Τhe effectiveness of socially assistive robotics in children with autism spectrum disorder

Abstract

Aim: The current study investigates the outcomes of the interaction between four elementary school pupils with autism spectrum disorders and a robot called Daisy.

Method: During structured and prepeared activities which were conducted by the social robot, as well as by a human partner.

Results: Results indicate positive outcomes during the interaction with the robot. Specifically, there were more incidences of eye contact, proximity and verbal interaction during sessions with the robot than during those with the teacher. Additional behaviors such as increased attention and ability to follow instructions improved during interaction with the robot. There was also a noted reduction in fidgeting.

Introduction

Autism Spectrum Disorder (ASD) is characterized by deficits in social and communication skills as well as stereotypical behaviors (DSM-5, APA, 2013). According to estimates 1 in 68 children will be affected in the USA by end of this decade (Christensen 2016) and 1 in 100 people globally (Center for Disease Control and Prevention 2016). Social and communication skills deficits in autism encompass a number of other deficiencies, like difficulty in recognizing body language or in demonstrating eye contact with the people they converse with (Jones and Klin 2013), and problems understanding the emotions of others as well as expressing their own (Baron-Cohen et al. 2013). Characteristics vary among individuals with ASD with regard to severity, they may range from mild to serious, thus there is no specific treatment for children to overcome these difficulties. Early, individualized intervention can help the child with ASD, so that he or she can improve the quality of life (Boyd et al. 2010, Scassellati et al. 2012, Webb et al. 2014). There is a vast amount of reference to different interventions, therapeutic methods and techniques used to ameliorate the quality of life of individuals with ASD making them as independent as possible (Syriopoulou-Delli and Cassimos 2013). Several alternative communication systems have been designed to precipitate spontaneous, functional communication of children on the autism spectrum or to introduce a sequence of events progressively utilizing visual devices in order that children achieve as much autonomy as possible in all areas of functionality (Schopler 2001, Mesibov and Shea 2009, Bondy and Frost 1998). Moreover, other approaches have been used such as that of having a non-human collaborator act as a mediator for the enhancement of the interpersonal interactions of the child with ASD. In addition, virtual reality has been used with children on the autism spectrum targeting their deficiencies such as fine motor skills (Zhao et al. 2017) assisting initially in the assessment and consequently in the improvement of these areas, as well as treatment through the computer (Scassellati et al. 2012).

Research into both diagnosis and treatment of people with ASD within the field of robotics has been popular since the previous decade. The field of robotics designed to help more in social rather than physical interactions is relatively recent. Known as Social Assistive Robotics (SAR) at the focus of interest is the treatment of children with ASD. Likewise, the current study aims to investigate the enhancement of social and communication skills of children with ASD when they participate in suitably prepared, structured activities. The interest of this study is that the tool used to carry out the activities besides the human user is the robotic partner – a stuffed toy robot shaped like a flower (Figure 1), which embodies characteristics of expression and speech. This robot in unison with the teacher carries out activities alternately with each child participating in the study. The data subsequently collected evidenced that it is possible to develop an innovative educational tool that will enhance the teaching and learning procedures and will develop the social skills of children with ASD.

Figure 1.

The social robot Daisy.

Social robots have been designed to appeal to and enhance the social behavior of those who interact with them (Kim et al. 2013). Previous investigations demonstrate that they can involve the student in interactive tasks as they form a motivation and impel them to participate in social interaction (Kozima et al. 2008, Scassellati et al. 2012). With the passage of time, different kinds of robots have been created each of which has been used to strengthen the social skills of children with ASD. Depending on the purpose of the study that is conducted, robots can be selected and used to achieve the intended goals. For example, a study that aims to teach mimicry skills would more likely choose a humanoid robot, and if the goal is to involve the student in a simple interactive exchange then a non-humanoid robot may be preferable. Accordingly, a humanoid robot is preferred when the purpose is the generalization of teaching behaviors, given that similarity to the human form facilitates this aim. Moreover, simplistic forms, may be used because they hold the attention of children with ASD successfully. Finally, it is possible for a session to begin with the use of a simple robot to get the student involved in the activity and after some sessions for it to be replaced with a more realistic robot to teach complicated behaviors. In this way, multiple goals may be attained (Ricks and Colton 2010).

