A 2025 multi-site study led by Dawson and colleagues evaluated a mobile app called “SenseToKnow,” designed for accessible, early autism screening in toddlers. Caregivers used the app at home, where it presented short videos and a bubble-popping game to elicit a range of social and attentional responses. These behaviours were analysed using computer vision and machine learning. Among 620 toddlers aged 16–40 months, the app showed high accuracy in identifying autism, performing comparably to established screeners. The results suggest digital approaches as a promising way to make early identification more accessible, objective, and equitable for all families.
Addressing Gaps in Early Autism Identification
Most early screening relies on parent questionnaires, such as the Modified Checklist for Autism in Toddlers (M-CHAT). While helpful, these checklists can produce variable results in real-world settings and may be less accurate for children from diverse backgrounds. Recent advances in digital health offer opportunities to enhance early identification. Computer vision and machine learning can objectively capture differences in attention, social communication, and play—without relying solely on parent report. Tools that use mobile devices can help reach more families, including those who face barriers to in-person evaluation.
Digital tools can help make early autism identification more accessible and responsive to family needs.
What Was Studied and What Was Found
Dawson and colleagues developed and tested the SenseToKnow mobile app, designed for caregivers to use at home with toddlers aged 16 to 40 months. The app features a series of short, engaging videos and a bubble-popping game, inviting children to show their interests, responses, and ways of interacting. The device’s camera records these moments, and algorithms analyze features like gaze, facial expression, movement, and how a child responds to their name. The app then calculates an individualized “autism likelihood” score using models trained on video data labeled by expert clinicians, drawn from children with and without autism, to detect behavioral patterns linked to early autistic development.
In this study, 620 toddlers and their families took part across multiple sites, representing a broad range of backgrounds. After the home assessment, all children received a thorough evaluation from clinicians using gold-standard measures like the Autism Diagnostic Observation Schedule (ADOS-2), with clinicians making diagnostic decisions based on holistic information.
Key findings: 1. The SenseToKnow app showed strong accuracy in recognizing autistic development, with an area under the curve (AUC) of 0.92, sensitivity of 83.0%, and specificity of 93.3%. 2. The app performed consistently across different ages, genders, and racial/ethnic backgrounds. 3. Notably, the digital screener identified some autistic children who were missed by established screeners like the M-CHAT. 4. Combining the app’s scores with traditional questionnaires further improved identification. 5. Caregivers found the app straightforward and quick, with most completing it in less than 20 minutes.
The authors highlight that digital behavioral tools can offer scalable, objective data to supplement parent and professional observations. By enabling screening at home, apps like SenseToKnow may support families in a wide range of communities and help reduce barriers to earlier recognition and support.
App-based tools empower families to participate in screening in comfortable, familiar settings.
Next Steps for Research and Practice
Dawson and colleagues encourage further research to understand how digital screening fits within broader systems of support and care. They recommend that future work: First to study long-term outcomes and practical benefits of app-based screening in diverse settings. Secondly to explore ways to integrate digital and traditional approaches, aiming for holistic, individualized pathways to support. And finally, to listen to caregivers, autistic individuals, and communities to ensure tools are inclusive, respectful, and address real-world needs. The authors emphasize that continued collaboration among families, clinicians, researchers, and autistic advocates is essential for creating tools that reflect a range of experiences and promote equitable access to support.
Next steps include making sure digital screening works well for all families and communities.
Conclusions
The SenseToKnow app offers a promising, user-friendly option for early autism identification, with high accuracy and broad applicability. By using digital behavioral patterns observed in naturalistic settings, such approaches can help reduce barriers and increase timely access to resources and support. Ongoing research will clarify how these tools fit within family-centered, strengths-based care.
Digital screening may help more families access information and support in ways that fit their needs.
Where next?
This conference, brings together four leading experts to explore distinct yet interconnected topics: early detection, co-occurring mental health challenges, participatory approaches, and evidence-based supports for adolescents and young adults. With a focus on practical tools, emerging models, and inclusive innovation, this conference is essential for professionals working to improve outcomes for autistic children, teens, and their families.
A phenomenal line-up includes: Professor Sven Bölte, Professor Geraldine Dawson, Associate Professor Georgia Pavlopoulou, and Professor Susan White.
Use the interactive programme below to gain an overview of the topic, meet the speakers, test your knowledge, and a whole lot more!
NB this blog has been peer-reviewed
References
• Dawson, G., Krishnappa Babu, P. R., Di Martino, J. M., Aiello, R., Eichner, B., Espinosa, S., et al. (2025). Validation of a Mobile App for Remote Autism Screening in Toddlers. NEJM AI, 1–12. https://ai.nejm.org/doi/abs/10.1056/AIcs2400510
• Pierce, K., Gazestani, V. H., Bacon, E., et al. (2022). Eye-tracking reveals abnormal visual preference for geometric images as an early biomarker of an ASD subtype associated with increased symptom severity. Biological Psychiatry, 92(8), 665–674. https://doi.org/10.1016/j.biopsych.2022.06.014
• Whitehouse, A. J. O., Varcin, K. J., Alvares, G., et al. (2017). Developmental behavioral interventions for children with autism spectrum disorder: An overview. Current Opinion in Psychiatry, 30(2), 91–97. https://doi.org/10.1097/YCO.0000000000000315
• Lord, C., Charman, T., Havdahl, A., et al. (2022). The Lancet Commission on the future of care and clinical research in autism. The Lancet, 399(10321), 271–334. https://doi.org/10.1016/S0140-6736(21)01541-5
• Dawson, G., & Bernier, R. (2013). A quarter century of progress on the early detection and treatment of autism spectrum disorder. Development and Psychopathology, 25(4pt2), 1455–1472. https://doi.org/10.1017/S0954579413000710
• de Leeuw, A., Happé, F., & Hoekstra, R. A. (2020). Editorial: Digital health tools in the assessment and intervention of autism spectrum disorder. Journal of Child Psychology and Psychiatry, 61(3), 227–229. https://doi.org/10.1111/jcpp.13155
• Wall, D. P., Dally, R., Luyster, R., et al. (2012). Use of artificial intelligence to shorten the behavioral diagnosis of autism. PLoS ONE, 7(8), e43855. https://doi.org/10.1371/journal.pone.0043855
About the author
Professor Francisco Musich, PhD is a Clinical Psychologist, Professor of Childhood Psychiatric and Neurological Disorder at Universidad Favaloro, Argentina, Head of the Department of Child and Adolescent Psychology at the Institute for Cognitive Neurology – INECO – Argentina, and Head of the Department of Psychopathology and Differential Diagnosis – ETCI – Argentina.
