• workshop_dense_100#slide 3

The mind map was built from a structured source base rather than from an informal brainstorm. In the academic-logic paper, the RAIDT programme is explicitly organised as an inputs-to-artefacts architecture: literature and policy debates, governance requirements, research questions, and a mid-range design theory lens feed the core artefacts. Those artefacts are then defined as the five pillars (Responsibility, Auditability, Interpretability, Dependability, Traceability), the run-level evidence pack, the run definition, scoring anchors, and influence-method configurations. This means the source basis for the map was an integrated conceptual scaffold linking problem framing, governance requirements, and artefact design.

The foundations and design-theory papers show how that scaffold was converted into RAIDT terminology. They argue that the run as the unit of governance is the missing governance object in high-stakes GenAI decision support, and that governance must be expressed through reconstructable evidence rather than narrative assurance alone. Accordingly, the map?s logic is grounded in a score profile built from the five pillars, in a run-level evidence pack that preserves what happened in one configured use event, and in influence methods as governance interventions whose effects must be evidenced at run level. The map therefore reflects not only topic coverage, but also the project?s academic positioning: a design science contribution, formalised as a mechanism-based mid-range design theory and prepared for a three-paper arc, sector playbooks, and journal submission packages.

In that sense, the mind map?s source basis is best understood as a synthesis device for the whole RAIDT programme. It draws together the conceptual background, the project positioning logic, and the operational grammar needed to move from responsible AI principles to inspectable governance practice. That is why the map naturally centres the run-level evidence pack, the score profile, anchors 1=missing / 3=partial / 5=audit-ready, and the claim that run-level governance evidence is the programme?s primary proof object.

A concrete public-service example would be a council team mapping a GenAI eligibility-advice system before field deployment. Using the same source basis described in the papers, the team would start with the governance requirement to justify one disputed advice event, then map the run as the unit of governance, the relevant policy clauses, and the five pillars (Responsibility, Auditability, Interpretability, Dependability, Traceability). They would then connect those nodes to the run-level evidence pack fields: prompt version, retrieval snapshot identifiers, model deployment ID, oversight record, and output hash.

In practice, that map would also show influence methods as governance interventions. If retrieval is added to improve policy grounding, the team would expect stronger Traceability and Auditability only if the evidence pack stores the exact rule text and identifiers. The resulting score profile, interpreted through anchors 1=missing / 3=partial / 5=audit-ready, gives the organisation a reviewable basis for deciding whether the configuration is fit for contested eligibility decisions.

• 11-RAIDT_Academic_Logic_M_v11
• 12-RAIDT_DSR_Theory_M_v8
• 08-RAIDT_Foundations_M_V50