Assessment & Research

Identifying children with autism spectrum disorder based on their face processing abnormality: A machine learning framework.

Liu et al. (2016) · Autism research : official journal of the International Society for Autism Research 2016
★ The Verdict

Eye-tracking plus machine learning hit 88% accuracy for autism detection, but newer quiz and audio tools now do the job cheaper and better.

✓ Read this if BCBAs who assess or refer young children for autism.
✗ Skip if Clinicians already using the newer 97-99% tablet screeners.

01Research in Context

01

What this study did

Liu et al. (2016) filmed kids' eyes while they looked at faces. The team fed the eye-movement data into a computer. The goal was to see if the machine could spot autism from gaze patterns alone.

No extra tasks or teaching. Just quiet watching and recording where the eyes went.

02

What they found

The program sorted autistic from non-autistic kids with 88% accuracy. That beat chance by a wide margin. It hints that odd face scanning can serve as an early red flag.

03

How this fits with other research

Marsack-Topolewski et al. (2025) now tops this score. Their four-model ensemble, tested from toddlers to adults, hits 97-99% accuracy using a short parent quiz. The new tool replaces pricey eye trackers with a tablet and still beats the 2016 result.

Yin et al. (2026) swapped eyes for ears. Ten minutes of baby sounds fed into a support-vector machine reached 93% accuracy. Both studies keep the machine-learning spirit but trade hardware for cheaper data.

Bone et al. (2015) warned that early ML autism claims often fail when retested. The 2016 paper answered by adding stricter stats and cross-checks, showing the field is learning from past flops.

04

Why it matters

You may soon screen without a long waitlist. If a quick eye-track or mic check can flag risk, you can refer families before red flags grow. Watch for tablet apps and audio tools moving out of the lab. Pilot one in your clinic and track how many kids you catch earlier.

Free CEUs

Want CEUs on This Topic?

The ABA Clubhouse has 60+ free CEUs — live every Wednesday. Ethics, supervision & clinical topics.

Join Free →
→ Action — try this Monday

Try a tablet-based autism risk quiz on your next intake and note if it matches your clinical impression.

02At a glance

Intervention
not applicable
Design
other
Population
autism spectrum disorder, neurotypical
Finding
positive

03Original abstract

The atypical face scanning patterns in individuals with Autism Spectrum Disorder (ASD) has been repeatedly discovered by previous research. The present study examined whether their face scanning patterns could be potentially useful to identify children with ASD by adopting the machine learning algorithm for the classification purpose. Particularly, we applied the machine learning method to analyze an eye movement dataset from a face recognition task [Yi et al., 2016], to classify children with and without ASD. We evaluated the performance of our model in terms of its accuracy, sensitivity, and specificity of classifying ASD. Results indicated promising evidence for applying the machine learning algorithm based on the face scanning patterns to identify children with ASD, with a maximum classification accuracy of 88.51%. Nevertheless, our study is still preliminary with some constraints that may apply in the clinical practice. Future research should shed light on further valuation of our method and contribute to the development of a multitask and multimodel approach to aid the process of early detection and diagnosis of ASD. Autism Res 2016, 9: 888-898. © 2016 International Society for Autism Research, Wiley Periodicals, Inc.

Autism research : official journal of the International Society for Autism Research, 2016 · doi:10.1002/aur.1615