AI-enabled interventions for sustained attention
Mean increase in expert-recorded attention span.
The question & approach
Children received personalized neurodevelopmental activities generated through the SafeMango app and administered by Child Care Experts. Baseline and follow-up measures included SNAP-IV ratings and expert-recorded attention span.
Mean increase in expert-recorded attention span.
What improved & what was observed
The manuscript reports a 124.3% average increase in expert-recorded attention span, with a paired t-score of −16 (p < 0.001). The correlation between app scores and SNAP-IV scores was −0.948. Correlation is an association between measures, rather than proof of causality.
What this proof of concept measured
| Measure | Reported result | Context |
|---|---|---|
| Expert-recorded attention span | 124.3% average increase | Pre–post change over three months |
| Expert-timed attention | t = −16; p < 0.001 | Reported paired comparison |
| SNAP-IV ratings | t = 14.1; p < 0.001 | Reported improvement in standardized ratings |
| App / SNAP-IV association | r = −0.948 | Alignment between measures with opposite score directions |
What the findings support
Children’s expert-recorded attention spans more than doubled on average. The parallel changes in expert observation and SNAP-IV ratings provide two perspectives on the improvement. The findings support the feasibility and promise of personalized developmental activities within an expert-supported intervention.
The conclusion, in context
This pilot demonstrates a measured change within the studied group. Without a control group it cannot establish how much change was caused by the intervention, isolate the AI contribution, or prove transfer to classroom achievement.
Study scope & next questions
This pre–post study did not include a randomized control group. Expert-delivered activities were part of the intervention. The findings cannot isolate the contribution of AI, establish long-term persistence, or guarantee outcomes for a current app.
Relevant products
Research alignment indicates the product area informed by this work; it does not mean every current product version was tested.
Study details and methods are available in the full whitepaper on request. Findings should be read within the study design, sample size, and limitations above.
400Mn data points. 50,000+ longitudinal users. A decade of learning.
These company-reported figures describe the foundation of ARRC. This study tests a specific intervention or model question in the sample shown above; its findings should be read at that scale.
Understand the science & data ↗Every child has potential.
Let’s build what comes next.
For schools, families, researchers, and learning partners.