• workshop_dense_100#slide 24

Within RAIDT, Responsible AI is best understood as the normative basis for governing organisational GenAI use, but not as the final governance method. The managerial and governance papers consistently present Responsible AI as a commitment to transparency, accountability, fairness, safety, and meaningful oversight, especially where AI supports consequential judgement under uncertainty. RAIDT accepts that starting point, yet argues that those commitments must be translated into inspectable evidence for real use rather than left at the level of principle, policy, or model description.

This is why RAIDT defines a governance method with two practical outputs: a run-level evidence pack and a score profile. The framework preserves the ethical orientation of Responsible AI while operationalising it through the five pillars (Responsibility, Auditability, Interpretability, Dependability, Traceability). These pillars do not discard earlier Responsible AI concerns; they reorganise them around reconstructable accountability for one use event. The anchors 1=missing / 3=partial / 5=audit-ready then make governance quality explicit. In that sense, Responsible AI matters in RAIDT because it tells the framework what should count as good governance, while RAIDT tells the organisation what must be captured to show that good governance occurred.

It matters especially for GenAI because deployed behaviour is shaped not only by the base model but also by prompts, retrieval, tools, adapters, and review actions. RAIDT therefore treats influence methods as governance interventions. The result is a more disciplined account of responsible use: values are retained, but they are tested against evidence from configured runs in actual organisational contexts.

Consider an HR team using GenAI to draft shortlist justifications. A purely principle-level Responsible AI stance might require fairness, transparency, and accountability, but it may still leave the team unable to explain months later which prompt template, criteria version, model deployment, or adaptation layer informed the justification. That is precisely where RAIDT becomes important.

Under RAIDT, the hiring interaction is governed as a run. The run-level evidence pack records the prompt template version, criteria version, model or adapter identifiers, output, and review steps. The score profile then shows whether the case was responsible, auditable, interpretable, dependable, and traceable. In practice, this matters because HR decisions are often contested after the event. RAIDT turns Responsible AI from a statement of intent into a reviewable governance record that can support challenge, oversight, and organisational learning.

• 01-Responsible_AI_for_Managerial_Decision-Making_Under_Uncertainty-V3
• 11-RAIDT_Academic_Logic_M_v11
• 15-RAIDT-IS-Governance_M_v07