S12.01 - Inputs

S12.01 ? Inputs

flowchart LR
    A[Original PhD brief]
    B[Responsible AI decision-support framing]
    C[GenAI taxonomy and design theory]
    D[Background story and journal positioning]
    A --> E[RAIDT
Run-level evidence framework]
    B --> E
    C --> E
    D --> E
    E --> F[[Inputs
Programme-anchoring concept]]
    F --> G[Run-level evidence logic]
    G --> H[Evidence pack]
    G --> I[Five-pillar score profile]
    H --> J[Reviewer reconstruction]
    I --> K[Governance readiness]
    H --> L[Organisational learning]
    I --> M[Policy and supervisory alignment]
    N[Healthcare and public services] --> F
    O[Education and enterprise productivity] --> F
    P[Playbooks, scoring rubric, paper architecture] --> F

? Star S12 - Programme Architecture and Supervisory Navigation

Star context: Clarifies the starting materials from which RAIDT is built, so supervisors can see how the programme?s architecture grows from the original research problem, conceptual framing, and governance commitments rather than from any single paper, tool, or implementation example.

Academic picture

Definition / background

Inputs are the foundational materials that establish the intellectual starting point of the RAIDT programme. In this item, the term refers to the original PhD brief, the responsible AI decision-support framing, the generative AI taxonomy, the background problem narrative, the design-theory orientation, and the journal-positioning work that together define what kind of contribution RAIDT is intended to make.

This matters because RAIDT is not just a collection of governance artefacts or papers. It is a coherent research programme about how generative AI governance can move from broad principles and static policy claims towards run-level evidence, reviewability, contestability, and audit readiness. The inputs explain why the programme is organised in this way and why the run is treated as the unit of governance.

Inputs are therefore different from components, outputs, or empirical findings. They do not yet constitute a run-level evidence pack, a score profile, or a validated implementation pattern. Instead, they define the conceptual boundaries within which those later outputs make sense. In supervision terms, this item helps distinguish origin, rationale, and framing from execution, testing, and downstream application.

Within RAIDT, the inputs belong inside the knowledge architecture because they anchor continuity across the whole programme. They connect the early research problem to later evidence structures, ensuring that the evidence pack and five-pillar score profile remain aligned with the programme?s original governance purpose rather than drifting into generic AI assurance language.

Why this concept matters

Inputs matter because complex doctoral programmes often become difficult to explain once they produce multiple papers, artefacts, examples, and sectoral applications. Without a clear account of the inputs, supervisors and reviewers may confuse the programme?s central idea with one implementation choice, one empirical paper, or one domain example.

For organisations using generative AI, this same confusion has practical consequences. Governance frameworks can become disconnected from the problem they were supposed to solve, producing evidence that is procedurally neat but conceptually misaligned. By making the inputs explicit, RAIDT preserves the chain from initial governance concern to operational evidence requirement.

The concept also prevents a weak version of responsible AI work in which principles are stated but not translated into a disciplined framework. Inputs show what commitments the programme starts from, while RAIDT shows how those commitments are operationalised at run level.

Key idea: Inputs matter because they preserve the conceptual centre of gravity that allows RAIDT to turn responsible GenAI governance from an abstract ambition into an operational evidence framework.

What this item captures

The original research problem that motivated RAIDT as a governance contribution rather than as a narrow technical artefact.
The responsible AI decision-support framing that positions generative AI use inside organisational accountability structures.
The taxonomy work that clarifies what kind of GenAI phenomena RAIDT is intended to govern.
The background story explaining why principle-level governance is insufficient on its own.
The design-theory orientation that supports RAIDT as a structured framework rather than an ad hoc checklist.
The journal-positioning logic that explains how the programme speaks to academic audiences across governance, systems, and applied AI discussions.
The conceptual boundaries that later shape evidence-pack design, scoring logic, and supervisory explanation.

Practical example / likely audience question

Audience question

Why revisit the beginning of the project when the real contribution is the RAIDT framework and its evidence outputs?

Answer

The concern behind this question is usually that origin material looks historical rather than analytically necessary. The direct answer is that revisiting the inputs is essential because it shows what RAIDT is actually for. Without the inputs, a reader may assume the programme is mainly about a checklist, a scoring device, or a set of implementation examples. In fact, the centre of gravity is responsible governance of generative AI in organisational work, with run-level evidence used as the mechanism for making governance claims reviewable.

A practical example is a reviewer who sees the five-pillar profile and assumes RAIDT is simply another assessment matrix. The inputs correct that reading by showing that the profile exists because earlier framing work identified a gap between governance principles and situated evidence. RAIDT handles this better than a generic AI governance approach because it does not stop at high-level values or policy statements; it ties the original problem formulation directly to evidence requirements at the level of a concrete run.

Practical example in RAIDT terms

In healthcare administration, imagine a GenAI system used to draft patient discharge summaries for clinician review. The run-level issue is not merely whether the model appears useful in general, but whether this specific configured use, at this specific time, for this specific task, can be governed responsibly.

The relevant inputs include the original responsible decision-support framing, the taxonomy distinguishing generative drafting support from autonomous decision-making, and the design logic that treats evidence as a requirement for governance claims. From those inputs, RAIDT can specify what evidence is needed for the run: prompt and configuration details, human review role, clinical context boundaries, logging, exceptions, and criteria for acceptable use.

The most affected pillars are Responsibility, Auditability, and Traceability, with Dependability and Interpretability also influenced through the way the run is structured and reviewed. The value of the inputs is that they make the evidence request intelligible. Instead of asking for generic assurance, RAIDT asks for evidence that matches the programme?s original governance problem.

Detailed link to RAIDT

The inputs layer links to RAIDT in four ways.

First, it connects directly to RAIDT?s core idea by defining the initial governance problem that RAIDT is meant to solve: responsible oversight of generative AI in organisational work through evidence rather than assertion.

Second, it links to the run because the early framing establishes why the run, rather than the model or policy statement alone, becomes the meaningful unit of governance analysis.

Third, it links to the evidence pack and score profile because the categories of evidence, scoring criteria, and interpretation logic must remain consistent with the original conceptual commitments.

Fourth, it links to reviewability, contestability, audit readiness, and organisational learning because a framework can only support these outcomes if its operational mechanisms remain traceable to a clear problem definition and governance rationale.

Inputs ? Run-level evidence logic ? Evidence pack design ? RAIDT score profile ? Governance readiness

Link to the five RAIDT pillars

Responsibility

Inputs shape how responsibility is framed before any scoring or evidence collection begins. They clarify who the framework is designed to support, what kind of decisions are in scope, and what accountable use means in context.