Research using robots with children on the autism spectrum, usually aims to engage the students and involve them in an activity. The training of children with autism is directed at having them learn the correct way to begin an activity and replacing the incorrect way. For this reason, robots have been designed that target the reinforcement of interaction, which were esteemed to successfully fulfill this aim (Ricks and Colton 2010). Additionally, major goals of teaching children with ASD are increasing attention span and enhancing social and communication behavior, including skills like eye contact, proximity and spontaneous imitation when they participate in structured activities with a robot, behaviors which form the foundation on which social skills are adopted and social interaction ameliorated (Scassellati et al. 2012). Investigations have demonstrated that the eye contact level of the children increases when they participate in activities conducted by a robot (Τapus et al. 2012, Simut et al. 2016), where they do not hesitate to go up close to the robot, even to touch it, behaviors which do not emerge so readily with the human partner. Finally, researchers have shown that children with autism produced speech to a greater extent when interacting with the robot compared to when the interaction took place with a human partner (Stanton et al. 2008, Kim et al. 2013). Generally, research demonstrates that, in cases of children with ASD, robots manage to induce a higher level of involvement in an activity.

To recap, the purpose of the current study is to examine

1. the difference between the intervention conducted by the teacher and the one carried out by a robotic partner with respect to the quality of interaction (eye contact, proximity and verbal interaction).

2. the difference between the intervention conducted by the teacher and the one carried out via robotic partner regarding the involvement of students in the teaching process (whether the children pay attention when spoken to, follow the given instructions and how often fidgeting occurs to the extent that it disrupts an activity).

The current study was carried out to examine the role of the Socially Assistive Robotics (SAR) as an innovative educational tool in the development of the social skills of children with autism as they participated in structured and suitably prepared activities, which in the present study was conducted by a social robot, Daisy, as well as a human partner in order subsequently to compare the results of the two different interventions. Participants in the study comprised four children with autism who are pupils at elementary school and are simultaneously assisted by a special support teacher. The study was carried out in a special education center and sessions were held outside of the regular school timetable. In total, eight 30-minute sessions each comprising four activities were held with each of the pupils. In those sessions the researcher examined the two aforementioned goals of the research and for this purpose they used data forms for frequency (Zirpoli 2005), to count the frequency of the appearance of specific behaviors within the referred period.

The positive outcomes of the current research would be significantly helpful and promising for the teaching process of children with ASD and, in addition to special education teachers in schools; they could also be implemented by parents at home and other specialists involved in the care of children with ASD with the aim of assisting them to showcase and reach their full potential.

Method

Participants

Participating in the research were four children with ASD, ranging from 7 to 12 years in age. They were all pupils in the general classroom at a state elementary school in Northern Greece where all were simultaneously assisted by a special support teacher. Outside of school they also received specialist help. The selection criteria for participation in the study were the following:

1. Age ranging from 7 to 12 years old

2. A diagnosis of ASD from a certified organization

3. A linguistic ability to form a sentence consisting of more than 3 words.

4. An IQ of ≥70. The IQ scores were assessed with the Greek WISC III, which is the Greek edition of the cognitive intelligence scale for children, Wechsler Intelligence Scale for Children between the ages of 6 and 16 years and 11 months.

Unstructured interviews were conducted with the pupils’ parents and the specialists who dealt with them outside school. Then pupil profiles were mapped out to establish an appropriate baseline level for the composition of the base group. The profiles also provided the foundation for the activities to be set out.

Pupil’s profile

Pupil 1 (7 yrs 5 mths). (1) He attends first grade and speech and occupational therapy after school. (2) His assessment showed adequately developed gross and fine motor skills, he holds a pencil correctly and performs activities using both hands. His vocabulary is sufficient for his age, but his speech is beset by the constant repetition of random expressions. He has deficiencies regarding seriation tasks, processing and encoding information. He cannot plan activities and has trouble giving fully developed answers. (3) With regard to behaviors, he is given to intense bouts of laughter/crying and he often cannot interpret the interlocutor’s facial signals. Levels of eye contact are very low – generally he avoids looking at others. He rarely follows instructions and is easily disrupted if he does. He has great difficulty concentrating and only tends to play very monotonous games.

