Original Article
DOI: http://dx.doi.org/10.1590/2446-4740.01316
*e-mail: carlos.valadao@aluno.ufes.br
Received: 23 February 2016 / Accepted: 04 July 2016
Analysis of the use of a robot to improve social skills in children
with autism spectrum disorder
Carlos Torturella Valadão*, Christiane Goulart, Hamilton Rivera, Eliete Caldeira,
Teodiano Freire Bastos Filho, Anselmo Frizera-Neto, Ricardo Carelli
Abstract Introduction: Autism Spectrum Disorder is a set of developmental disorders that imply in poor social skills,
lack of interest in activities and interaction with people. Treatments rely on teaching social skills and in such
therapies robotics may offer aid. This work is a pilot study, which aims to show the development and usage
of a ludic mobile robot for stimulating social skills in ASD children. Methods: A mobile robot with a special
costume and a monitor to display multimedia contents was designed to interact with ASD children. A mediator
controls the robot’s movements in a room prepared for interactive sessions. Sessions are recorded to assess the
following social skills: eye gazing, touching the robot and imitating the mediator. The interaction is evaluated
using the Goal Attainment Scale and Likert scale. Ten children were evaluated (50% with ASD), using as
inclusion criteria children with age 7-8, without use of medication, and without tendency to aggression or
stereotyped movements. Results: It was observed that the ASD group touched the robot about twice more in
average than the control group (CG). They also looked away and imitated the mediator in a quite similar way
as the CG, and showed extra social skills (verbal and non-verbal communication). These results are considered
an advance in terms of improvement of social skills in ASD children. Conclusions: Our studies indicate that the
robot stimulated social skills in 4/5 of the ASD children, which shows that its concepts are useful to improve
socialization and quality of life.
Keywords: Autism spectrum disorder, Social skills, Social robots, Assistive technology.
Introduction
Children with Autism Spectrum Disorder (ASD)
may lack social and communication abilities, which
are fundamental for quality of life and interaction
with other people (Kim et al., 2013). These children
usually have difculties in displaying and perceiving
emotions and social clues, a situation that complicates
even more their lack of communication (Duquette et al.,
2008; Kim et al., 2013; Scassellati et al., 2012).
The aetiology of ASD remains undened and studies
show that this condition may have multifactorial
causes such as genetics and environment (Rutter,
2005). The ASD comprehends the classic autism, the
Asperger Syndrome and the PDD-NOS (Pervasive
Developmental Disorder Not Otherwise Specied)
(Brasil, 2013; Lord et al., 2000).
Considering its spectrum nature, ASD appears
in different levels and symptoms, although the main
aspect, remarkable in all kinds of ASD, is the lack of
socialization and communication skills. In addition,
individuals affected by ASD have difculties in
understanding and expressing emotions, engaging in
eye contact, joining interactive activities (imitation and
joint attention), among other social-like behaviours,
and sensitivity to physical contact (touching)
(Scassellati et al., 2012). In the Asperger Syndrome,
there is a normal intellectual developing (Lord et al.,
2000), with paired language and intellect, but these
children also present social interaction decits and
restricted interests and behaviours (Klin, 2006).
ASD epidemiology varies from 0.62% to 2.64% of
the population, values that change due to the techniques
and data used to diagnose autism (Elsabbagh et al.,
2012; Leventhal et al., 2013). According to the Autism
Society (Autism…, 2016), the Central of Disease
Control and Prevention (Centers…, 2013) estimates
that one in each 68 children is born with autism in
the USA (1.47%), where the autistic population is
more than 3.5 million people. The Diagnostic and
Statistical Manual of Mental Disorders estimates
that an average of 1% of the world population has
autism spectrum disorder (American…, 2013). Some
statistics shows the prevalence of autism worldwide:
in France, it is estimated 0.67% of the population has
ASD (Autism…, 2012); in Canada, 0.68% (Norris et al.,
2006); in Singapore, 1% (Autism…, 2013); in the
UK, 1.1% (The National…, 2015); in Japan, 1.61%
Volume 32, Number 2, p. 161-175, 2016
Valadão CT, Goulart C, Rivera H, Caldeira E, Bastos TF Fo, Frizera-Neto A, Carelli R
(Honda et al., 2005); and in South Korea, 2.64%
(Leventhal et al., 2013).
