Predictive Roles of Psychometric and Learner Characteristics on the Learning of Students with Autism
TL;DR
Results suggest that although students’ preferred learning modalities and their executive functioning have the greatest explanatory power, predictor value, and correlation to their performance (rubric and quiz scores) in unplugged computing lessons, the Elastic Net with leave-one-student-out cross validation did not successfully predict rubric scores or quiz performance, and the correlations were not statistically significant.
Abstract
Very few studies have designed and implemented coding curricula for students with special needs, and even fewer have explored computing for students with autism [1, 6, 8]. This study is the first to investigate unplugged computing specifically for students with autism. To this end, 10 accessible unplugged computing lessons were designed, developed, and implemented for nine sixth-grade students with Autism Spectrum Disorder (ASD) at a school for students with ASD. To protect participant confidentiality, individual-level psychometric and learner characteristic data were replaced with a multivariate synthetic dataset (n = 30) in this study. At the start of the study, the students completed surveys for their learner characteristics, which included BRIEF-2, State-Trait Anxiety Inventory for Children (STAIC), Locus of Control, and learning preference (visual/auditory/kinesthetic; individual vs. group). During the intervention, students completed a quiz for each lesson and teachers rated students’ learning using one or more rubrics per lesson. Baseline performance and individual learning trajectories were modeled using linear mixed-effects models and evaluated predictive validity with Elastic Net under leave-one-student-out cross-validation. Results suggest that although students’ preferred learning modalities and their executive functioning have the greatest explanatory power, predictor value, and correlation to their performance (rubric and quiz scores) in unplugged computing lessons, the Elastic Net with leave‑one‑student‑out cross validation did not successfully predict rubric scores or quiz performance, and the correlations were not statistically significant.