Pupil 2 (10 yrs 1 mth) is a high functioning child on the spectrum. (1) He attends fourth grade and outside school is supported by a speech therapist, occupational therapist, and a psychologist. (2) His assessment showed adequate gross and fine motor skills and sufficient understanding of the concepts of space and time. He tends to use the right-hand side of his body. Although his vocabulary is average for his age, he has problems expressing himself, making semantic connections and processing language rationally. (3) Regarding behaviors, he has trouble with any team activities and his behavior often has to be curbed as he cannot handle loss and obstructs games regardless of the effect on his peers. He is insensitive to the feelings and needs of others.

Pupil 3 (6 yrs 8 mths) (1) He attends the first grade and outside school is supported by a speech therapist and has extra lessons with a special education teacher. (2) His assessment showed adequate gross motor skills, but fine motor skills are impaired and he has difficulty holding a pencil. Laterality is unclear, but he uses his right hand to write with. His language tends to be incoherent and unstructured and his vocabulary is limited. (3) The behaviors of the pupil are stereotypical; moderate eye gaze, he avoids contact and collaboration with others, his activities are repetitive, he reacts negatively to changes in routine, he cannot stay still or focus for long and tends to flap his arms and stand on one leg.

Pupil 4 (12 yrs 2 mths) (1) He attends fifth grade and outside school is supported by a speech therapist, special needs teacher and a psychologist. (2) His assessment showed adequate gross and fine motor skills and a competence with abstract concepts. He also has little difficulty with seriation and processing. Communication is problematic as his short-term memory is weak, he cannot give developed answers and does not recognize jokes or metaphors. (3) As for behaviors, he has little grasp of social conventions, nor can he interpret others’ feelings or needs. He is easily distracted and lacks focus.

Before the research, parental consent was given, and parents were informed that pupils could leave the process at any time and for any reason.

Experimental design

The intervention was carried out in the special education center outside of the regular school timetable and lasted five months from November 2016 to April 2017. A single-case design intervention was carried out in the current study as the purpose of the research was to investigate differences in the performance of each pupil individually in the matter of definite variables (Kratochwill and Levin 2014). There was a familiarization phase (A) and after that, both the teacher (B) and the robot (B′) conducted the same activities 1,2,3,4 with the same protocol and identical steps and phrases so that the two sessions differed only in the type of interaction partner. The participants were divided in two groups (Pupils 1 and 3 and Pupils 2 and 4) conducting the same activities. The session for each pair followed a different sequence so that two participants followed the sequence B-B′-B′-B and the other the sequence B′-B-B-B′ in order to predict methodological issues.

Settings

The study was carried out in four (4) different rooms in the special education center which the children visited outside of the regular school timetable. On one side of the room was the work table where the activities were conducted. The child sat at one side of the table and the partner (teacher or robot) sat at the other side. Two more chairs were located in the room. The first was in front of the work table and the second was placed at the side of it. The purpose of the chairs was to have a better view of the child’s performance. In the session with the robot, an extra chair was located for the operator of the robot. The operator controlled the robot remotely via a tablet without making that visible to the children. In the room, there were also three boards on the wall (next, behind, and in front of the child) where the partner could locate the items needed for the implementation of the activities. There was also a mat with stuffed toys which were used to carry out some activities.

Procedure

Each of the four pupils participated in eight (8) sessions, four of which were carried out by the robot (R) and the other four by the teacher (T) in order to later compare the results of the study and to find the differences between the two interventions, those which were conducted by a social robot and those conducted by a human partner. The said tasks were designed to be appropriate for the special needs and capacities of each pupil. The data were collected based on Unstructured interviews that had been held with the parents and the specialist that dealt with each pupil outside of the regular school timetable and then each pupil’s profile was mapped out and activities were organized based on the children’s strengths and weaknesses. Each session lasted thirty (30) minutes in which every child participated in activities with specific social goals (Table 1).

Table 1.

The aims of the activities.

