28678
Robot-Assisted Learning-By-Teaching Pedagogy for Improving Social Skills
An increasing number of children with autism spectrum disorder (ASD) are attending mainstream schools. The majority of these children receive regular interventions that provide opportunities to practice social skills within group settings, one-on-one intervention with an educator or targeted therapy with a specialist. However, it is still uncommon for a child with autism to have the opportunity to develop the social skills acquired during interventions and foster a deep, generative understanding of these skills by engaging others in peer tutoring. The learning-by-teaching pedagogy offers these children a number of potential social benefits that may facilitate long-term learning and improved outcomes but there are numerous challenges make the deployment of peer tutoring difficult in practice.
Objectives: In this first study, we evaluated the potential benefits of a novel robot-assisted, learning-by-teaching approach for advancing social skills using a set of well-defined social skills tasks for school-aged children with ASD. We investigated the feasibility and comparative benefits of implementing the paradigm in a child-to-robot context versus a child-to-human-confederate scenario. Further, we examined the relative differences in engagement between conditions and potential performance implications correlated to higher engagement in each study condition.
Methods: Ten children between 6-10 years of age (m=8.75 yrs.) who were diagnosed with ASD based on evaluation with gold standard DSM-5 criteria and met or surpassed a minimum threshold (>=80) for IQ, as determined by the Differential Ability Scales-General Conceptual Ability (DAS-GCA) assessment (m=107.5), were invited to participate in this randomized control trial. We empirically evaluated performance on a set of social skills tasks and engagement level during two study conditions: robot and human-confederate. We computed Pearson bi-variate correlations to evaluate potential connections between IQ, performance and engagement and performed Fisher’s z-transformation on correlations to investigate the significance of differences between resulting correlations. Finally, we conducted multiple linear regression to examine the predictive value of age, IQ and engagement for predicting performance in both study conditions.
Results: Results show that 80% of study participants performed better in the robot condition (mean performance: robot=63%, confederate=37%) and 90% of all participants were significantly more engaged in the robot condition (mean engagement: robot=61%, confederate=32%). Verbal IQ and IQ were significantly correlated with overall performance in the confederate condition (r=0.794, p<0.05 and r=0.726, p<0.05, respectively), but no correlation resulted in the robot condition between Verbal IQ/IQ and overall task performance or in any individual script. Multiple linear regression was conducted to predict overall performance from IQ, age and engagement. Although age, IQ and overall engagement did not predict performance in the confederate condition, these variables statistically significantly predicted overall performance in the robot condition, F(3,6)=5.399, p<0.05, R2=0.730.
Conclusions: Results from this study confirm the feasibility of this new robot-assisted intervention and support the significant potential of this approach for promoting social skills. These findings further underscore the attentional value of employing a robot in an autism intervention and the potential therapeutic benefit of employing a social robot as a peer student.