A Game-Based Eye-Tracking Task for Inclusive Educational Assessment in Children with Autism Spectrum Disorder and Dyslexia: An Exploratory Study

Background: Inclusive educational assessment requires tools that capture how learners with diverse cognitive profiles engage with tasks. This study examines whether a game-based, gaze-contingent eye-tracking task can generate relevant indicators for inclusive assessment of autistic children and children with dyslexia. Methods: Forty-five children (15 autistic, 15 with dyslexia, 15 neurotypical) completed a Unity 3D eye-tracking task assessing implicit joint attention, inhibitory control, and task engagement. Six gaze-based metrics were analyzed alongside standardized measures of joint attention, theory of mind, working memory, and processing speed. Group differences in the eye-tracking metrics were examined using permutation-based analysis of covariance controlling for chronological age and Full-Scale IQ, and associations with clinical measures were examined using a small set of pre-specified, Holm-corrected Spearman correlations. Results: Significant group differences were observed for five of the six metrics (partial η2 = 0.33–0.81). The autistic group showed the highest values on difficulty-related metrics and differed significantly from both the neurotypical and dyslexia groups on nearly every measure, while the neurotypical and dyslexia groups did not differ significantly from each other. One correlation remained significant after correction for multiple comparisons: task efficiency in the autistic group was linked to a behavioral correlate of theory of mind. Conclusions: Gaze-based indicators within a game-based task can differentiate autistic children from neurotypical and dyslexic peers and show a preliminary behavioral correlate of theory of mind in autism specifically. These findings offer narrow, preliminary support for the value of process-oriented tools for inclusive educational assessment. Future research should examine feasibility, validity, and psychometric stability in larger, matched samples and in authentic school contexts.