This pilot study explored the integration of Artificial Intelligence (AI) and Augmented Reality (AR) technologies to assist children with Autism Spectrum Disorder (ASD), who typically experience difficulties with social skills, particularly in social-emotional reciprocity. Given the scarcity of intelligent learning systems tailored to this demographic, this research developed the Augmented Reality System for Social-Emotional Reciprocity Recognition and Inference Learning (IAR-SERRI). This system leverages machine learning to analyze unfamiliar emotions and devise adaptive teaching strategies, enhancing learning opportunities for children with ASD. The IAR-SERRI system consists of three primary modules: (1) an AI Learning Module that utilizes Random Forest classification to pinpoint challenging emotions and scenarios based on user responses; (2) a Database Module that logs user details, response patterns, and outcomes from AI analyses; and (3) an AR Module that offers an interactive environment with 3D facial animations and social scenarios to aid in recognizing and inferring six fundamental emotions: happiness, fear, sadness, anger, surprise, and disgust. A pre-experimental one-group pre-test–post-test design was employed, involving five ASD children aged 6–13, to evaluate the efficacy of the IAR-SERRI system. Results showed that the system effectively identified learning challenges via the AI module and significantly improved the participants’ social-emotional abilities, with post-test scores averaging 11 points higher than pre-test scores (p = 0.043). Despite limitations due to the small sample size and IQ prerequisites, this pilot study underscores the potential of combining intelligent tutoring with AR to foster personalized, engaging learning experiences for individuals with ASD. Future research may extend this approach to include lower-functioning individuals with ASD.