What happens when AI changes the evidence of learning?
Artificial intelligence is moving faster than most university committee cycles. I say that with respect, because governance needs care, but it also needs pace when the learning environment changes so quickly.
The question is no longer whether AI belongs in higher education. It is already there. Students are using it. Staff are exploring it. Employers are adopting it. Professional practice is changing around it. The more serious question is this: can universities still make defensible claims about student learning, academic standards and graduate capability in an AI-enabled world?
That question is uncomfortable because it is not only technical. It reaches into the credibility of awards, the validity of assessment, the meaning of academic judgement, the fairness of the student experience, the protection of academic integrity, and the confidence that employers, professional bodies, students and the public place in universities.
The mistake would be to treat AI only as a student conduct problem. That view is attractive because it gives universities familiar tools: misconduct panels, declarations, detection reports, handbook updates and perhaps a very determined spreadsheet. However, it misses the deeper issue.
AI does not merely make misconduct easier. It forces universities to ask whether their assessments still generate credible evidence of learning.
My argument is direct: AI in higher education is not mainly a plagiarism crisis. It is an evidence crisis.
It challenges the chain of inference that links a submitted artefact to a student’s actual capability. If that chain is weak, the assessment is weak. If the assessment is weak, the award becomes vulnerable. If the award becomes vulnerable, institutional trust becomes fragile.
Why is the real question not “How do we stop AI?”
Universities should not approach AI with panic. Panic is rarely a good educational strategy. Nor should they approach it with technological worship. Universities have survived printing presses, calculators, search engines, virtual learning environments and, with admirable courage, PowerPoint presentations with far too many bullet points. They can survive AI as well, but only if they govern it properly.
The next stage of AI-enabled education requires three things at the same time: intellectual honesty, assessment redesign and standards-led governance.
Without intellectual honesty, universities pretend that old assessments still mean what they used to mean.
Without redesign, they keep asking students to complete tasks that no longer test what they think they test.
Without governance, they produce local improvisation, inconsistent fairness and reputational risk.
So the central question is not: How do we stop AI?
That is the wrong question, and in many cases it is not realistic.
The better question is this: How do we design a university system in which AI may be used intelligently, openly and ethically, while every award remains defensible?
That word, defensible, matters.
Can we defend the task?
Can we defend the evidence?
Can we defend the academic judgement?
Can we defend the student experience?
Can we defend the award if challenged by a regulator, an employer, a professional body, an external examiner or the student themselves?
This is where AI becomes strategic. It is not just a classroom issue. It is a quality assurance issue, a governance issue, a digital transformation issue and, ultimately, a public trust issue.
What does the current evidence show about student AI use?
The first reason this matters is that student behaviour has already changed.
The HEPI and Kortext Student Generative AI Survey 2026 reported that AI use among full-time UK undergraduates is now near universal. It found that 95 per cent of students use AI in at least one way, and 94 per cent use generative AI to help with assessed work [1].
That does not mean 94 per cent of students are cheating. We must be careful and fair. It means that the learning environment has changed. It also means that any policy built on the assumption of rare or exceptional AI use is already out of date.
The same evidence points to a governance gap. Students may be using AI widely, yet institutional encouragement, approved tool access and clear guidance do not always match the level of use [1]. Jisc’s 2025 work also shows that staff and institutions are experimenting with AI, but that staff often need clearer guidance, proper development, discipline-specific examples and time for assessment redesign [2].
In plain English, the policy train has left the station, but some colleagues are still looking for the platform.
This is not a criticism of staff. It is a criticism of under-supported change. Universities cannot ask colleagues to redesign assessment for a new technological era while offering only a policy PDF, a webinar recording and good wishes.
Why must assessment prove learning, not merely collect polished work?
Universities do not only teach content. They certify achievement.
A university degree is a public claim. It says that a person has demonstrated knowledge, skills, judgement and capability at a particular level. That claim must remain meaningful.
The National Student Survey remains relevant here because it asks students about clarity of marking criteria, fairness of assessment, opportunities to demonstrate learning, timely feedback and the usefulness of feedback [3]. AI does not remove these questions. It makes them sharper.
If students do not understand when AI may be used, if staff apply rules differently across modules, or if assessment briefs do not explain the purpose of the task, then fairness and learning both suffer.
The issue is not only whether a student used a tool. The issue is whether the assessment system helps students understand what genuine learning looks like.
