Effectiveness of Artificial Intelligence-mediated pedagogical strategies on emotional self-regulation in children with autism: a systematic review

Abstract

Emotional dysregulation can increase vulnerability to anxiety in Autism Spectrum Disorder (ASD). Given the limitations of traditional education in addressing this particularity, Artificial Intelligence (AI) offers real-time adaptation systems. The objective of this study was to evaluate the level of efficacy of AI-mediated pedagogical strategies in contrast with traditional teaching methods for strengthening emotional self-regulation in children with ASD. He Scopus and Web of Sci-ence databases were explored for the 2016–2026 period under the PICO methodologi-cal framework and PRISMA 2020 guidelines; additionally, the protocol was registered in OSF. Based on the eligibility criteria, the initial 177 records were filtered, yielding 13 studies for qualitative synthesis and, additionally, 4 controlled trials to verify efficacy. The RoB 2 and ROBINS-I tools helped explore the risk of internal bias. Due to its capacity to redistribute cognitive or affective load, artificial intelligence can be used as a neuroaffective tutor. Compared to traditional interventions, augmented reality reported a reduction in irritability of up to 90%. Anxiety levels are significantly re-duced through virtual immersion. Robotics, when prompting prosociality and positive emotions with p = 0.0076, enables an improvement compared to human control types. A compliance rate higher than 91% was achieved, favored by the "friendship" consid-eration that the children established with the devices. Artificial intelli-gence-based intervention has proven promising compared to some traditional inter-ventions in modulation due to its algorithmic predictability, which prevents sensory overload. Rather than marginalizing the educator, these technologies reconfigure their role in the strategic orchestration of the classroom.