Artificial intelligence in neurodiverse education: A PRISMA-ScR systematic scoping review (2017–2025)

AbstractArtificial intelligence (AI) is increasingly used to support neurodiverse learners and offers new opportunities for early identification, personalized instruction, and cognitive accessibility. However, evidence remains dispersed across disciplines and educational settings. This systematic scoping review maps how AI has been designed, implemented, and evaluated for learners with autism spectrum disorder (ASD), attention-deficit/hyperactivity disorder (ADHD), dyslexia, and related neurodevelopmental differences. Following PRISMA-ScR, studies published between 2017 and 2025 were systematically searched across major education, computing, and interdisciplinary databases and charted by population, setting, AI modality, purpose, and outcomes. Included studies were synthesized into five domains: (1) diagnosis and early identification, (2) learning enhancement, (3) personalized intervention and assistive technologies, (4) cognitive accessibility and inclusive design, and (5) stakeholder perspectives. Across domains, AI-supported approaches commonly reported improved screening performance, more adaptive pacing and feedback, and increased learner engagement and autonomy through assistive and neuroadaptive tools. At the same time, recurring risks were identified, including dataset bias, privacy and surveillance concerns, limited real-world validation, and insufficient inclusion of neurodivergent voices in design and evaluation. Overall, the literature suggests that AI is most promising when implemented to advance inclusion and learner agency rather than automate decision-making. Future work should prioritize participatory, neurodiversity-affirming design, transparent model development, and robust evaluation in authentic educational contexts.