Still regarding the epidemiology worldwide, the
gender ratio of children diagnosed with ASD are of
4.2 boys for each girl, which shows a higher prevalence
in male population (Fombonne, 2009). In Brazil the
epidemiologic studies are scarce, with an estimate of
500,000 people with autism (0.29%), based on the
2000 Census (Brasil, 2013). However, there is no
current ofcial numbers of ASD people in Brazil.
In addition, the diagnostics of autism is not simple
and can be based on several factors and biological
data, such as clinical, molecular and neuroimaging
ndings, among others (Rossignol and Frye, 2012).
Children affected by ASD can rely on behavioural
treatments that can help them to live a better life
by teaching them social skills. Studies, such as
(Scassellati et al., 2012), suggest that such intervention
therapies should start as early as possible to achieve
its maximum efciency. The treatment is based on
stimulating the child to interact with other people and
using several tools and strategies, such as toys and
activities that resemble games. It is usual for children
to receive different stimuli, for example different kind
of sound (Lamas et al., 2009), although most of these
studies focus on social and communication skills
(Scassellati et al., 2012). Currently, several studies try
to nd ways to bring new stimuli to those children,
in order to make them achieve better outcomes in the
cognitive and social development. For example, robots
and games are used to stimulate the development
of social skills, such as verbal, imitation and tactile
sensitivity (Cabibihan et al., 2013; Kim et al., 2013),
in order to investigate how these children react to
different kinds of stimuli and how these stimuli could
be useful for therapies and treatments (Duquette et al.,
2008; Michaud et al., 2003).
Robots able to stimulate social and cognitive
abilities in ASD children can be classied into
anthropomorphic, non-anthropomorphic and non-
biomimetic (Cabibihan et al., 2013; Scassellati et al.,
2012). The rst category resembles some human
features in its appearance. The second one does not
resemble human features, but biological features, such
as animals. Finally, the third category is composed
of robots, whose appearances do not resemble any
biological creature. Examples of those robots are
PLEO, Keepon and Paro (non-anthropomorphic)
(Kim et al., 2013; Kozima et al., 2009; Scassellati et al.,
2012; Wada et al., 2004), KASPAR, ROBOTA, NAO
(anthropomorphic) (Cabibihan et al., 2013; Dautenhahn,
2003; Robins and Dautenhahn, 2014) and Pekee
and Roball (non-biomimetic) (Michaud et al., 2003;
Salter et al., 2006). All of them are used in autism
therapy, trying to assist children in demonstrating and
perceiving emotions, as well as interacting with others.
In the eld of robotics, these devices are known as
social-assistive robots, which is addressed to help people
to express emotions and offer them the opportunity
of human-robot interaction (Scassellati et al., 2012).
In fact, ASD children may react better to
robots, due to its predictability, in contrast with the
unpredictable nature of humans (Cabibihan et al.,
2013; Duquette et al., 2008). Thus, it is possible
to design a robot to interact in different ways with
these children and give its architecture a ludic
aspect, therefore, providing the robot the possibility
of offering a pedagogical and developmental aid to
these children (Duquette et al., 2008; Kim et al., 2013;
Kozima et al., 2009). An example is a toy-like robotic
platform named ROBUS (ROBot of Université de
Sherbrooke), used in research with ASD in Canada
(Michaud and Clavet, 2001). Other examples of
robotic aid in therapies of ASD children is shown in
researches (Cabibihan et al., 2013; Kim et al., 2013;
Kozima et al., 2009; Robins and Dautenhahn, 2014;
Scassellati et al., 2012).