	Aims of the activities
	Pupil 1 and 2
1st activity	The understanding and comprehension of the four basic emotions (happiness, sadness, anger, fear) and the enhancement of eye contact.
2nd activity	The purpose of the second activity was to understand the usefulness of pointing and the appropriate operational and expressive gestures.
3rd activity	The purpose of the third activity was to recognize the four basic emotions (happiness, sadness, anger, and fear) in form of facial expression, to be capable of connecting them with social situations and follow spatial commands.
4th activity	The purpose of the fourth activity was the child's understanding of desirable behavior in a specific social situation.
	Pupil 2 and 4
1st activity	The purpose of the first activity was emotional expression and the implementation of the rules of play.
2nd activity	The purpose of the second activity was understanding the correct way to make conversation.
3rd activity	The purpose of the third activity was to enable the pupils to carry out spatial commands to express their emotions and cultivate empathy.
	Pupil 2
4th activity	The purpose of the fourth activity was to enhance the ability to make decisions and solve problems.
	Pupil 4
4th activity	The fourth activity aimed to enable him to use different social rules in different social situations and furthermore to follow instructions in new activities.

Before the task sessions started, the pupil was exposed to a familiarization phase which lasted 5 minutes. In this phase, the pupil entered the room where the robot was located; the robot remained silent for 2 minutes only blinking its eyes and looking right and left whilst waiting for the pupil to get closer. After that time, if the child would not approach, the robot would said to him “Hello”, “What’s your name?”, “My name is Daisy”. In this phase, the pupil was allowed to touch the robot and talk with it. The same steps were followed in the familiarization phase with the teacher. This phase consisted of two 5-min sessions carried out in successive weeks. In the first session, the four pupils were exposed to one of the two conditions (the teacher or the robot). In the second session, the pupils were exposed to the other condition. The two pupils who had been exposed to the teacher in week one followed by the robot in week two began the intervention with the teacher. Whereas the two pupils who followed the sequence of robot (week one) teacher (week two) started the intervention with the robot partner. After that phase and one week later, the sessions with specified learning objectives for each pupil were initiated. Each time the participants were exposed to a pre-specified activity with a sequence of steps, carried out by both the human and robotic partner with a one-week interval. They followed the same steps in order that the two different interventions differed only regarding the type of interaction partner. Each session started with the partner greeting the pupil, followed by the activities and ended with a farewell. In the session with the robot, after the final step, the robot closed its eyes and the child understood that the session was over.

The social robot Daisy

The tool that was used in the individual sessions was a non-humanoid robot in the style of a stuffed toy, called Daisy (Figure 1) which incorporates speech and expression characteristics. Its shape resembles that of a flower and within its petals two speakers are hidden which amplify the sound. A tablet is embedded in the center which provides the robot with an expressive face. Two large eyes that are capable of focusing in the center as well as looking to the left and right enable the robot to maintain contact with the child and give a different sense of interactive communication. The face is filled in with eyebrows and a big, brightly-colored mouth which opens and closes during speech. One second tablet is wirelessly connected to the first so that the researcher may operate the robot, giving it instructions to utter the appropriate words, sentences and expressions in order to manipulate the game and constantly cater to the requirements of the activity.

Data recording section

In order to monitor the conditions of the study, it was deemed necessary firstly to observe the children when they were interacting with the robot and secondly to write down personal notes while the pupil interacted both with the robot, Daisy, and with the teacher. Specifically, in the research data forms for frequency were used (Zirpoli 2005), to count the frequency of the appearance of specific behaviors within the referred period. The behaviors which were noted were firstly (1) eye contact (EC) that was measured based on the number of times during the session that the child turned either to the teacher or to the robot or looked them in the face, (2) secondly, proximity (PR) which was determined based on the number of times each child got up from his seat to approach the partner conducting the session and if he wanted to touch the partner and lastly (3) verbal interaction (VI) which examined the degree to which each child spoke spontaneously either to the teacher or to the robot. All of which were later sorted to measure the quality of the interaction (QI). Certain behaviors aiming to assess pupil involvement in the teaching process (ITP) were grouped together. These behaviors were (1) difficulty in paying attention (PA) when spoken to, measured by the number of times he did not answer the questions straight away or the question had to be repeated twice or more, or if he interrupted the partner to say something irrelevant. (2) The inability to sit (ITS) still to the extent that it disrupts an activity was measured by the number of times the pupil got up from his seat or left the defined-space of the activity (1m. from the intervention table), and (3) difficulty following instructions (FI). This last behavior was assessed based on the number of times each pupil followed the rules and the instructions of the game straight away or they had to be repeated twice or more.

