Cognitive-Adaptive AI Framework for Behavioral Crisis Prediction in Children with Autism

Abstract

Behavioral crises in children with autism often come on sharply and are difficult for caregivers and clinicians to intervene in a timely way. In this study, the proposed cognitive-adaptive artificial intelligence (AI) system integrates reinforcement learning (RL) with multimodal Internet of Things (IoT) sensing and real-time emotional context modeling to forecast the escalation of a crisis. Pilot studies on 12 weeks, comprising a sample of 60 autistic children, showed an improvement of 23 percent in detecting early crisis and a reduction of 17 percent in false positives as compared to baseline models. The framework constantly acquires contextual information patterns - based on physiological, behavioral, and environmental information - to change its decision policies in real-time. The findings demonstrate the revolutionary aspect of humanistic, data-intensive AI systems in proactive management of autism care.