Ethics, integrity, and responsible use

The honest starting point

Higher education institutions are at different stages on AI policy. Some have detailed, considered frameworks. Many have interim guidance that is being revised. Some have nothing yet. Wherever your institution sits, your professional obligations do not wait for the policy to arrive.

The questions that matter most are not "is AI allowed?" — they are "what am I responsible for when I use it?" and "what am I asking my students to be responsible for?" This page tries to give you clear thinking on both.

Academic integrity — the fundamental question

The fundamental question in higher education is not whether AI is a form of cheating. It is whether the assessed work demonstrates what it is supposed to demonstrate. That has always been the question. AI makes it more urgent.

Here is a useful way to frame it. If a student submits an essay, a report, or a project that was substantially generated by AI without appropriate disclosure, what has been lost? The answer is: evidence of the student's own reasoning, analysis, and capability. The assessment is no longer measuring what it was designed to measure. That is the problem — not AI itself.

1. ✓Generally acceptable: using AI to understand a concept, get feedback on a draft the student has written, brainstorm ideas before developing them independently, practise exam questions, or identify gaps in understanding. These are legitimate learning uses.
2. ✓Acceptable for staff: using AI to draft lecture materials, feedback templates, rubrics, module documentation, or exam questions — provided you review, verify, and take professional responsibility for the output.
3. ✗Not acceptable: a student submitting AI-generated work as their own in any assessed context — essays, reports, dissertations, portfolios, reflective writing — without appropriate disclosure as defined by your institution's policy.
4. ?Context-dependent: AI-assisted editing, restructuring, or paraphrasing of student writing. Some institutions permit this with disclosure; others do not. Your policy needs to be explicit, and students need to know before they submit — not after.

What "AI disclosure" means in practice

An increasing number of institutions require disclosure when AI tools have been used in assessed work. This is the right direction — transparency is better than either a blanket ban or silent use. But disclosure requirements are only useful if they are specific enough to be actionable.

"I used AI" tells you nothing useful. A good disclosure statement specifies which tool was used, for what purpose, at what stage of the work, and what the student then did with the output. For example: "I used Claude to generate an outline for this essay. I then wrote the essay from scratch using that outline as a structural reference."

If your institution does not yet have a disclosure template, it is worth developing one — both for assessed student work and for your own professional practice as a researcher and teacher.

Postgraduate supervision and AI

Supervising a postgraduate student who is using AI raises questions that undergraduate assessment does not. A PhD thesis is an original contribution to knowledge — and the question of what "original" means when AI tools are part of the research workflow is genuinely unresolved.

Some practical principles that hold regardless of how the policy develops.

1. 1Have the conversation early. Establish at the start of supervision what AI use is permitted, what must be disclosed, and what the student's institution and relevant discipline body requires. Do not leave this until a problem arises.
2. 2Original contribution remains the standard. A thesis must demonstrate the student's own intellectual contribution. AI can support the research process — literature scanning, writing assistance, data analysis support — but the thinking, the argument, and the conclusions must be the student's.
3. 3Research ethics may be implicated. If AI tools are used to process research participant data, analyse sensitive datasets, or generate content involving real people or organisations, research ethics approval may need to cover this explicitly.
4. 4The oral examination still matters. A viva or oral defence remains the most reliable way to verify that the thesis represents the student's own understanding. If the student cannot explain and defend what they submitted, that is the conversation.

Bias, accuracy, and your professional responsibility

AI tools reflect the data they were trained on — which includes cultural biases, gaps in representation, and assumptions embedded in the sources from which the models learned. In a higher education context, this matters in specific ways.

AI-generated reading lists may over-represent English-language, Western, and historically dominant perspectives. AI-generated feedback may apply assumptions about "good academic writing" that do not hold across disciplines or cultures. AI-generated content about contested topics may present one position as more settled than it is.

None of this is an argument against using AI. It is an argument for always remaining the professional in the room. You review the output. You apply your disciplinary expertise. You take responsibility for what goes out under your name or your module.

AI detection — an honest assessment

AI detection tools are not reliable enough to use as evidence of academic misconduct. This is not a minority view — it is the emerging consensus among researchers studying these tools.

Detection tools have meaningful false positive rates. They have flagged legitimate student writing as AI-generated — including work by students whose first language is not English, students who write in a formal register, and students who follow structured academic conventions closely. Using detection output as evidence in a misconduct process exposes your institution to serious fairness and legal risk.

The more reliable approach is assessment design that makes AI shortcuts less useful and less attractive — and conversation when something seems inconsistent with what you know of the student. The assessment design page covers this in detail.