Prior to the interventions, trials took place to accustom the observers with the task of completing the observation form and with the variables that would be measured. Before the session started the two observers had the two different observations forms in front of them. One form measured the quality of interaction and one measured the involvement of the pupils in the activity. Each observer also had a voice recorder. Every time that the child manifested one of the behaviors as defined it was noted on the relevant observation form. After the session each of the two observers listened to the recordings again to verify that the measurements were valid and also to add information such as the exact number of words spoken. Then the two observers compared their results, as detailed in the paragraph below, and the percentage of agreement between them was calculated, giving a reliability average of 92.1%. After each intervention, the results originating from the observation forms were calculated and two sets of values were created, one for the teacher-led (TL) procedure and one concerning the robot-led procedure (RL). After that the results were compared with the aim of finding the differences between the two different conditions (TL and RL).

Observers’ congruence

To ensure reliability of results the sessions were monitored by two observers, who carefully followed the development of the predetermined behavior-aims whilst simultaneously recording them. The observers noted each separate appearance of the defined behaviors. At the end of each session, the observation data forms were examined and compared, so that only duplicate information was included in the final results. Agreement was checked when both observers recorded the presence or absence of the behavior. Disagreement is defined in the case that the first observer wrote down the behavior and the second did not. Interobserver agreement was calculated using the number of agreements divided by the number of the total agreements and disagreements, multiplied by 100. This gave the percentage of the agreement between both observers. The reliability level ranged between 86.7% and 98.3%, with an average of 92.1% (Table 6).

Table 6.

Observers’ congruence.

Behavior type	Agreement %
Quality of interaction	TL	RL
Eye contact (EC)	86.7	87.1
Proximity (PR)	93.6	93.8
Verbal interaction (VI)	98.3	95.1
Involvement in the teaching process
Difficulty paying attention when spoken to (PA)	88.2	93.9
Inability to sit to the extent that it disrupts the activity (ITS)	91.8	96.8
Difficulty following the given instructions (FI)	88.3	91.9

Data analysis

Analyses were performed by SPSS 20. In order to test a hypothesis, a paired samples t-test (at 95% confidence level) was performed for all three components of the quality of interaction (Eye contact, Proximity, and Verbal interaction) and the involvement which is a compound connotation composed by ‘Difficulty paying attention when spoken to’, ‘Fidgeting to the extent that it disrupts the activity,’ and ‘Difficulty following the given instructions’ (Table 2).

Table 2.

Paired samples statistics for the quality of interaction between the two conditions.

		Mean	N	Std. deviation	Std. error mean
Pair 1	TL–Eye contact	4.125	16	1.544	0.3860
	RL–Eye contact	8.125	16	2.363	0.5907
Pair 2	TL–Proximity	1.000	16	2.066	0.5164
	RL–Proximity	3.125	16	1.669	0.4171
Pair 3	TL–Verbal interaction	2.063	16	1.389	0.3472
	RL–Verbal interaction	4.188	16	2.287	0.5717

TL: teacher-led procedure and RL: robot-led procedure.

Results

As it can be observed in the following tables (Tables 3 and 4), there is a statistically significant difference for all three components of the quality of interaction between Teacher-performed activity and Robot-performed activity:

Table 3.

Paired samples test for the quality of interaction.

		Paired differences					T	df	Sig. (2-tailed)
		Mean	Std. deviation	Std. error mean	95% Confidence interval of the difference
					Lower	Upper
Pair 1	TL–EC RL–EC	−4.0000	1.5492	0.3873	−4.8255	−3.1745	−10.328	15	0.000
Pair 2	TL–PR RL–PR	−2.1250	2.0290	0.5072	−3.2062	−1.0438	−4.189	15	0.001
Pair 3	TL–VI RL–VI	−2.1250	1.2583	0.3146	−2.7955	−1.4545	−6.755	15	0.000

EC: eye contact; PR: proximity; RL: robot-led procedure; TL: teacher-led procedure; VI: verbal interaction.

TL: teacher-led procedure and RL: robot-led procedure.

Table 4.

Paired samples statistics for the involvement in the teaching procedure between the two conditions.