This is why I see the current moment as strategic rather than merely disruptive. Many assessments were already over-reliant on final products, under-attentive to process, thin on authentic application and slow to connect with professional practice. AI has simply walked into the room, smiled politely, and exposed the furniture.
The serious response is not to remove curiosity from learning. It is to design assessment that asks students to think, justify, test, revise, defend and learn.
A polished artefact may be impressive, but if it cannot be connected to student reasoning, it is not enough.
Why is assessment an inference system?
To govern AI properly, we need to define the real object of governance.
It is not the essay.
It is not the software.
It is not even the declaration at the end of a submission form.
The real object of governance is the inference we make from evidence.
Every assessment makes an inference. A student submits a report, sits an examination, completes a design project, gives a presentation, performs in a laboratory, defends a dissertation or submits a portfolio. From that evidence, academics infer what the student knows, understands and can do.
That inference must be valid, reliable, fair and explainable.
Generative AI disturbs this inferential chain because it can help produce the artefact without necessarily producing the learning. This does not mean that every AI-assisted artefact is invalid. It means universities must know what role AI played, what the student contributed, what learning outcome is being assessed, and what evidence supports the judgement.
This is the mystery inside the obvious fact.
The obvious fact is that AI can generate work.
The deeper issue is that universities were often assessing work as if provenance and process were obvious.
They are not obvious anymore. In truth, they were never fully obvious. AI has simply made the uncertainty visible.
So I propose a simple test for AI-era assessment governance:
Can we defend the inference?
Can we explain why this task provides evidence of this learning outcome, at this level, under these conditions, for this award?
If the answer is unclear, the problem is not merely AI. The problem is assessment design.
What does UK regulation require?
The UK governance picture is not one single regulatory machine. England, Scotland, Wales and Northern Ireland have different arrangements, and serious leadership must respect that. However, across these systems, the direction is clear: credible standards, quality assurance, student experience, evidence and accountability remain central.
In England, the Office for Students Condition B4 is especially important. It requires students to be assessed effectively. It requires assessments to be valid and reliable. It also requires academic regulations to be designed so that relevant awards are credible [4].
These are not decorative phrases. They are practical tests.
Valid assessment means that students demonstrate the knowledge and skills intended by the design of the assessment.
Reliable assessment means that students demonstrate those knowledge and skills consistently across students and over time.
Credible awards must reflect what students have actually achieved.
This gives university leaders direct questions.
If an assessment brief allows extensive undeclared AI generation, and the task no longer distinguishes between student capability and tool capability, is the assessment valid?
If different lecturers interpret AI rules differently, is the system reliable?
If award outcomes remain strong while evidence of student capability becomes weaker, are awards still credible?
These questions are not abstract. They sit at the centre of academic governance. They should be discussed by programme teams, academic boards, quality committees, external examiners, digital education teams and governing bodies.
Why must quality assurance be embedded, not added later?
The QAA UK Quality Code 2024 and current guidance are important because they emphasise strategic, embedded and evidence-based approaches to quality and standards [5]. The responsibility is not only with individual module teams. It is institutional.
This matters because AI cannot be solved by leaving every lecturer alone with a policy link and a prayer. Academic judgement must remain human, but the system supporting that judgement must be clear, consistent and proportionate.
QAA’s Academic Integrity Charter also remains relevant. It positions academic integrity as a baseline for qualifications that are genuine, verifiable and respected [6]. In the AI era, integrity is not simply about catching wrongdoing after the event. It is about designing conditions where honest learning is clear, supported and fair.
Across the wider UK, the same theme appears in different ways. Scotland’s Tertiary Quality Enhancement Framework places emphasis on learning, teaching, assessment, student engagement, externality, data and evidence [7]. Medr’s Strategic Plan 2025 to 2030 in Wales emphasises learner outcomes, continuous improvement, flexible pathways and evidence-based decision-making [8]. The Scottish Credit and Qualifications Framework continues to connect credit, level, learning outcomes, assessment and quality assurance [9].
Flexible learning does not mean flexible standards in the careless sense. It means flexible pathways with clear evidence of achievement.
What are the main governance risks universities must face?
I see seven governance risks that universities must face honestly.
1. Validity risk
If a task can be completed largely through AI without the student showing the intended knowledge or skill, then the assessment may no longer support the claim attached to it.
This is not solved by asking the student to tick a box. It is solved by redesigning the task and the evidence base.
2. Reliability risk
Different staff, modules and schools may interpret AI use differently.
One tutor allows AI for brainstorming. Another allows drafting support. A third bans everything. A fourth says nothing because the handbook was last updated when Twitter was still Twitter.
