Difficulties with play skills have been well documented in children with Autism Spectrum Disorder (ASD; Wuff, 1985; Baron-Cohen, 1987; Lewis and Boucher, 1988; Jarrold et al., 1993; Charman et al., 2000; Williams et al., 2001). These difficulties are seen in sensory motor play, manipulative play, physical play, pretend play, and social play (Boucher, 1999). Consistent with this view, many studies have focused on teaching play skills to children with ASD (Jung and Sainato, 2013). Previous intervention studies have used video and live modeling (Jahr et al., 2000; MacDonald et al., 2005), pivotal response training (Stahmer, 1995; Thorp et al., 1995), activity schedules (Morrison et al., 2002; Machalicek et al., 2009), or social stories (Barry and Burlew, 2004). Researchers have also combined these strategies with contingent reinforcement (Jung and Sainato, 2013). These studies have found training increases engagement in appropriate play behavior and cooperative play in children with ASD. On the other hand, few studies have examined the effectiveness of robotic device use in teaching play skills to children with ASD, although robotic devices can automatically and immediately reinforce appropriate play behavior.

Robotic devices have been used to increase social-communication behaviors, such as joint attention (Warren et al., 2015; Simut et al., 2016) and imitation (Duquette et al., 2008), in children with ASD. These studies have focused on the use of both humanoid robots and non-humanoid toy-like robots, such as KASPER (Robins et al., 2009), Keapon (Costescu et al., 2015), NAO (Huskens et al., 2015; Warren et al., 2015), Probo (Simut et al., 2016), Robota (Billard et al., 2007), or Tito (Duquette et al., 2008). However, these robots mainly provide feedback as a result of the behavior of a child. Given that the facilitation of social play involves two people using toys, it may be necessary to consider directly providing feedback as a result of the behavior of both a child and the other individual.

Paired robotic devices might encourage cooperative behaviors, such as turn taking (Nunez et al., 2016). In this approach, remotely connected paired devices provided feedback separately as a result of child’s own behavior as well as the other individual’s behavior. Huskens et al. (2013) suggested robotic devices should be deployed as mediators to promote social interaction between a child with ASD and another individual. However, to our knowledge, no studies have examined how automatic feedback via paired robotic devices affects social play behaviors. In addition, as Diehl et al. (2012) have pointed out, most studies using robots for children with ASD have not used an experimental design, such as an experimental group design or single subject experimental design.

When considering play behaviors, we need to recognize two types. First are those related to social play, such as ball contact and looking at a therapist’s ball (Bass and Mulick, 2007). Second are those related to social interaction, such as eye contact and positive affect. The purpose of the current study is to examine whether automatic feedback via paired robotic devices facilitates social play behaviors in children with ASD, and to explore how social play relates to social interaction.

If the paired robotic devices can immediately provide automatic feedback contingent on child’s social play behaviors, it is possible that automatic feedback increases the social play behaviors in children with ASD. Therefore we hypothesized the following relationship between behavior contribution and feedback:

It is possible that social play will also facilitate social interaction, and then we could expect that:

Alternatively, it is also possible that social play directs the child’s attention away from the therapist toward the activity, and thus we could expect that:

To directly test these hypotheses a single AB design was used in a pilot study to make inferences about the effects of feedback made by colored lights and vibration via paired robotic devices on social play behaviors in three boys with PDDs. In this experiment, we used a rapidly changing reversal design with the same experimental condition as the pilot study. By using this experimental design, we further evaluated whether and what types of social play behaviors are facilitated by the feedback provided by remotely connected paired devices in children with PDDs.

This study investigated the effects of automatic feedback in the form of colored lights and vibration produced via paired robotic devices, COLOLO, in social play and interaction in children with ASD. The frequency of ball contact and looking at the therapist’s ball were higher in the with automatic feedback condition than in the without automatic feedback condition, supporting Hypothesis 1. On the other hand, the frequencies of eye contact and positive affect for all children with ASD did not consistently increase or decrease in the with automatic feedback condition, thus the results indicated lack of support for both Hypothesis 2a and Hypothesis 2b. Therefore, when using the paired robotic devices, the children with ASD appear to have exhibited increases in social play behaviors using toys and but not increases in behaviors associated with social interaction.

