Early development · Step 9 of 12
Numeracy
In this section we use the Berlin Numeracy Test — a validated short test of statistical and probabilistic reasoning that predicts performance in evidence-weighted decision-making. It runs pending an assessment-psychologist scope-of-use sign-off (in review at v1.4.1) and contributes evidence to J2 (Calibrated trust and probabilistic reasoning). The real assessment presents the full battery with adaptive follow-up items when needed.
An example item might be:
An AI-assisted résumé screener produces a 'strong candidate' flag. Historically, 10% of applicants are genuinely strong candidates. When shown a strong candidate, the tool flags them 90% of the time. When shown a weak candidate, it also incorrectly flags them 20% of the time. A random résumé receives a 'strong candidate' flag — what is the probability that this candidate is actually strong?
- 33%
- 90%
- 70%
- 10%
Illustrative item — the demo shows one representative question per section, not the full battery. In this persona, this answer anchors at L1 for J2.
The real assessment runs the full Berlin Numeracy Test, pending an assessment-psychologist scope-of-use sign-off (in review at v1.4.1).