This creates confusion, inconsistency and appeal risk.
3. Equity risk
Some students have paid tools, faster devices, better prompting confidence, stronger informal networks or family members who understand digital systems. Others have limited access, fear of misconduct allegations, or no clear institutional support.
If AI becomes part of learning, unequal access becomes an academic standards issue as well as a student experience issue.
4. Academic integrity risk
The old model assumed that misconduct could often be detected by textual similarity, suspicious phrasing or source irregularity. That world has changed.
Detection tools may sometimes inform human judgement, but they cannot carry the whole burden of proof for a modern assessment system.
5. Staff capability risk
Policies are easy to write and hard to implement.
Staff need time, development, examples, approved tools, legal and ethical guidance, and permission to redesign assessment properly.
6. Data and ethics risk
AI tools raise questions about personal data, copyright, intellectual property, bias, accessibility, environmental cost, procurement and vendor dependency.
A university that adopts AI without strong data governance is not modern. It is just fast in the wrong direction.
7. Assurance risk
Governing bodies and senior teams need more than anecdote. They need evidence that assessment redesign is happening, students understand the rules, staff are trained, external examiners are asking the right questions, and awards remain defensible.
Activity is not assurance.
A working group is not assurance.
A colourful strategy diagram is not assurance, although it may lift the mood in a long committee meeting.
Why can AI detection not carry the weight of academic standards?
A particular risk deserves separate attention: the temptation to treat AI detection as the answer.
It is not the answer.
Jisc’s 2025 update makes clear that there have been major advances in AI capability, but no comparable major development in detection technology itself [10]. This is not a small operational inconvenience. It changes the logic of governance.
A detector may sometimes provide a signal. It may support a wider conversation. It may help academic staff notice a concern. But it cannot become the foundation of fairness, validity or academic judgement.
If universities rely too heavily on detection, they create several problems. They risk false accusations. They risk unequal treatment. They risk an arms race between tools. They risk teaching students that the real aim is not to learn, but to avoid being caught by software.
That would be a poor educational message and a weak ethical position.
The better approach is assessment design.
Students should be asked to show process, explain choices, defend reasoning, demonstrate competence and acknowledge tool use where it is permitted.
Universities should build tasks in which student thinking is visible, not tasks in which suspicion becomes the main method of assurance.
In simple terms, detection may have a supporting role, but it cannot be the spine of academic standards.
The spine must be assessment validity, human judgement, transparent rules and evidence-rich design.
What assessment philosophy should universities adopt?
I recommend a new assessment philosophy built around three words: trace, test and trust.
Trace
Trace means that assessment should, where appropriate, make the learning process visible.
This does not mean drowning students and staff in paperwork. It means designing proportionate evidence: plans, drafts, version history, laboratory notes, design logs, annotated prompts, reflective commentaries, project milestones and decision records.
The point is not surveillance. The point is educational provenance.
Test
Test means that assessment should include moments where students demonstrate capability under conditions that make sense for the outcome.
Sometimes that will be a supervised task. Sometimes it will be a viva. Sometimes it will be live problem-solving, practical demonstration, code walkthrough, design critique, clinical reasoning, studio review, laboratory performance or professional presentation.
The test should match the construct, not the convenience of the timetable.
Trust
Trust means that the institution builds a system students and staff can understand.
Rules should be clear, proportionate and fair. Students should know when AI is prohibited, when it is permitted, when it must be acknowledged, and when critical use of AI is part of the assessment itself.
Staff should know how to mark such work, how to respond to concerns and how to support students without turning every class into a courtroom.
This philosophy changes the assessment question.
We no longer ask only: what did the student submit?
We ask: what evidence shows that the student learned, reasoned, judged, created, tested, defended and improved?
That is a richer educational question.
How can universities classify AI use in assessment?
A practical model is to classify assessment tasks into four clear AI-use categories.
Category A: AI-restricted
Students must demonstrate foundational knowledge or professional competence without AI assistance, except for approved accessibility support.
This may be appropriate for core calculations, basic diagnostic reasoning, essential disciplinary knowledge, safety-critical skills or threshold professional competence.
Category B: AI-supported
Students may use AI for specified support, such as brainstorming, planning, feedback on structure, language refinement or revision prompts, with clear disclosure.
The student remains responsible for accuracy, argument, sources and final judgement.
Category C: AI-integrated
The assessment deliberately requires responsible AI use.
