Clinical Validation

Why clinical accuracy matters

When a platform produces scores that shape decisions about a child's education, accuracy isn't a feature. It's a prerequisite. A wrong standard score could put a child in the wrong program, cause them to miss support they need, or lead to interventions that don't fit their actual profile. We built the entire scoring pipeline around one rule: if we can't guarantee the accuracy of a calculation, we don't show it.

Clinical calculations done right

All clinical calculations in Cognistase are performed by a dedicated medical calculation engine. This isn't AI. It's traditional software that applies validated clinical formulas exactly the way a psychologist would by hand. Standard scores, percentiles, confidence intervals, and discrepancy indices are all calculated the same way every time. No randomness, no approximation, no guessing.

No AI does the math

Consumer AI tools sometimes get arithmetic wrong. When you're computing developmental scores, that's unacceptable. Our architecture prevents this by design: no language model performs any calculation in the clinical pipeline. The AI reads documents and extracts information. The math is done by validated, auditable code.

Evidence tiers and confidence indicators

Every data point in the system carries two markers: an evidence tier (how the data was collected) and a confidence indicator (how reliable the analysis is). A standard score from a WISC-V given by a neuropsychologist carries Tier 1 evidence and high confidence. A self-report from an 8-year-old carries Tier 4 evidence and lower confidence. Both are valuable. Both are clearly labeled so anyone reading the profile knows exactly what they're looking at.

Supported assessment frameworks

The clinical validation system works with recognized instruments used in clinical practice for gifted and twice-exceptional assessment:

• WISC-V (Wechsler Intelligence Scale for Children, 5th edition)
• BRIEF-2 (Behavior Rating Inventory of Executive Function, 2nd edition)
• Vineland-3 (Vineland Adaptive Behavior Scales, 3rd edition)
• SRS-2 (Social Responsiveness Scale, 2nd edition)
• Additional validated instruments as they become relevant

How we validate

Every scoring algorithm is checked against published clinical norms before we deploy it. We compare our outputs against known reference datasets to make sure they're accurate. Error rates are tracked and published. When we add a new assessment instrument, it goes through a full validation cycle before users can access it.

Independent scientific review

Our methods are reviewed by an independent advisory committee of researchers and clinicians who specialize in giftedness and developmental psychology. They have access to our scoring algorithms, validation results, and error rates. Their job is to challenge us, not confirm us.

Nothing is hidden

All scores, calculations, and the formulas behind them are fully visible. You can see exactly how a standard score was calculated, which data points went into it, and what clinical norms were used as reference. If a score seems off, you can trace the calculation step by step. This isn't a future promise. It's how the system works today.