival skill. The Journal of Creative Behavior 51, 4 (2017),
330–334.
[31]
Janet C Read. 2008. Validating the Fun Toolkit: an instrument for measuring
children’s opinions of technology. Cognition, Technology & Work 10, 2 (2008),
119–128.
[32]
Fanny Riedo, Philippe Rétornaz, Luc Bergeron, Nathalie Nyeler, and Francesco
Mondada. 2012. A two years informal learning experience using the thymio
robot. In Advances in Autonomous Mini Robots. Springer, 37–48.
[33]
Ken Robinson and John Rafter Lee. 2011. Out of our minds. Wiley Online Library.
[34]
Maha Salem, Friederike Eyssel, Katharina Rohlng, Stefan Kopp, and Frank
Joublin. 2011. Eects of gesture on the perception of psychological anthropomor-
phism: a case study with a humanoid robot. In International conference on social
robotics. Springer, 31–41.
[35]
R Keith Sawyer. 1998. The interdisciplinary study of creativity in performance.
Creativity Research Journal 11, 1 (1998), 11–19.
[36] Robert J Sternberg. 1999. Handbook of creativity. Cambridge University Press.
[37]
Robert J Sternberg. 2005. Creativity or creativities? International Journal of
Human-Computer Studies 63, 4-5 (2005), 370–382.
[38]
Apurv Suman, Rebecca Marvin, Elena Corina Grigore, Henny Admoni, and Brian
Scassellati. 2016. Prior behavior impacts human mimicry of robots. In 2016 25th
IEEE International Symposium on Robot and Human Interactive Communication
(RO-MAN). IEEE, 1057–1062.
[39]
Ellis Paul Torrance. 1966. Torrance tests of creative thinking: Norms-technical
manual: Verbal tests, forms a and b: Figural tests, forms a and b. Personal Press,
Incorporated.
[40]
E Paul Torrance. 1968. A longitudinal examination of the fourth grade slump in
creativity. Gifted Child Quarterly 12, 4 (1968), 195–199.
[41]
E Paul Torrance. 1988. The nature of creativity as manifest in its testing. The
nature of creativity: Contemporary psychological perspective (1988).
[42]
Randi Williams, Hae Won Park, Lauren Oh, and Cynthia Breazeal. 2019. Popbots:
Designing an articial intelligence curriculum for early childhood education. In
Proceedings of the AAAI Conference on Articial Intelligence, Vol. 33. 9729–9736.
[43]
Peter Wittenburg, Hennie Brugman, Albert Russel, Alex Klassmann, and Han
Sloetjes. 2006. ELAN: a professional framework for multimodality research. In
5th International Conference on Language Resources and Evaluation (LREC 2006).
1556–1559.
[44]
Karolina Zawieska and Brian R Duy. 2015. The social construction of creativ-
ity in educational robotics. In Progress in Automation, Robotics and Measuring
Techniques. Springer, 329–338.