Students may be asked to compare outputs, identify hallucinations, verify sources, critique bias, improve prompts, challenge assumptions and explain why the final human judgement is stronger than the initial machine output.
Category D: AI-audited
Students submit a short process record explaining what tools were used, why they were used, what was accepted, what was rejected and how the final submission remained their own intellectual work.
This is not a bureaucratic game. It is a learning contract.
It tells students what kind of intellectual agency is expected. It tells staff what evidence to judge. It tells external examiners and regulators that the institution has thought about validity, reliability and fairness.
What does practical assessment redesign look like?
Assessment redesign does not mean killing curiosity. It means protecting learning.
In engineering, a student may use AI to generate early design options, but must submit a design decision log showing assumptions, constraints, rejected alternatives, calculations and safety considerations. The final assessment could include a short oral defence: why did you choose this design, what assumptions could fail, and how would you test it?
The learning outcome is not button-pressing. It is engineering judgement.
In business, students may ask AI to summarise market data, but must compare the output with original sources, identify unsupported claims and produce a decision memo. The assessment should reward source criticism, strategic reasoning and ethical awareness, not fluent summary.
A very smooth answer that cannot survive questioning should not score highly simply because it has attractive headings.
In computing, AI code assistance may be allowed for certain stages, but students must explain the architecture, test cases, security risks and debugging process. A code walkthrough can reveal very quickly whether the student understands the system.
AI may write code. It cannot yet defend design trade-offs with the nervous honesty of a student who has actually built something. Let us not rush to give it that job.
In humanities and social sciences, essays can still matter, but they need stronger scaffolding. Students might submit a research question, source map, annotated bibliography, draft argument, peer review response and final essay. They may use AI to generate counterarguments, but must evaluate whether those counterarguments are historically, philosophically or methodologically sound.
In teacher education, students could analyse AI-generated lesson plans against curriculum aims, inclusion requirements, cognitive load and safeguarding principles. This teaches responsible AI use while preserving the professional judgement expected of teachers.
Across all these examples, the principle is the same.
Universities should place friction at the point where learning happens.
Not pointless friction. Not administrative suffering, which universities have already mastered beautifully.
Educational friction: the kind that requires students to think, justify, test, revise and defend.
Why does a university need an AI operating model, not only an AI policy?
A serious university response needs an operating model, not simply an AI policy.
I would propose a Standards and AI Assessment Governance Framework with six components.
1. Create a cross-institution AI and assessment group with real authority
This group should include academic leadership, quality assurance, registry, digital learning, information governance, library services, student representation, disability support, careers and, where relevant, professional accreditation leads.
If AI is owned only by one enthusiastic digital education team, the institution may produce excellent workshops but weak system change.
2. Require every programme to map assessment against learning outcomes and AI-use categories
Programme teams should answer:
Which outcomes require independent demonstration?
Which outcomes require responsible AI use?
Which assessments are most vulnerable to invalid inference?
Where do students demonstrate process, not only product?
Where can professional or authentic tasks strengthen evidence?
3. Update the policy architecture
AI guidance must connect to academic regulations, assessment regulations, misconduct procedures, external examiner guidance, programme approval, periodic review, data protection, accessibility and procurement.
A lonely AI policy is like a door without a building around it. It may look useful, but it does not protect much.
4. Build staff capability
This means structured development, discipline-specific examples, peer review of assessment redesign, support for assessment literacy and proper time.
Universities should not ask staff to redesign the future of education between marking, meetings, recruitment calls and the printer failing again.
5. Support students as partners
Every student should receive clear AI induction.
Every assessment brief should state the AI-use category.
Students should learn how to acknowledge AI use, verify output, protect data, avoid intellectual dependency and preserve their own voice.
6. Create an assurance dashboard
It should track programme redesign coverage, staff development participation, student understanding of AI rules, external examiner comments, appeal trends, misconduct patterns, accessibility implications, approved tool access and examples of good practice.
These measures will not be perfect, but they move governance from anecdote to evidence.
What should external examiners, professional bodies and boards ask?
External examiners and professional, statutory and regulatory bodies need to be brought into this conversation more explicitly.
If a programme redesigns assessment for AI, external examiners should not merely comment on samples of marked work after the event. They should be asked whether the assessment pattern gives defensible evidence of programme outcomes.
Professional bodies also matter. In accredited courses, the question is not only whether a student can produce a good submission. The question is whether they can practise safely, ethically and competently.
AI can support professional practice, but it must not conceal the absence of professional judgement.
