• workshop_dense_100#slide 18

The score profile in RAIDT is the scored representation of one run's governance status. RAIDT produces two practical outputs: a run-level evidence pack and a score profile. The pack records what happened; the score profile rates that recorded run across the five pillars (Responsibility, Auditability, Interpretability, Dependability, Traceability). Because RAIDT treats the run as the unit of governance, the score profile is attached to a specific configured use rather than to a model in the abstract. The papers describe this as an anchored 1-5 profile grounded in inspectable artefacts. Using the deck's shorthand, the anchors are 1=missing / 3=partial / 5=audit-ready. The profile may be summarised statistically, but RAIDT retains the full five-pillar pattern so that trade-offs remain visible.

It matters because RAIDT is trying to solve an evidence problem, not merely a reporting problem. Policies, model cards, or fluent explanations can suggest that a system is responsible, but they do not show whether a particular run can be reconstructed, challenged, or improved. The score profile turns the run-level evidence pack into a comparable governance judgement. This lets organisations identify recurrent weaknesses, distinguish configuration effects from vendor effects, and test how influence methods as governance interventions change outcomes. It also makes governance claims falsifiable: if required evidence is absent, the relevant pillar should score poorly. In that sense, the score profile is not decorative reporting. It is the mechanism by which RAIDT makes governance readiness inspectable, reviewable, and operational in practice.

A healthcare team uses GenAI to summarise a chest-pain consultation into Symptoms, Diagnosis, Treatment, and Red Flags. The run-level evidence pack records the constrained system role, the ban on unsafe invention, the requirement to state uncertainty, the model and prompt versions, and the clinician oversight step. When the run is scored, Responsibility may be high because safe constraints and escalation are explicit, while Auditability depends on whether prompts, outputs, and checks are fully logged and hashed.

This example shows why the score profile matters. A fluent clinical summary alone would say little about whether the use was governable. The score profile shows whether the run is merely useful-looking or genuinely ready for review and contestability. It also gives the organisation a basis for improving the workflow before broader deployment.

• 00-RAIDT_Wording_v2
• 11-RAIDT_Academic_Logic_M_v11