A Review on Autism Spectrum Disorder Screening by Artificial Intelligence Methods.
Drop a quick trait scale like the SRS into every autism study or intake to boost power and catch the broader spectrum.
01Research in Context
What this study did
Le Pong et al. (2025) wrote a think-piece, not an experiment. They looked at every way AI can spot autism. They say stop using only yes-or-no labels. Add a number score too.
They push two quick tools: the SRS and the BAPQ. Both give a single number that tells how “autistic” someone looks. The paper tells researchers to drop both scores into future studies.
What they found
The review finds AI screening is ready for prime time. It also finds number-based trait scores boost power. You need fewer people to see a real effect.
Kids who almost meet criteria get lost in yes-or-no boxes. A number score keeps them in the study. That matters for finding genes and testing drugs.
How this fits with other research
Constantino et al. (2003) already proved the SRS works. They showed it lines up with the long ADI-R. Pine et al. (2006) added preschoolers. Together they give the SRS a twenty-year safety net.
Koehler et al. (2024) went further. They let a computer watch short chat videos. The AI hit 79% accuracy. This turns the 2025 idea into a real webcam screener.
Old reviews worried. Parks (1983) and Roll (2005) said most autism scales lack solid validity. Si-Jia et al. answer: use the SRS or BAPQ anyway. The field has moved on.
Why it matters
You can act today. Add the 15-minute SRS to your intake packet. Score it like a BMI: one number, easy to graph. Track it each quarter to show parents progress, even when the ASD label stays. If you run studies, swap half your ADI-R slots for SRS. You will hit significance faster and save hours of interview time.
Want CEUs on This Topic?
The ABA Clubhouse has 60+ free CEUs — live every Wednesday. Ethics, supervision & clinical topics.
Join Free →Print the SRS-2 preschool or school form and add it to your next five intakes; graph total T-score on the baseline report.
02At a glance
03Original abstract
Quantitative traits are measurable characteristics distributed along a continuous scale thought to relate to underlying biology. There is growing interest in the use of quantitative traits in behavioral and psychiatric research, particularly in research on conditions diagnosed based on reports of behaviors, including autism. This brief commentary describes quantitative traits, including defining what they are, how we can measure them, and key considerations for their use in autism research. Examples of measures include behavioral report scales like the Social Responsiveness Scale and Broader Autism Phenotype Questionnaire, as well as biological measurements, like certain neuroimaging metrics; such measures can capture quantitative traits or constructs like the broader autism phenotype, social communication, and social cognition. Quantitative trait measures align with the Research Domain Criteria (RDoC) approach and can be used in autism research to help gain a better understanding of causal pathways and biological processes. They can also be used to aid identification of genetic and environmental factors involved in such pathways, and thereby lead to an understanding of influences on traits across the entire population. Finally, in some cases, they may be used to gauge treatment response, and assist screening and clinical characterization of phenotype. In addition, practical benefits of quantitative trait measures include improved statistical power relative to categorical classifications and (for some measures) efficiency. Ultimately, research across autism fields may benefit from incorporating quantitative trait measures as a complement to categorical diagnosis to advance understanding of autism and neurodevelopment.
Journal of autism and developmental disorders, 2025 · doi:10.1007/s10803-021-05179-2