Field classification
The evidence-standard landscape
Different evidence standards answer different questions. The useful distinction is not which evidence matters, but what each evidence type can and cannot establish.
| Category | Question answered | Useful for | Does not by itself prove |
|---|---|---|---|
| Runtime action evidence | What did the system or agent do? | Runtime control, enforcement, agent governance, incident review | Full business decision-state, business authority, organisational commitment |
| Governance evidence taxonomy | What evidence exists about an organisation’s controls or governance posture? | Risk pricing, capital assessment, underwriting, governance review | The basis of a specific consequential decision |
| Audit and telemetry records | What events were observed? | Monitoring, debugging, investigation | Decision relevance, authority, judgement or commitment |
| Model governance records | What model, approval or control framework existed? | AI governance, model risk management, documentation | What governed an individual decision at execution |
| Decision evidence standard | What evidence, authority, conditions, system contribution, human judgement and commitment formed the decision? | Replayability, review, dispute, handover, delegated authority, claims, underwriting and assurance | That the decision was automatically correct, lawful, fair or compliant |
Runtime and governance evidence may form part of a decision-evidence record, but they do not by themselves establish the full decision-state.
Why the distinction matters
In a consequential review, an organisation may need to connect system events, source evidence, authority, conditions, human judgement and commitment. Treating one evidence type as a substitute for the whole record creates avoidable reconstruction burden.