L03 / Interactive laboratory

AI validity-range explorer

Explore why prediction uncertainty should grow outside the region represented by training and validation evidence.

Educational model

Values are synthetic and dimensionless unless labelled. This laboratory illustrates relationships; it is not a solver, design tool, validation result or client-project output.

Reference behaviourModel predictionIndicative uncertainty
01

What it demonstrates

Uncertainty should increase when a query moves away from the region supported by training and validation evidence.

02

What it does not demonstrate

The band is an educational construction, not calibrated probabilistic confidence and not evidence for a deployed model.

03

Practical question

Can the system recognize unsupported inputs, reduce its authority and route the decision to a safer fallback?

Apply the principle to a real requirement

Define the decision, evidence and operating boundary.

Talk to CX24