		Mean	N	Std. deviation	Std. error mean
Pair 1	TL–Difficulty paying attention when spoken to	5.438	16	2.658	0.6644
	RL–Difficulty paying attention when spoken to	2.813	16	1.759	0.4399
Pair 2	TL–Inability to sit	3.438	16	2.449	0.6122
	RL–Inability to sit	1.563	16	1.931	0.4828
Pair 3	TL–Difficulty following the given instructions	4.188	16	2.482	0.6206
	RL–Difficulty following the given instructions	3.063	16	1.843	0.4607

TL: teacher-led procedure; RL: robot-led procedure.

• For Eye contact: The mean for Teacher is 4.125 (SD = 1.544) and for Robot it is 8.125 (SD = 2.363), p = 0.000

• For Proximity: The mean for Teacher is 1.000 (SD = 2.066) and for Robot it is 3.125 (SD = 1.669), p = 0.001

• For Verbal interaction: The mean for Teacher is 2.063 (SD = 1.389) and for Robot it is 4.188 (SD = 2.287), p = 0.000

As it can be observed in the following tables (Tables 5 and 6), there is a statistically significant difference for all three components of involvement in the teaching procedure, between Teacher-performed activity and Robot-performed activity:

Table 5.

Paired samples test for the involvement in the teaching procedure.

		Paired differences					T	df	Sig. (2-tailed)
		Mean	Std. deviation	Std. error mean	95% Confidence interval of the difference
					Lower	Upper
Pair 1	TL–PA RL–PA	2.625	2.125	0.5313	1.4925	3.7575	4.941	15	0.000
Pair 2	TL–ITS RL–ITS	1.875	1.310	0.3276	1.1768	2.5732	5.724	15	0.000
Pair 3	TL–FI RL–FI	1.125	1.668	0.4171	0.2360	2.0140	2.697	15	0.017

PA: difficulty paying attention when spoken; RL: robot-led procedure; TL: teacher-led procedure.

• For ‘Difficulty paying attention when spoken to’: The mean for Teacher is 5.438 (SD = 2.658) and for Robot it is 2.813 (SD = 1.759), p = 0.000

• For ‘Fidgeting to the extent that it disrupts the activity’: The mean for Teacher is 3.438 (SD = 2.449) and for Robot it is 1.563 (SD = 1.931), p = 0.000

• For ‘Difficulty following the given instructions’: The mean for Teacher is 4.188 (SD = 2.482) and for Robot it is 3.063 (SD = 1.843), p = 0.017

Discussion

Τhe purpose of the current study is to examine (1) Τhe difference between the intervention conducted by the teacher and the one carried out by a robotic partner with respect to the quality of interaction (eye contact, proximity, and verbal interaction) and (2) Τhe difference between the intervention conducted by the teacher and the one carried out via robotic partner regarding the involvement of students in the teaching process (whether the children pay attention when spoken to, follow the given instructions and how often fidgeting occurs to the extent that it disrupts an activity.) As for the quality of interaction, the current study demonstrates that the quality of interaction concerning children with ASD is of a higher level when they interact with a robot than with the teacher. In particular, there is a statistically significant difference for all three components of the quality of interaction between Teacher-performed activity and Robot-performed activity, showing that the quality of interaction is better for the Robot-performed activity. The high levels of eye contact exhibited in the sessions with the robot were not evident to this extent in the sessions with the teacher. Those results are in accordance with other research findings which demonstrated that children with autism display high levels of eye contact when they interact with a robot as compared to a human partner (Tapus et al. 2012, Kim et al. 2013).

All the pupils participating in the research, seemed to approach the robot more, compared with the human partner. It was observed that they stayed closer to the robot as opposed to the human, with whom they maintained the same distance. Those results confirm other research which indicates that the robot causes the pupil to display proximity behavior in the activity (Kozima et al. 2007), advocating that the children approach a robot with greater ease and interact with it more than with a human being. According to the bibliography, children with autism do not initiate social interaction as they are withdrawn (Wing 2000) and they do not seek to approach other people. The results of the current study indicate that the robot can trigger approach behavior in children with ASD. Those children who face difficulties in personal interactions approached the robot and communicated with it. This occurs because the robot manifests emotional expressions in a simple way which is easily recognizable to children with ASD. In addition, the children do not get bored or feel overwhelmed by those expressions (Kozima et al. 2007; Yun et al. 2016).

