• workshop_dense_100#slide 92

The eight branches connect back to the centre because, across the programme architecture described in the papers, each branch is a different warrant for the same core claim: responsible GenAI governance must be grounded in the run where risk materialises. Read in that light, the branches are not separate agendas but mutually reinforcing routes back to one centre. The literature-gap branch shows why principle-heavy and system-level approaches leave a residual run-level evidence gap. The governance-theory and audit-lineage branches explain why reconstructable artefacts matter for accountability. The technical branch shows that prompts, retrieval augmentation, adapters, and alignment controls alter behaviour at run time. The design-science and measurement branches explain how the run-level evidence pack and score profile make those effects observable through the five pillars (Responsibility, Auditability, Interpretability, Dependability, Traceability).

The remaining branches return to the same centre from applied directions. The empirical-validation branch asks whether repeated runs under different configurations actually produce different governance outcomes. The policy-interoperability branch shows how run evidence becomes the measurement hinge linking obligations to verifiable artefacts. The sector-calibration or implementation branch shows how the same framework is adapted across healthcare, finance, public service, cybersecurity, HR, and other domains without losing its central logic. In short, every branch asks a version of the same question: what must be captured, scored, and reviewed so that a specific GenAI use can be reconstructed and governed? That is why the centre remains stable even as the surrounding branches vary in method, audience, or application.

A healthcare deployment in a regulatory sandbox makes the centre-branch relationship tangible. One branch concerns policy alignment, so the trust maps evidence fields to oversight and record-keeping duties. Another concerns measurement, so the team applies anchors 1=missing / 3=partial / 5=audit-ready when scoring the run. Another concerns technical configuration, so retrieval and alignment settings are logged. Another concerns empirical validation, so the same scenario is repeated to assess variance.

Although these activities look different, they all connect back to the same centre: treating the run as the unit of governance. Each branch contributes a different justification or operational pathway, but all depend on the same run-level evidence pack as the object that can be inspected, scored, audited, and improved.

• 08-RAIDT_Foundations_M_V50
• 11-RAIDT_Academic_Logic_M_v11