Roughly one in 100 children worldwide has autism spectrum disorder, and many families wait months for an assessment that decides whether early intervention starts in time. Researchers at the Hong Kong University of Science and Technology built an AI platform meant to shorten that wait.
Called ASD Super Agent, the system combines early screening, personalized intervention advice and continuous family follow-up in one tool aimed at parents rather than clinicians. The team tested it with nearly 500 families and visited special education centers and psychiatric hospitals across Hong Kong, the Greater Bay Area and Beijing.
Home video instead of a clinic queue
Parents record short videos of a child’s everyday activity, complete standardized questionnaires and enter basic developmental details. The platform returns a risk assessment and a report families can bring to a professional evaluation.
In preliminary testing the screening model reached an area under the curve of 0.92, where values near 1 indicate stronger separation between children with and without risk. The researchers stress the system remains under development and needs clinical validation before deployment.
Twenty intervention models plus local practice
Recommendations rest on a 20+1 framework. The 20 covers recognized approaches embedded in the system, including Applied Behaviour Analysis, Naturalistic Developmental Behavioural Interventions and TEACCH structured teaching. The +1 lets therapists and special education providers add their own methods to the knowledge base.
Parents describe challenges such as limited eye contact or repetitive behavior and receive tailored suggestions with academic references.
It handles voice interaction in Cantonese, Putonghua and English, and keeps posture processing on-device to limit data leaving the phone.
A million records behind the model
The team has assembled a multimodal database of video, posture and behavioral data now exceeding 1.1 million records. The platform is positioned as an aid to assessment, not a replacement.
