S7.01 - Design_science_research

S7.01 ? Design science research

flowchart LR
    A[High-level AI governance principles] --> B[Need for operational oversight]
    A2[Descriptive or policy-only approaches] --> B
    B --> C[RAIDT
Run-level evidence framework]
    C --> D[[Design science research
Artefact logic for RAIDT]]
    D --> E[Run-level evidence]
    E --> F[Evidence pack]
    E --> G[RAIDT score profile]
    F --> H[Reviewer reconstruction]
    G --> I[Governance readiness]
    D --> J[Organisational learning]
    D --> K[Policy alignment]
    L[Healthcare]
    M[Public services]
    N[Finance]
    O[Education]
    P[Templates, logging, dashboards]
    L --> D
    M --> D
    N --> D
    O --> D
    P --> D

? Star S7 - Academic Theory and Design Logic

Star context: Positions RAIDT within Academic Theory and Design Logic by showing that RAIDT is not only a descriptive account of GenAI governance, but a designed and evaluable governance artefact intended to solve a real organisational problem through run-level evidence and structured assessment.

Academic picture

Definition / background

Design science research is a research tradition, especially prominent in Information Systems, that seeks to address important real-world problems by designing, justifying, and evaluating artefacts. Those artefacts may be models, methods, constructs, decision procedures, frameworks, or socio-technical arrangements rather than only software applications. The central idea is pragmatic but academically rigorous: knowledge is produced not only by describing or explaining the world, but also by designing an intervention that improves it and by learning from how that intervention performs.

In RAIDT, design science research matters because the governance problem is not solved by description alone. Organisations do not merely need a theory saying that GenAI should be responsible, auditable, interpretable, dependable, and traceable. They need a usable governance artefact that structures evidence at the level of a run, produces a reviewable evidence pack, and generates a score profile that supports oversight, comparison, and improvement. RAIDT therefore fits design science research because it is intentionally constructed to solve a practical governance problem while also contributing conceptual and methodological knowledge.

This distinguishes design science research from several nearby ideas. It is not the same as a purely explanatory theory, whose primary goal is to account for observed phenomena. It is not the same as routine system development, because its purpose is scholarly contribution through justified design and evaluation. It is also not simply a compliance checklist, because the artefact must embody defensible design logic about what evidence should exist, why it matters, and how that evidence supports governance decisions. In RAIDT, design science research provides the overarching logic that connects conceptual design to operational governance.

The concept belongs inside RAIDT because RAIDT is best understood as a designed governance artefact for the era of generative AI. It defines the run as the unit of governance, specifies what evidence should be collected, and translates that evidence into practical outputs that organisations can inspect and contest. That is precisely the kind of problem-solution orientation that design science research is intended to support.

Why this concept matters

Design science research matters because it gives RAIDT a defensible answer to the question, ?Why build this framework in this form?? Without that answer, RAIDT could be misread as an arbitrary model, a loose conceptual taxonomy, or a policy aspiration without an operational core. Design science research clarifies that RAIDT exists because organisations face a concrete governance problem: they must make GenAI use reviewable at the level where work is actually performed.

The concept also prevents a common confusion between principles and mechanisms. Many governance discussions stop at desirable values, but design science research pushes the project toward artefact logic, use context, evaluation, and improvement. That shift matters in GenAI governance because risks often emerge in specific runs, prompts, settings, user roles, data conditions, and task contexts. A design-science framing therefore justifies why RAIDT focuses on run-level evidence rather than only policy-level statements.

If this concept were missing, RAIDT would be easier to dismiss as either too theoretical for practitioners or too applied for theory-building audiences. Design science research avoids that false split by showing that a well-designed artefact can itself be a scholarly contribution while also delivering practical governance value.

Key idea: Design science research matters in RAIDT because it justifies building a governance artefact that turns abstract AI principles into reviewable run-level evidence and operational assessment.

What this item explains

Why RAIDT is framed as a designed governance artefact rather than only a descriptive theory.
Why run-level evidence, evidence packs, and score profiles are legitimate design outputs in a research contribution.
How RAIDT links practical problem-solving with academic rigour in Information Systems and organisational governance.
Why governance for generative AI requires a structured artefact and not only narrative guidance.
How evaluation and refinement become part of the contribution, rather than an afterthought.

Practical example / likely audience question

Audience question

Why describe RAIDT as design science research rather than simply calling it an AI governance framework or an Information Systems theory?

Answer

The concern behind this question is usually that design science research may sound like an unnecessary label, or that it implies RAIDT must be a software tool to qualify. The direct answer is that RAIDT is called design science research because its core contribution is the design of a governance artefact intended to solve a practical problem: how to make individual GenAI uses inspectable, assessable, and improvable in organisational work.

A generic governance framework might list principles, controls, or policy recommendations. A purely explanatory theory might clarify causes, mechanisms, or outcomes. RAIDT goes further by specifying a concrete governance object: the run-level evidence pack and associated score profile. For example, if a public-sector team uses a GenAI assistant to draft citizen communications, RAIDT asks what evidence should exist for that run, how it should be reviewed, and how the organisation should interpret strengths and weaknesses across the five pillars. That is an artefact-centred, problem-solving contribution, which is exactly why design science research is the right positioning.

RAIDT handles this better than a generic AI governance approach because it does not stop at saying organisations should be accountable. It gives them a designed structure for evidencing accountability in context. That move from aspiration to artefact is the essential design-science step.

Practical example in RAIDT terms

Consider a healthcare setting where a hospital team uses a generative AI system to draft discharge-summary explanations for patients. The GenAI use case is helpful because it can improve speed and clarity, but the run-level issue is that each generated output depends on prompt wording, patient context, model configuration, source materials, human review, and final sign-off.

In RAIDT terms, the organisation would need evidence showing who initiated the run, what task was intended, what context and source materials were used, which model and settings were active, what output was produced, what human checks were applied, and what revisions were made before the summary reached the patient. Responsibility is affected because staff roles and sign-off obligations must be clear. Auditability is affected because the organisation must be able to reconstruct the decision path. Interpretability is affected because clinicians must understand why the output is acceptable or problematic. Dependability is affected because the organisation needs confidence that the process performs consistently. Traceability is affected because inputs, outputs, versions, and review steps must be connected.

Design science research improves governance readiness here because it explains why RAIDT should be designed as a practical artefact for capturing, organising, and assessing this evidence. Without that design logic, the hospital may have policy statements about safe AI but still lack a repeatable way to examine what actually happened in a specific discharge-summary run.

Detailed link to RAIDT

Design science research links to RAIDT in four ways.

First, it explains RAIDT's core identity as a purposeful governance artefact designed to address the practical problem of governing generative AI in organisational work.
Second, it supports the choice of the run as the unit of analysis and intervention, because governance becomes meaningful when evidence is attached to specific uses rather than to abstract system claims.
Third, it justifies the creation of structured outputs such as the run-level evidence pack and five-pillar score profile as designed objects that support evaluation and comparison.
Fourth, it strengthens reviewability, contestability, audit readiness, and organisational learning by making the framework inspectable, improvable, and defensible as an intervention rather than only a concept.

Design science research ? Run-level evidence ? Evidence pack ? RAIDT score profile ? Governance readiness

That chain matters because it shows how an academic design logic becomes an operational governance pathway. Design science research is the reason RAIDT can claim both scholarly relevance and practical utility.

Link to the five RAIDT pillars

Responsibility

Design science research supports Responsibility by requiring RAIDT to be intentionally oriented toward a meaningful organisational problem with clear roles, purposes, and decision consequences.