Considering ball contact, Hypothesis 1 was positively supported. The findings are in lines with one of the pioneering works, which has demonstrated that a spherical mobile robot, Roball, may increase a child’s interaction with a ball by providing automatic feedback consisting of motion, messages, sounds, and an illuminating interface (Michaud et al., 2005). This suggests that automatic feedback of vibration (tactile stimulus) might function as a reinforcer for ball contact behavior. However, we also used light feedback (visual stimulus). There is a possibility that light feedback also functions as a reinforcer for child’s ball contact. Therefore, in a future study, we would evaluate which modality of feedback has a stronger effect on increasing ball contact.

Considering frequency of looking at the therapist’s ball, the first Hypothesis was also positively supported. This indicates that attention to shared play materials might be increased by light feedback via paired robotic devices. Although we used vibration feedback, this feedback was not contingent upon the child looking at the therapist’s ball, but contingent on the child looking at his own ball. Thus, light feedback provided via remotely connected paired devices may increase attention to the play materials of peers in children with ASD.

Concerning Hypotheses 2a and 2b of the current study, our results did not support either of these hypotheses. Neither the child’s eye contact nor their positive affect consistently increased as a result of the feedback in the form of light and vibration. The result can be easily interpreted because the feedback was not contingent upon the child’s responses. In addition, however, increases in eye contact and positive affect were observed in the without automatic feedback condition for two children. A potential explanation for this outcome could be the frequency of the reinforcement provided by the therapist. As far as ecologically validity is concerned, in the procedure of this experiment, the therapist provided verbal/physical praise for the child’s eye contact and positive affect throughout the session. This may have led to increased opportunity for praise for the therapist in the without automatic feedback condition in which the frequencies of child’s ball contact was lower. To improve this aspect of the intervention, we recommended that future studies include the combined use of other wearable devices, such as an eye tracker (Ye et al., 2012) or a face reader device we have developed for detecting smiles from facial EMG signals (Gruebler and Suzuki, 2014), in order to provide contingent feedback for child eye contact and/or positive affect.

There were several limitations to the current study. First, we used a single subject experimental design with three children with ASD in this study, and this limits the generalizability of the results to the larger population due to limitations inherent in single subject experimental designs, such as absence of statistical analysis and inference. Further studies are required, including use of a group experimental design with larger sample sizes. Second, although we used an ABAB design to minimize carryover effects, because the experiment sessions were administered across 2 days, we were not able to eliminate ordering and novelty effects. It is possible that the novelty of the interaction affected the increase in the dependent variables on the 1st day (first set of AB trials) and on the 2nd day (second set of AB trials), due to the time that has elapsed between the first and the second session. Further studies must seek to eliminate ordering and novelty effects through blocked and longitudinal study designs. Third, we need to be cautious about interpreting the observed increases in children’s ball contact and looking at the therapist’s ball as the result of automatic feedback functioning as a reinforcer. It was unclear whether automatic feedback functioned as an antecedent stimulus or a reinforcer for children’s ball contact and looking at the therapist’s ball. Further research is warranted to identify the function of automatic feedback via the implementation of a yoked condition. Fourth, we only used the feedback of light (visual stimulus) and vibration (tactile stimulus). Future studies will be required to use other modalities, such as sound (auditory stimulus).

Nevertheless, the current findings establish that feedback via paired robotic devices can facilitate some aspects of social play behaviors in children with ASD, whereas previous studies have focused on examining differences between a human and a robot as an interaction partner (e.g., Costescu et al., 2015; Srinivasan et al., 2015; Simut et al., 2016), or investigating the effects of teaching by the robot (e.g., Billard et al., 2007; Warren et al., 2015). As Huskens et al. (2013) have suggested, it would be interesting to see more studies on this topic; in other words, there is a wide range of necessities for further investigation. While we are hopeful that clinical applications of paired robotic devices may demonstrate significant enhancement of social play for children with ASD at an early developmental stage, it is important to note that future research should reveal both whether and how the paired robotic devices contribute to increasing various forms of social play behaviors in children with ASD.

This study was carried out in accordance with the recommendations of Keio University’s Institutional Review Board with written informed consent from all parents of participants.

SM, EN, MH, JY, and KS designed the research. SM and EN performed the research. SM analyzed the data and wrote the article.

The authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.