Answers are grounded in this paper and cite their sources.
Objectives This scoping review aims to systematically map how Artificial intelligence (AI)-driven reinforcement learning is employed to deliver personalized education for individuals with autism, with a focus on cognitive development.Methods Guided by Arksey and O’Malley’s framework and the PRISMA-ScR guidelines, a comprehensive search was conducted across seven databases, including PubMed, Scopus, and IEEE Xplore. Studies were included if they involved individuals with autism spectrum disorder, implemented RL-based systems in educational or cognitive training contexts, and reported outcomes related to personalization or cognitive enhancement. Data were synthesized using descriptive statistics and thematic analysis.Results Nineteen studies were included, encompassing a wide range of reinforcement learning (RL) approaches (e.g. Q-learning, deep Q-networks, actor–critic models) deployed across heterogeneous platforms such as mobile applications, socially assistive robots, virtual reality environments, and wearable technologies. While many studies reported improvements in learner engagement and selected cognitive outcomes (e.g. attention, executive function, memory, and social cognition), the evidence base was characterized by substantial methodological variability, small sample sizes, short intervention durations, and limited use of standardized outcome measures. Comparative effectiveness across platforms and algorithms was rarely examined, and ethical, cultural, and implementation challenges were inconsistently reported.Conclusions Although RL-based educational systems show promise for supporting individualized and adaptive learning in autism education, current evidence remains fragmented and largely exploratory. The positive trends observed should therefore be interpreted with caution. Future research should prioritize rigorous study designs, longitudinal evaluation, transparency in algorithmic decision-making, and culturally responsive implementation to strengthen the reliability and translational value of AI-driven personalization for neurodiverse learners.
No takes yet. Share an insight, caveat, or question.
Dhar et al. (2026) studied this question.
Synapse has enriched 10 related papers. Consider them for comparative context: