Testing depth: metrics and thresholds, system audits, alignment/repeatability/adversarial testing, and resource allocation
Testing an AI system is layered: metrics against thresholds, audits of the surrounding system, checks that results repeat, and adversaries trying to break it. Each layer catches what the previous one cannot.
Why this matters for the AIGP exam
Testing scenarios ask which kind of testing a described activity is, and which kind a described failure needed. The layers are the vocabulary.
The essentials
- Metrics and thresholds. Quantitative performance — accuracy, precision and recall, error rates across subgroups — measured against acceptance thresholds set before testing, not after seeing the results.
- System audits. Structured examination of the whole system and its process: data handling, documentation, controls — not just model outputs.
- Repeatability. The same inputs and process should produce consistent results; a system whose behaviour cannot be reproduced cannot be meaningfully evaluated, debugged or audited.
- Adversarial testing and red-teaming. Deliberate attempts to make the system fail: crafted inputs, prompt attacks, misuse scenarios. It finds the failure modes cooperative testing never triggers.
- Resource allocation is part of test design. Testing effort is finite; directing it proportionally to risk — deeper testing for higher-stakes systems — is itself a governance decision the exam treats as testable.
What the exam asks
Classify a described testing activity, match a failure to the layer that should have caught it, and defend risk-proportionate test effort.
Going deeper
This page is the condensed version. The full topic — with the detail above expanded and practice questions attached — is in the app, inside Domain III. Domain I, a quarter of the course, is free to try first.
Back to the AIGP study guide.
AI Governance Study is an independent study aid. It does not represent a government entity: it is not affiliated with, endorsed by or authorised by any government, government agency or regulatory authority, and it does not provide government services or legal advice. Laws and frameworks are described in our own words — the official texts are listed at official sources. It is also not affiliated with, endorsed by, or sponsored by the IAPP. The AIGP name is used only to identify the exam this material helps you prepare for.