• workshop_dense_100#slide 20

In RAIDT, the statement 'what RAIDT is not' is a boundary-setting definition rather than a rhetorical aside. It means that RAIDT should not be misconstrued as a prompt-engineering exercise, a RAG implementation, an auditability-only framework, a model card, a generic checklist, or a guarantee of legal compliance. The papers place this boundary around the framework because RAIDT is specifically a run-level evidence method for governed organisational use of GenAI. Its core purpose is to turn a particular use event into a reconstructable and reviewable object through the run-level evidence pack and a score profile across the five pillars (Responsibility, Auditability, Interpretability, Dependability, Traceability).

This negative definition matters because many adjacent artefacts solve only part of the governance problem. Model cards provide model-level transparency, but not run-specific provenance. RAG can support Traceability, but only if retrieval evidence is preserved. Prompting can support Interpretability, but prompting alone does not establish Responsibility, Dependability, or Traceability. A checklist can record declared controls, yet RAIDT requires evidence of what actually happened in the run as the unit of governance.

The practical significance is methodological and organisational. By stating clearly what RAIDT is not, the framework avoids overclaiming and keeps scoring properly bounded. The anchors 1=missing / 3=partial / 5=audit-ready do not certify correctness or legal sufficiency; they indicate how far a run can be inspected, challenged, and improved. That boundary protects scholarly clarity and prevents misuse of RAIDT as a marketing label or compliance shortcut.

Consider a bank using GenAI to draft an adverse-action explanation after a credit refusal. A project team might believe that a structured prompt and a compliant-looking template are enough to show responsible use. RAIDT says otherwise. The organisation still needs a run-level evidence pack showing the prompt version, model configuration, decision criteria referenced, generated explanation, and the review steps that determined whether the draft could be used.

Why does that matter? Because a customer may challenge the explanation later. If the bank only kept the final text or relied on a model card, it could not show how the explanation was generated or whether reasons were linked to documented criteria. RAIDT's boundary statement prevents this confusion. It clarifies that the framework is not a polished explanatory template and not a legal guarantee; it is an evidence-centred method for reconstructing and governing the specific run.

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