This is especially important in engineering, health, education, law, social work, finance and computing.
Boards and senior leaders should ask better questions.
Not only: do we have an AI policy?
That is too easy.
Better questions are:
How do we know students understand it?
How do we know staff apply it consistently?
Which programmes have redesigned assessment?
What evidence has changed?
What are external examiners saying?
Where are the appeal risks?
Where are the equity risks?
What is our position on approved tools?
What is the cost of not providing access?
This is where leadership maturity shows.
A weak institution looks for quick reassurance. A strong institution looks for uncomfortable evidence.
That is not pessimism. It is proper assurance.
What does leadership require?
My leadership stance is clear.
I am not arguing for a defensive university that treats every AI tool as a suspicious visitor. Nor am I arguing for a fashionable university that adopts every tool because it has a demonstration video and a procurement discount.
The right stance is innovation with a spine.
Universities should be open to AI where it improves learning, feedback, accessibility, student support, employability and efficiency.
They should be strict where AI threatens validity, fairness, privacy or professional competence.
That is not contradiction. That is judgement.
Good leaders in this area must hold two ideas together.
First, students need AI literacy because their future workplaces will expect it.
Second, students still need independent disciplinary understanding because AI without human judgement is not intelligence; it is acceleration. It can accelerate good thinking, and it can accelerate nonsense with equal enthusiasm.
The Institution of Engineering and Technology has argued that education should not only teach AI use, but also ethics, transparency, environmental impact and understanding of how the technology works [11]. That is the graduate attribute universities should want: not passive consumers of outputs, but critical users of powerful systems.
My recommendation is practical.
Build institutional confidence by being honest.
Do not promise that AI detection will save standards.
Do not pretend that old assessment patterns are all fine.
Do not terrify students into secrecy.
Do not leave staff alone to improvise.
Create a governed space where redesign, integrity and innovation work together.
What must a university award still mean?
AI does not remove the responsibility of the university. It sharpens it.
A university award must still mean that a student has demonstrated knowledge, understanding, skills and judgement at the required level.
It must still mean that assessment was valid, reliable and fair.
It must still mean that standards were protected through human academic judgement, supported by strong systems.
It must still mean that the institution can defend its decisions.
But protecting standards does not mean freezing assessment in the past. Some old assessment designs were never as strong as universities liked to believe. AI has made the weakness harder to ignore. That is uncomfortable, but useful.
It gives universities an opening to design better assessment: more authentic, more transparent, more process-aware, more ethically grounded and more connected to the world graduates now enter.
The universities that lead this next phase will not be those that shout the loudest about misconduct or buy the most impressive tools. They will be those that ask better questions.
What is the learning outcome?
What evidence proves it?
What role may AI legitimately play?
What must remain independently demonstrated?
How do we protect fairness?
How do we support staff?
How do we give students confidence without giving them permission to outsource thinking?
If universities answer these questions well, AI will not diminish higher education. It will force universities to clarify what higher education is for.
That, for me, is the leadership opportunity: not panic, not hype, and not a detector-shaped comfort blanket, but a serious redesign of assessment, governed by standards, guided by evidence and carried by human judgement.
Self-reflective critical questions for university leaders
- When students submit AI-assisted work, what evidence proves that the capability belongs to the student rather than mainly to the tool?
- Which learning outcomes must students demonstrate independently, and which should explicitly require responsible AI use because that reflects modern professional practice?
- Are institutional AI rules fair to students who differ in access, confidence, disability support, prior digital experience and institutional guidance?
- Could current assessment design be defended to an external examiner, regulator, employer, professional body and student appeal panel?
- Are universities using AI to deepen learning, feedback and critical judgement, or simply to make old processes faster and more convenient?
- What forms of human academic judgement must remain protected if university awards are to remain credible, respected and publicly trusted?
Final reflection
AI in UK higher education governance is not a narrow technology issue. It is a test of academic seriousness.
The duty of universities is not to preserve old assessments because they are familiar. It is to preserve standards because they are valuable.
The duty is not to pretend that every piece of writing is untouched by technology. It is to ensure that when a degree is awarded, the university can still say with confidence what the student has genuinely achieved.
The duty is not to choose between integrity and innovation. It is to govern innovation so that integrity survives.
The universities that lead the next phase of AI-enabled education will not simply buy more tools or issue the sternest warnings. They will ask better questions, redesign assessment intelligently, support staff and students honestly, and build governance systems strong enough to protect trust.
