• qa_deck_100#slide 33 · Socio-technical governance and accountability lineage

Information Systems governance shapes RAIDT by supplying its core governance logic: decision rights, control, accountability, coordination, and the information artefacts through which these are enacted. Across the RAIDT papers, the problem is not that organisations lack AI principles, but that governance remains too abstract or too coarse when generative AI behaviour is configured at run time. RAIDT therefore adapts IS governance to configured GenAI use by defining the run as the unit of governance and by turning one situated use event into a bounded governance object. The run-level evidence pack is the practical expression of that move, because it preserves the prompt, configuration, retrieved context, outputs, checks, and oversight decisions that allow a specific use to be reconstructed and challenged later.

IS governance also shapes RAIDT methodologically. Instead of treating logs as technical residue, RAIDT treats them as organisational evidence that must travel across managers, compliance teams, auditors, and operational staff. The score profile across the five pillars (Responsibility, Auditability, Interpretability, Dependability, Traceability) makes these governance conditions comparable, while the anchors 1=missing / 3=partial / 5=audit-ready give reviewers a disciplined basis for judging governance readiness. In the same way, RAIDT treats influence methods as governance interventions rather than as engineering choices alone: structured prompting, retrieval augmentation, adaptation, and alignment matter because they alter both system behaviour and what can be reviewed. In that sense, RAIDT is shaped by IS governance as an evidence-centred, socio-technical extension of organisational control for live GenAI work rather than as a model-level checklist.

A public-service team uses GenAI to advise on eligibility rules. Under a conventional policy-only approach, the organisation may have an approved system and a written governance policy, yet still fail to show which rule version informed a disputed recommendation. RAIDT applies IS governance more concretely. The run-level evidence pack records the prompt, model version, retrieval snapshot of the exact policy clause, hashes, timestamps, and the reviewer?s oversight decision.

That evidence then supports a score profile that different actors can use for different governance purposes. Operations can check whether the advice stayed within the workflow boundary, compliance can inspect whether the correct rule text was retrieved, and an appeals function can test whether the case is reconstructable. If the retrieval snapshot is missing, Auditability and Traceability cannot credibly reach an audit-ready state, even if the answer sounds fluent.

• 12-RAIDT_DSR_Theory_M_v8
• 17-RAIDT-Sociotechnical_M_v6