• qa_deck_100#slide 4 · RAIDT core and mindmap orientation

Inside RAIDT, the score profile is the evaluative layer applied to the run-level evidence pack. RAIDT treats the run as the unit of governance, so governance readiness is not inferred from a model card or a policy statement; it is assessed for one configured use in one context, at one time. The run-level evidence pack captures the run identifier, prompts, model and tool versions, retrieval snapshots where used, outputs, hashes, and review steps. The score profile then rates that evidential record across the five pillars (Responsibility, Auditability, Interpretability, Dependability, Traceability). In the deck's shorthand, this is expressed through anchors 1=missing / 3=partial / 5=audit-ready, which the papers frame as missing evidence, partial evidence, and complete evidence sufficient for review, reconstruction, and justified use.

Governance readiness therefore works as an observable property of recorded evidence, not as a narrative claim about good intentions. A run with strong documentation but weak uncertainty handling may score well on Auditability and Traceability yet remain weaker on Responsibility or Interpretability. That is why RAIDT keeps the full score profile rather than collapsing all trade-offs into one number, even if a composite mean is sometimes reported. The profile supports comparison across runs, scenarios, and domains; it also supports remediation, because low scores point to the control that is missing. This is also where influence methods as governance interventions matter: prompting, retrieval augmentation, PEFT or LoRA, and alignment controls can raise or lower pillar scores only when their effects and artefacts are captured in the evidence pack.

In a public-service eligibility workflow, a staff member asks GenAI to interpret a benefits rule. The run-level evidence pack records the prompt, model version, retrieval query, the exact policy passages retrieved, their document identifiers and hashes, and the human review step before advice is issued. When scored, the run may receive high Auditability and Traceability because an auditor can reconstruct which rule text informed the advice. If the output is clear but does not explain uncertainty about borderline cases, Interpretability may remain only partial.

That is the practical value of governance readiness in RAIDT. The organisation can see that this run is reviewable and contestable, but not equally strong on every pillar. Instead of claiming that the system is simply governed, the score profile shows which governance properties are present in this particular run and which must be improved before the workflow is used more widely.

• 00-RAIDT_Wording_v2
• 08-RAIDT_Foundations_M_V50
• 11-RAIDT_Academic_Logic_M_v11