If they do that well, AI will not weaken higher education. It will force higher education to clarify its purpose.
In my view, that is not a threat. It is a demanding and useful invitation.
References
[1] Higher Education Policy Institute and Kortext. Student Generative AI Survey 2026. HEPI Report 199. Published 12 March 2026. Available at: https://www.hepi.ac.uk/reports/student-generative-ai-survey-2026/
[2] Jisc. Trends in assessment in higher education: considerations for policy and practice. Published 2025. Available at: https://www.jisc.ac.uk/reports/trends-in-assessment-in-higher-education-considerations-for-policy-and-practice
[3] Office for Students. National Student Survey 2025. Published 2025. Available at: https://www.officeforstudents.org.uk/publications/national-student-survey-2025/
[4] Office for Students. Condition B4: Assessment and awards. Available at: https://www.officeforstudents.org.uk/publications/regulatory-framework-for-higher-education-in-england/part-v-guidance-on-the-general-ongoing-conditions-of-registration/condition-b4-assessment-and-awards/
[5] Quality Assurance Agency for Higher Education. UK Quality Code for Higher Education 2024. Published 2024. Available at: https://www.qaa.ac.uk/quality-code
[6] Quality Assurance Agency for Higher Education. Academic Integrity Charter for UK Higher Education. Available at: https://www.qaa.ac.uk/about-us/what-we-do/academic-integrity/charter
[7] Scottish Funding Council. Scotland’s Tertiary Quality Enhancement Framework: TQEF Toolkit. Published 2025. Available at: https://www.sfc.ac.uk/assurance-accountability/learning-quality/scotlands-tertiary-quality-enhancement-framework/tqef-toolkit/
[8] Medr. Strategic Plan 2025–2030. Published 2025. Available at: https://www.medr.cymru/en/strategic-plan/
[9] Scottish Credit and Qualifications Framework Partnership. Quality assurance and framework information. Available at: https://scqf.org.uk/
[10] Jisc National Centre for AI. AI Detection and assessment: an update for 2025. Published 24 June 2025. Available at: https://nationalcentreforai.jiscinvolve.org/wp/2025/06/24/ai-detection-assessment-2025/
[11] Institution of Engineering and Technology. AI in education position paper. Published 2025. Available at: https://www.theiet.org/media/vummykiu/1243-ai-in-education.pdf
[12] Department for Education. Generative artificial intelligence in education. Updated 12 August 2025. Available at: https://www.gov.uk/government/publications/generative-artificial-intelligence-in-education/generative-artificial-intelligence-ai-in-education
[13] Department for Education. Generative AI: product safety standards. Updated 19 January 2026. Available at: https://www.gov.uk/government/publications/generative-ai-product-safety-expectations
[14] Jisc National Centre for AI. Staff perceptions of AI 2025. Published 2025. Available at: https://nationalcentreforai.jiscinvolve.org/wp/
[15] Jisc. AI maturity toolkit for tertiary education. Available at: https://www.jisc.ac.uk/guides/ai-maturity-toolkit-for-tertiary-education
[16] Ofqual. Principles of AI use in marking. Published 14 January 2026. Available at: https://www.gov.uk/government/publications/principles-of-ai-use-in-marking
[17] Quality Assurance Agency for Higher Education. Reconsidering assessment for the ChatGPT era: QAA advice on developing sustainable assessment strategies. Published 2023. Available at: https://www.qaa.ac.uk/docs/qaa/members/reconsidering-assessment-for-the-chat-gpt-era.pdf
[18] Quality Assurance Agency for Higher Education. Quality Code Advice and Guidance: Principle 1, Taking a strategic approach to managing quality and standards. Published 2025. Available at: https://www.qaa.ac.uk/the-quality-code/2024/advice-and-guidance-2024/quality-code-advice-and-guidance-principle-1
[19] Quality Assurance Agency for Higher Education. An evidence-based toolkit on leveraging Generative AI to support the graduates of the future. Published 2025. Available at: https://www.qaa.ac.uk/membership/benefits-of-qaa-membership/collaborative-enhancement-projects/generative-ai/exploring-the-opportunities-that-generative-artificial-intelligence-offers-for-higher-education/an-evidence-based-toolkit-on-leveraging-generative-ai-to-support-the-graduates-of-the-future
[20] Russell Group. Principles on the use of generative AI tools in education. Published 2023. Available at: https://www.russellgroup.ac.uk/policy/policy-briefings/principles-use-generative-ai-tools-education
SHARE