This paper describes a research that employs
a robot to stimulate social skills in ASD children,
promoting eye gaze, touch, and imitation, besides
interaction with people. The robot used is called
MARIA (acronym for “Mobile Automatic Robot
for Interaction with Autistics”) and this work is a
pilot study with ve ASD children and ve children
from a control group, CG (without ASD), based on
interactive session with the robot MARIA. In such
sessions, the children’s behaviours are analysed, and
their social skills and quality of child-robot interaction
are measured through a quantitative scale (GAS)
and a questionnaire using Likert scale. A search
conducted in 2016 in SCOPUS, Web of Science
and IEEE Xplore databases for “autism spectrum
disorder”, “social robots” and “assistive technology”
resulted in no other work with similar approach
related to ASD children interacting with robots, and
no other work was found with similar features as
the robot here presented: protocol including a robot
self-presentation; robot with a mix style of human and
non-biomimetic appearance; the way of evaluating
the results, including questionnaires and scales, in
addition to specic aspects related to socialization,
such as number of times that the children gaze the
robot, touch the robot, and imitate the mediator.
Methods
Volunteers
The volunteers of this study were ve ASD
children and ve without ASD (control group),
with ages ranging from 7 to 8 years old. The ASD
children are with AMAES (acronym in Portuguese
162
Res. Biomed. Eng. 2016 June; 32(2): 161-175
Robot to improve social skills in ASD children
for “Association of Friends of Autistics of Espírito
Santo”). The children from the control group are
with EMEF-UFES (Municipal Elementary School
of Vitoria at Federal University of Espírito Santo).
To the ASD children attend the sessions, the inclusion
criteria used were: children should have an ASD
diagnostic made by a physician. On the other hand,
the exclusion criteria were: children should not have
experienced any traumatic episode or phobias, nor
neurological diseases simultaneously with ASD or
other syndromes that affect signicantly the motor
and behavioural development; expressive stereotyped
movements; constant tendency to aggressiveness; or
use of medication that affects the neurological system.
These exclusion criteria are important, since
this is a pilot study and the features aforementioned
could affect directly the analysis that are made
during the experiments, as most of the ASD children
take medications in a regular basis, which can alter
signicantly their behaviours. Besides, most of them
present stereotype movements, tendency to aggression
or agitation, fact that notably reduces the available
number of volunteers. This turns the volunteer selection
extremely difcult, adding the fact some parents do
not accept their children take part on the sessions or
simply they do not show up in the sessions.
The volunteer ASD children were diagnosed by
medical professionals before entering in AMAES and
were under psychological, speech and/or pedagogical
therapy in such association. Before selecting the
children, an extensive study and analysis together
with the institutional pedagogues were made in
order to determine which children could participate.
After making the analysis with the pedagogues, we
were able to select ve ASD children to take part of
the experiments whose details are described in the
list below:
• V1: Female, 7 years old. Diagnosed with
autism on 03/02/2011.
• V2: Male, 8 years old. Diagnosed with autism
on 06/20/2011.
• V3: Male, 8 years old. Diagnosed with autism
on 05/07/2012.
• V4: Male, 7 years old. Diagnosed with autism
on 08/14/2012.
• V5: Male, 7 years old. Diagnostic dened in
01/2012.
All parents (from both groups) signed a consent
form, previously approved by the Ethical Committee of
the Federal University of Espírito Santo (#1,101,769).
The parents, caretakers, family members, and/or legal
guardians were invited and were present to watch the
experimental sessions.
This manuscript presents a proof-of-concept.
Therefore, the authors aim to extract valuable information
while aiming at a smaller group (10 subjects in total, with
50% being ASD children) when compared with more
complex clinical trials. Such conduct is quite common
in early stages of validation of a prototype. In fact,
works, such as (Robins et al., 2004; Shamsuddin et al.,
2012; Wainer et al., 2014), use smaller number of
volunteers to make a proof-of-concept. The analysis
made using the metric scales, which will be further
explained, brings precious and important information,
even with a small sample. Therefore, it is possible to
evaluate the robotic system (explained in the following
section) as a tool for applications in therapies with
children with ASD and, then, showing that the robot
is useful to promote child-robot and child-mediator
interaction.
