Parent’s AI Coach (PaiCoach): A Single-Case Experimental Study of an AI-Supported Parent Coaching System for Autistic Children

PURPOSE: Artificial intelligence (AI) may enhance the scalability and immediacy of parent coaching by providing automated, performance-based feedback within naturalistic routines. In this study, we examined the effects of Parent's AI Coach (PaiCoach), an AI-supported parent coaching system, on parent implementation of evidence-based practices in improving autistic children's communicative responsiveness during shared book reading. METHOD: Participants included four mother-child dyads with minimally verbal autistic children between 45 and 56 months of age. A concurrent multiple-baseline across participants single-case experimental design was used. Primary dependent variables included parent implementation fidelity and child communicative responsiveness. Visual analysis and Tau-U effect sizes were used to examine functional relations between the introduction of PaiCoach and changes in parent and child outcomes. RESULTS: Visual analysis showed functional relations between the introduction of PaiCoach and increases in parent implementation fidelity with corresponding improvements in children's responsiveness. Through human validation, we show that PaiCoach adopts 68.6% of AI-generated analyses without modification, whereas 31.4% required some expert editing. Social validity and usability findings indicate high parent satisfaction with PaiCoach and its good usability. Overall, findings provide preliminary evidence for the feasibility of AI-assisted parent coaching and the continued importance of human oversight in current AI-supported systems.