Responsible Generative AI · Governance · Doctoral Research

Standards-Aligned Responsible Generative AI and Human Agency

A doctoral research route connecting responsible Generative AI principles with standards-aligned controls, human oversight, privacy, transparency, AI literacy, and socio-technical evaluation.

Doctoral research program · active · framework under development

Risk

What can go wrong?

Signals

What do we know?

Oversight

Who reviews decisions?

Govern

How is the system improved?

Standards-to-evidence research map

A conceptual route from governance expectations to controls, human oversight, and testable evaluation.

Scope

Role and problem

My role: Doctoral researcher at Charles Darwin University, supervised by Jon Mason. The public route describes the research architecture and evidence plan without presenting work-in-progress as completed findings.

Responsible Generative AI is often discussed as a list of principles. The harder systems problem is operational: how should institutions translate responsibility into technical controls, governance mechanisms, evaluation criteria, human decision boundaries, and evidence that can be tested in practice?

Architecture

System flow

Use case and stakeholders

Standards and governance mapping

Risk and data assessment

Technical controls

Human oversight boundaries

Transparency and AI literacy

Evaluation protocol

Review and iteration

Results and scope

Public record

Standards

Operational alignment

Map high-level responsible-AI expectations into concrete technical and organisational requirements that can be reviewed and tested.

Human

Agency and oversight

Define where human judgement, escalation, contestability, and non-delegable decisions should remain explicit in GenAI-supported workflows.

Evidence

Evaluation design

Assess privacy, transparency, traceability, AI literacy, risk controls, and system usefulness as connected socio-technical properties.

Signal	Scope	Interpretation
Standards	Operational alignment	Map high-level responsible-AI expectations into concrete technical and organisational requirements that can be reviewed and tested.
Human	Agency and oversight	Define where human judgement, escalation, contestability, and non-delegable decisions should remain explicit in GenAI-supported workflows.
Evidence	Evaluation design	Assess privacy, transparency, traceability, AI literacy, risk controls, and system usefulness as connected socio-technical properties.

Shared here: This page is a public research-scope record. It distinguishes the active research questions and evaluation architecture from empirical findings that are still to be produced.

Contribution

Define a standards-aligned research architecture that connects technical system design with governance rather than treating them as separate layers.
Investigate human agency, oversight, privacy and data governance, transparency, AI literacy, and risk controls as interacting design requirements.
Develop an evaluation approach that can distinguish aspirational principles from controls and evidence that work in context.

Lessons

Governance becomes useful when it changes architecture, evaluation, or decision boundaries.
Human oversight needs a designed role, escalation path, and evidence of effectiveness.
Responsible-AI claims should be tied to testable controls and context-specific evidence.

Limitations

The doctoral research is at an early stage; no empirical PhD findings are claimed yet.
The initial empirical domain is teacher education and professional learning, so transfer to other sectors must be demonstrated rather than assumed.
Standards alignment does not by itself prove safety, effectiveness, or compliance; evaluation remains necessary.

Stack

Responsible Generative AI
AI Governance
Human Agency
Human Oversight
Privacy
Transparency
AI Literacy
Socio-Technical Systems
Evaluation