Robotic system
The methodology followed in this study was based
on customizing a mobile robot, in order to give it a
playful and toy-like appearance to interact with the
children, stimulating their social skills, such as eye
gaze, touch, and imitation, as those features usually
are remarkably faulty in ASD children. Thus, taking
into account the potential interaction aid that robots
can offer to these children, the robot MARIA was
designed in the Federal University of Espirito Santo
(UFES), Brazil. An important aspect of the robot
design is that it has a child-size height, which means
the volunteers from both groups would interact with
a robot with their similar height. Added to the size
feature, the robot MARIA is a mobile robot, which
differs from several robots presented in the literature
that do not present mobility, such as (Kim et al., 2013;
Kozima et al., 2009; Wainer et al., 2014). To develop
the robot MARIA, some guidelines were used: the
balance between human and robot (mechanical)
style followed the recommendations found presented
in the work (Giullian et al., 2010). In addition, the
colourful aspect was taken into consideration, since
it has great importance as shown in (Paron-Wildes,
2005). The robot features, further explained, were
designed to maximize the interaction potential
between the robot and the child. In addition to the
robot shape itself, moving objects also catches the
attention of children with ASD, thus, mobility is an
another feature incorporated to the robot MARIA
(Cabibihan et al., 2013).
To assembly this robot, a mobile platform was
used, which was complemented by other accessories.
The devices used in the robot MARIA are: (1) Pioneer
163Res. Biomed. Eng. 2016 June; 32(2): 161-175
Valadão CT, Goulart C, Rivera H, Caldeira E, Bastos TF Fo, Frizera-Neto A, Carelli R
3-DX differential drive mobile robot that allows the
robot MARIA movements. It has three wheels, one
free and two other motor wheels, and its kinematics is
differential with non-holonomic constraints; (2) 12”
monitor used to display animations; (3) Sound speakers
used to emit sounds of the animations; (4) Internal
notebook used to control the display and speakers;
(5) Wi-Fi link to allow the internal computer connection
to the master computer through VNC (Virtual Network
Computing - a protocol designed to allow accessing a
computer through network using graphical interface,
as if the user were in front of the real computer) via
Wi-Fi connection.
To hide the electronics and to make the robot
friendlier, a costume was designed using soft breboard,
colourful uorescent papers, and ethylene-vinyl
acetate foam (EVA). The costume gives a ludic shape
to the robot, in order to turn the robot more attractive
to the children. MARIA costume has human female
features (eyes, nose, mouth, wig and eyelashes),
stylized into a robotic-toy style, thus featuring a mix
of anthropomorphic and non-biomimetic appearance.
Figure 1 shows MARIA and a schematic diagram
of the system functioning. The robot and the master
computer (controls both the internal robot computer
and the embedded robot computer; the embedded
computer is actually a notebook that is on the robot
and has no direct connection with the robot, only
with the display and sound speakers) are connected
to the router and, therefore, the master computer
is able to control, send, and receive information
from the robot. The monitor is connected directly
to the robot’s computer (embedded notebook) and
is indirectly controlled by the master computer.
All these connections are made using a router and
wireless connection. The robot functioning is shown
in Figure 1a and the robot image is displayed in
Figure 1b.
The robot MARIA, specially developed for social
interaction with ASD children, has some remarkable
features, compared to others in literature. First,
most of the robots presented in the literature are
smaller than the children, such as NAO, KASPAR,
KEEPON, PLEO, ROBUS (Anzalone et al., 2014;
Duquette et al., 2008; Kozima et al., 2009; Robins and
Dautenhahn, 2014). MARIA has the average height
of the volunteers (1.35 m), which, according to the
literature (Robins et al., 2006), makes the interaction
easier. In addition, most of the robots used to help
ASD children are static and neither use videos to
stimuli them nor make its self-presentation before
starting the experiments. These features distinguish
our work from the others presented in the literature.
