Age estimation

What the clause says

Estimation derives age from features or behaviour inherent in the person, often through biometric analysis. The framework points to presentation attack detection and to the measurement of error near the threshold.

What it supplies to a comparison

Estimation is where the accuracy distribution, the challenge age and the attack resistance have the most effect on the comparison. A single accuracy number is not enough to characterise it.

Evidence to keep

• The error distribution by age band and by demographic group.
• The presentation attack test report with the instruments used.

What the evidence looks like

For 4.3.2, the first artefact is The error distribution by age band and by demographic group.. The second is The presentation attack test report with the instruments used.. If 4.3.2 has no first artefact, the practice is unproven at 4.3.2, and the report should record the absence rather than infer the practice from a policy.

Questions for the provider

1. What is the error rate within two years of the threshold?
2. Which groups were represented in the test crew, and which were not?

How it connects to the draft

Age estimation belongs to 4 characterisation, and it gives the analysis one fixed point: Estimation is where the accuracy distribution, the challenge age and the attack resistance have the most effect on the comparison. A single accuracy n. A comparison that cannot place a result against 4.3.2 is measuring something the framework has not defined, and the report should say so.

In the register, 4.3.2 is read beside the unit that tests it, and where it sets an objective rather than a number, that unit turns the objective into a question and an evidence artefact.