CAMARONERAS REGISTRADAS Y APROBADAS INSTITUTO NACIONAL DE PESCA - I.N.P
CALLE 25 DE JUNIO S/N ENTRE NAPOLEON MERA Y CALLEJON BANAO -
Children in the robot group spent the greatest amount of time attending to the robots used in the study (see Figures 4-3B and 4-4B). This trend was seen during a majority of the conditions within the training session (see Figure 4-4B). We had hypothesized that children would engage in higher levels of attention directed towards the robot during the early training sessions given their intrinsic interest in robots and technology (Diehl et al., 2011; Robins et al., 2006). However, we expected that across training weeks the robot would act as an effective mediator for interactions between children and their social partners, and children would increase the length of attentional episodes directed towards their interaction partners. In contrast to our hypotheses, children continued to devote maximum attention towards the robot across training sessions, which might have severely restricted children’s opportunities to engage with their social partners. To the best of our knowledge, ours is the first study to systematically compare the effects of prolonged 8-week human vs. robot-mediated therapies on the social communication skills of children with ASDs. Our findings suggest that the music group demonstrated greater social attention levels compared to the robot group. Our findings do not fit with the current literature on effects of robotic
human vs. robot mediator suggested that children with ASDs found robots engaging, and directed greater attention to the robot compared to the human mediator (Bekele, Crittendon, Swanson, Sarkar, & Warren, 2013; Duquette et al., 2008; Ismail, Shamsudin, Yussof, Hanapiah, & Zahari, 2012a). We also observed that children directed greater attention towards the robot at the cost of attention to their social partners. However, it is important to note that the ultimate goal of all robot-child interactions is not to enhance children’s interactions with the robot, but rather to use the robot as a mediator to engage in interactions with social partners (Ricks & Colton, 2010). In line with this motivation, several anecdotal studies have reported improvements in shared attention, imitation, turn taking, and joint attention skills in children with ASDs following interactions with a robot in a triadic context (Kozima et al., 2007; Robins et al., 2004; Robins et al., 2005; Stanton et al., 2008; Warren et al., 2013; Werry et al., 2001). However, only two of these studies assessed the effects of repeated sessions of robot-child interactions over a prolonged period of time (Kozima et al., 2007; Robins, Dautenhahn, Te Boekhorst, & Billard, 2004). Kozima and colleagues introduced their creature-like robot, Keepon, within the playroom of a day care center for children with ASDs, but unlike our study, children were free to approach and interact with the robot at any time during their routine activities based on their will (Kozima et al., 2007). In a more structured interaction protocol, Robins and colleagues observed the reactions of 4 children with ASDs over 100 sessions of interactions with robots; however, on an average each trial lasted between 3 and 5 minutes and was terminated as soon as the child showed boredom with the context (Robins et al., 2004). We argue that a reasonable test of the potential utility of robots as therapy tools in autism should be based on the current standards for autism interventions. Current autism interventions are very intensive and are provided for around 30-40 hours per week (Landa, 2007; Vismara & Rogers, 2010). Hence, we evaluated the effects of an intense 16-session protocol of robot-child interactions provided over 8 weeks on children’s social communication skills which suggested that the robot in fact limited children’s opportunities for social interactions.
Although social attention levels in the robot group were significantly lower than those seen in the music group, they were nevertheless higher than the levels seen in the standard-of-care academic group (see Figures 4-3A-C and Table 4-3). Similar to the design of the music group, children in the robot group also engaged in dual and multilimb synchrony-based games. The joint action-based activities might have promoted social monitoring of the adult model (see Figure 4-5). Moreover, while practicing each movement sequence, the first trial involved the robot, the child, and the model all moving together, but following this, we encouraged children to focus on synchronizing with the adult model in the absence of robotic movement. This phase might have encouraged children to monitor the actions of the adult model. We also noticed that unlike the music group, the robot group engaged in greater responsive compared to spontaneous social attention episodes with their social partners (see Figure 4-5 and Table 4-5). We observed that given children’s visual fixation on the robot, it was very hard to get them to disengage from the robot and attend to their social partners. The trainers and models had to repeatedly bid children to attend to them and monitor their actions which might have led to the higher amounts of responsive attention in this group.
In terms of training-related changes in attention patterns, the robot group demonstrated a reduction in attention towards the robot across training sessions with a concurrent increase in attention towards elsewhere and non-social objects (see Figure 4-3B). We think that these findings reflect a reduction in engagement and progressive boredom with the context. Similar findings were found in an 18-day field trial that used an interactive humanoid robot, Robovie, in first and sixth grade TD children to teach them words. The authors found a sharp reduction in interest towards the robot during the second week of interaction in some of the children (Kanda, Hirano, Eaton, & Ishiguro, 2004). Our own previous work with TD children using a smaller humanoid robot, I-sobot, also suggested that over time the limited capabilities of the robot could not sustain children’s engagement across training sessions (Srinivasan et al., 2013). In the current study, though we used state-of-the-art Nao robots, we think that the limitations of our current robotic technology might have led to decrease in engagement over time. The Nao robot has 25
degrees of freedom, but the movement repertoire of the robot is still limited compared to that of a child. Moreover, the robot’s movements are noisy and much slower than those of a human. Once triggered, the robot cannot adapt its movements to those of the child during an ongoing movement sequence. In terms of social interaction capacities, the robot’s verbiage was unclear and children had difficulty understanding the robot’s speech. Moreover, even though we used the online speech capabilities of the robot, there was some time lag between the child’s verbiage and the robot’s response since the adult trainer had to manually type in the responses of the robot. Hence, the lack of contingent responding and dynamic adaptive capacities severely limited the robot’s ability to become an effective mediator between the child and their social partners. Instead, the robot promoted visual fixation towards itself and non-social
attention towards objects at the cost of interactions between children and their social partners.