Moreover, spontaneous vocalization of the pupils was noted, even though in some sessions it did not appear frequently, and it was taken into consideration, because as the bibliography demonstrates, children on the autism spectrum find it difficult to express their emotional states and thoughts (Gonela 2006). What was observed is that the robot captured the children’s attention and in this way, they talked to it more than the teacher. That does not mean that such behaviors were not manifest in the session with teacher, but they were fewer in total and had a different tone. These results are in accordance with those of other research which finds that verbal interaction of children with autism increases, either to a greater or to a lesser extent when they interact with a robot (Kim et al. 2013, Pop et al. 2014).

Based on our analyses, children’s involvement in the teaching approach was, in comparison with the human teacher, greater in the session with the robot than the one with the teacher. Following the t-test results, the current research noted that there is a statistically significant difference for all three components of the involvement of the students in the teaching procedure between Teacher-performed activity and Robot-performed, showing higher values of difficulties for the Teacher-performed activities. According to dedicated literature, children with autism cannot concentrate on or pay attention to a task for a long time. Nevertheless, they can focus on things that they are interested in (Boucher 2009), but find it difficult to do the same when a human is talking to them or something is beyond their sphere of interest. Findings about attention in this study demonstrate that all children without exception, even if they had no difficulties in sessions with either robot or teacher, almost always displayed more attention to the robot in some activities than with the teacher. This occurred because it was not necessary to repeat the instruction for each part of the activity. Other research confirms these findings which show that the robots’ speech capacity attracts children to play with it, which in turn can help to enhance their social skills (Bharatharaj et al. 2017).

In the session with the robot, the pupils moved about and fidgeted less than they did with the human partner. Those findings are in accordance with research conducted by Stanton et al. (2008) showing that a strong mutual and genuine interaction was observed more with the robot than with a toy of the same appearance. As fewer stereotypical behaviors emerged, so the subjects spent more time fully focused on the procedure. To continue, involvement in the activity was measured also by the ability of the pupil to follow the given instructions. The ability to follow the given instructions was better in the robot performed activity, since more repetitions were needed to have the same instruction followed when it was expressed by the teacher than by the robot.

The use of the robot in the teaching procedure of children with ASD seems to be effective in enhancing specific social and communication behaviors which are not achieved by the human, but it is possible that they are less effective in other behaviors where human presence is necessary. It is possible to achieve eye contact and proximity to a greater extent in the interaction with the robot more so than with the human, nevertheless skills like mimicry and gesture and the usual functions of everyday life have been shown to a great degree in human interaction (Duquette et al. 2007). To sum up, SAR does not aim to supplant the human-specialist, but to create a different role so that with the assistance of the robot agent the intended positive outcomes can be jointly achieved in those areas where the child with autism is deficient.

Limitations of the study

This study uses an innovative tool for teaching students with ASD. Despite the use of novel apparatus and the notable improvement of particular social skills in the current investigation, there are also some limitations that have to be mentioned. The number of the participants was small and the duration of the implementation of the activities with each pupil was two months. The results, in other words, were based on the observations that were accomplished within this period. The positive results which emerge in this study may not actually be attributable to the capacity of the robot to engage the pupils with autism in the activity and the robot’s intrinsic ability to strengthen the subject’s weaknesses but instead be due to the fact that a novel approach was used which attracted their attention, a characteristic occurrence that tends to happen with anything new and where interest gradually wanes as the novelty wears off. Furthermore, the method used for data collection was qualitative for an in depth understanding of the phenomenon being studied and the procedure used was that of case study and the methodology used was observation and analysis of the materials. To conclude, more controlled studies are necessary to extract more indisputable results for the use of the robot in the treatment of the children with ASD.

Suggestions for further research

It would be beneficial to carry out an equivalent study to address the questions of the current study in the near future. The sample of the research could include a larger number of children with ASD and last for a longer period. In this way the reliability of the results would be ensured as they would be based on findings that concern far more subjects and this research could verify if those behaviors are maintained over time. Future research could also use the current robot, enriched with more capabilities. For example, it would be interesting to investigate the current study with the robot having a greater degree of autonomy. In the current study the remote operator formed sentences which he could choose at his convenience, he could also move the eyes on the robot’s face. However, if additional parts of the robot, such as the petals, could move or if the robot was positioned on a pivot which moved in accordance with the robot’s speech it is possible that different results would be obtained.

Disclosure statement

No potential conflict of interest was reported by the authors.