Table 1 shows a comparison between the robots
and the experiments conducted with those robots
and the robot MARIA, making clear that MARIA
is an innovative robot, with elements no previously
addressed.
MARIA is a child-size mobile robot with similar
height of the volunteers and with multimedia
devices, besides including self-presentation and
other behaviours, which will be further explained.
Regarding the number of children, the hypothesis we
want to prove is if a mobile robot with the similar
height and with self-presentation could promote
an interaction in a low number of sessions without
any previous training from the child. The main
idea of this work is to make a proof-of-concept
application, in order to know whether the robot
MARIA (child-size robot) with multimedia content
can make ASD children develop their social skills.
As shown in some works, such as (Giullian et al.,
2010), the robot appearance inuences directly in the
children’s behaviour and reactions. Mobile objects
and colourful aspects are also important aspects for
children with ASD, regarding the environment, as
shown in (Paron-Wildes, 2005), where the author
says 85% of the children with ASD can see the
colours in a more vibrant intensity.
In addition, the research conducted in (Giullian et al.,
2010) suggests some guidelines to make a robot that
will be visually attractive and functional to both
children and therapist. There, the authors discuss
the idea of using a child-size robot is positive, since
it may encourage the children to interact with it.
In addition, the balanced similarity with human
and robot, i.e., not being so human-like and, at
same time, not being so mechanical-like, makes
the robot more attractive to the children. If it looks
too human, it may lead the children into fear and/or
lack of interest, while if it looks too mechanical,
the child would be more interested in examining it,
instead of interacting with it (Giullian et al., 2010).
Summarizing, the robot should have the children’s
height to allow eye-level interaction and, at the same
time, it should look a mix-style between human
and mechanical structure to make the children
interested in the robot. In (Giullian et al., 2010),
it is also commented the importance of the robot
having a predened choreography in its memory,
such as the self-presentation, which is in the section
“Experimental sessions”. Complementing those
previous guidelines, the robot shape, colours, and
size were also dened according to suggestion of
caretaker and health professionals from AMAES
and children’s parents. Thus, one of the innovations
of this work is to use the robot with these whole
guidelines regarding height, size, and colours.
164
Res. Biomed. Eng. 2016 June; 32(2): 161-175
Robot to improve social skills in ASD children
This helps the child getting used to the robot
and also allows the therapist not worry with the
robot and focus on the therapy instead on the robot
functioning. It is important to emphasize the children
should not have any previous contact with the robot
before the tests, thus making the self-presentation
essential to make the child comfortable and willing
to play with the robot.
Their suggestions led us to build the robot with
the average 7-8 years old children height and with
a plenty number of different colours and textures,
making it more attractive to the children. This may
help the interaction of the children with other children
and they can see the robot as a friend. In addition, the
self-presentation, detailed in “Experimental sessions,”
Phase 1, is essential to create a sense of safety in the
child that interacts with the robot. The self-presentation
is also another innovation of this work.
Summarizing, the robot aspect and the Phase 1 of
the experiments were important to make the children,
who have never seen the robot before, get used to it.
Afterwards, in the next part of the experiment, they
were more condent in interacting with the robot.
Experimental sessions
ASD children have difculty in processing large
number of information and stimuli as it may overload
them, which may lead them not to behave naturally.
Therefore, the experiments need a simple setup and
should not be complex, otherwise their natural behaviour
towards the robot could not be analysed. Thus, the
experiments conducted were not complex, and as well
as not easy for the ASD children, in order to provide
signicant and important information regarding
their behaviours and whether they interact with the
robot. In addition, for not overloading the children,
the experiments were split into two distinct phases:
Figure 1. Schematic of the robot operation and details of the robot structure.
165Res. Biomed. Eng. 2016 June; 32(2): 161-175