SYNTHETIC THERAPEUTIC SCAFFOLDING FOR SKILL GENERALIZATION IN AUTISTIC CHILDREN

Autistic children may learn a functional response during instruction without using it when people, settings, or demands change. Synthetic Therapeutic Scaffolding (STS) is proposed as a clinician-governed approach to this problem. An AI avatar would provide brief, restricted rehearsal of an individualized skill, followed by practice with people and gradual removal of synthetic support. A targeted narrative synthesis identifies relevant evidence from conversational agents, virtual reality, and implementation platforms while examining the limits of its application to children. The proposed model connects varied practice and prompt fading with independently observed human transfer. Its requirements include accessible communication, ongoing child assent, controlled outputs, adult intervention, and defined privacy and safety procedures. A concurrent multiple-baseline study would evaluate the supervised package, followed by a comparison with human-led rehearsal matched for time, opportunities, and adult attention. Generalization is the primary outcome; maintenance, acceptability, fidelity, burden, and adverse events require separate assessment. Existing evidence supports investigation of this approach but does not establish its effectiveness or safety. Participatory design, technical testing, and feasibility work must precede efficacy evaluation.