• workshop_dense_100#slide 33

In the RAIDT papers, design science research is defined in classic Information Systems terms as the creation and evaluation of artefacts that address important organisational problems while generating reusable knowledge. That definition matters because the generative AI governance problem identified by the papers is not merely descriptive. Organisations need a way to design governance for specific configured uses whose outputs shape records, recommendations, and decisions. Design science is therefore the appropriate methodological home because it supports both artefact construction and theory building.

RAIDT is the concrete example. It is not presented as a general ethics manifesto or a pure logging scheme. It is a conceptual but operationally oriented governance artefact that theorises run as the unit of governance. The artefact couples a run-level evidence pack with a score profile across the five pillars (Responsibility, Auditability, Interpretability, Dependability, Traceability), using anchors 1=missing / 3=partial / 5=audit-ready to make governance readiness inspectable. The associated design theory then explains why particular evidence fields, review steps, and configurations should improve reviewability under specified boundary conditions.

This matters in RAIDT for three reasons. First, it turns responsible-AI aspirations into inspectable governance objects. Second, it makes cumulative research possible, because different settings can evaluate the same artefactual logic rather than inventing new governance categories each time. Third, it allows organisations to examine influence methods as governance interventions, asking not only whether outputs improve, but whether responsibility, auditability, interpretability, dependability, and traceability improve in ways that can later be reconstructed and challenged.

In a finance setting, a lender uses generative AI to draft an adverse-action explanation for a declined credit application. A design science approach does not stop at saying explanations should be fair and clear. It designs the governance artefact needed for this use: the run-level evidence pack records the prompt template, policy criteria, model version, any retrieved policy text, the generated explanation, reviewer action, and timestamps; the score profile shows whether the run is ready for reliance.

That matters because a disputed explanation can then be examined as a governed event. If reason statements are linked to documented criteria, Interpretability improves. If the exact policy clause and configuration are preserved, Auditability and Traceability improve. If evidence is incomplete, the run can be escalated before release. In RAIDT terms, design science converts a difficult governance aspiration into a practical, reviewable organisational capability.

• 12-RAIDT_DSR_Theory_M_v8
• 11-RAIDT_Academic_Logic_M_v11