It is compelling universities to confront the question of what is ‘higher’ about higher education beyond its linear progression from the compulsory education sector.  

Most universities define a graduate skills-set which encapsulates a holistic view of the outcomes their graduates will be able to demonstrate, often described as graduate attributes. These commonly include academic skills such as creativity, problem solving and analytics and critical thinking, human-centred skills such as adaptability, team working and collaboration, resilience and self-awareness and technical skills such as digital competencies.

However, academic curricula and assessment has, by and large, focussed on subject knowledge and skills and not the explicit development and assessment of these graduate attributes. The obvious question is how universities are actively, and intentionally, developing these skills in their graduates through their subject-focused curricula.

Vocational and technical disciplines which are preparing students for specific graduate careers have more robust responses to these questions. However, the challenge remains for academic disciplines more broadly to articulate how they are developing, and assessing, higher graduate skills.

Now that generative AI can perform… [lower order] functions for us, we can accelerate the trajectory of learning and assessment to focus on developing future skills, transferable graduate skills such as problem-solving abilities, building students' capacities for critical thinking, self-awareness and collaborative work, to name but a few. This renewed focus on what we might see as the core values of learning should surely be seen as a positive thing. 

Hughes and Linsey (2025) How AI may make us rethink the purpose of HE, QAA Blog.

Questions about the development of higher graduate skills are compounded by the fact that assessment in higher education has, for the most part, privileged the written word, valorising academic writing. With developments in generative AI, academic writing has become more easily accessible and therefore can be afforded a lower skill level. The intellectual argument remains sacrosanct, but (academic) writing becomes ostensibly obsolete as a measure of intellectual excellence. We argue this in our project blog:

If machine ‘intelligence’ can write an essay in an academic style or a report in a corporate voice identikit in terms of expression and tone, then why should we value, recognise and reward that ability as a higher skill?... When we have AI to do it for us, it appears it's no longer something we need to be able to do for ourselves.

Hughes and Linsey (2025) How AI may make us rethink the purpose of HE, QAA Blog.

These challenges must prompt a more explicit drive to define and assess the skills, as well as the knowledge, being developed in modules and courses, ensuring that these are effectively translated into the assessment criteria.

The question becomes ‘Is this assessment authentic and does it build the skills that are required in an Al enabled world?' 

How does AI support graduates of the future- an update on project progress 

A critical consideration of what is ‘higher’ about higher education inevitably leads us to question our ‘quality’ measures in the higher education sector. Generative AI is already challenging the appropriateness of some of our current quality frameworks. For example, the English Higher Education regulator, the Office for Students, expects providers to assess spelling, punctuation and grammar to “ensure the quality of students’ education” (OfS, 2021).

However, if Generative AI can deliver in this space, how can we argue that this is a ‘higher skill’ that needs to be assured?

Perhaps the most hotly debated challenge of Generative AI in the higher education sector is how universities ensure their graduates have gained the pre-requisite knowledge and skills required for their conferred degrees, and the authenticity of students’ work. 

One of the first actions that our institutions took after the public launch of Chat GPT in November 2022 was to secure academic integrity by identifying the submission of unacknowledged AI generated content as plagiarism or commissioning and providing guidance to students on how to avoid breaches of academic integrity.  

However, generative AI remains a significant threat to the integrity of university degrees. Universities cannot console themselves that AI generated assessments are not rigorous enough to pass. There is growing evidence that AI- generated submissions are going undetected and the grades that they are being awarded are gaining higher marks than those achieved by real students. Whilst higher education must critically engage with Generative AI, universities must also find ways to secure their assessment to ensure that students have met their programme learning outcomes. 

The resilience of assessment has been a key focus for our universities. Partners have begun to rethink assessment, some creating explicit tools to help staff stress test their assessments. As Generative AI technologies evolve, initial workarounds are becoming less reliable, and many types of assessments will present a low barrier to misuse.

As a result, we argue that the most sustainable approach will be to focus on authentic assessment principles where students are disincentivised to use AI as a tool to ‘do’ their assessment but are supported to use it effectively and appropriately (where it is integral to the assessment), with a greater emphasis on the process of completing the assessment and the skills set that this requires.

The opportunities Generative AI offers to higher education and the graduates of the future. 

Debates around the use of generative AI has inevitably focused on detection, whether with the assistance of a detection software or not. Whilst there are differing attitudes to detection software there is a universal recognition of the challenge of hybrid writing where standard tools used by students, and sanctioned by our institutions, have AI embedded within them.

Generative AI creates significant epistemological challenges in terms of how knowledge is both created, circulated and validated. 

Growing evidence supports the view that using generative AI to create outputs lessens the impact of human agency on epistemic development and creates what Sarkar (2024) defines as “mechanised convergence” where users have a strong tendency to accept AI output without critical judgement. The danger here, also discussed in more detail in challenge 5, is that knowledge circulated and validated becomes homogenised, reflecting dominant narratives and discourses, and threatening to negate multiple perspectives.

Generative AI can certainly empower…But by embracing this writing tool are we continuing to accelerate the tendency for a digitally determined realm of discursive homogeneity? Might we be making all those diverse voices start to sound the same – and even to say the same things? 

Hughes and Linsey (2025) How AI may make us rethink the purpose of HE, QAA Blog  

Perhaps more worrying is that users of generative AI are not aware of this dangerous predisposition. This is also the case in the field of education where, without guidance and support, students are using AI outputs without the required critical engagement.

Discussions with students in our own institutions have found that a significant number of students are not aware of how often Generative AI systems produce inaccurate or made-up facts, statistics or citations, produce biased outputs…Added to this is the lack of understanding by students (and staff) on how digital media of all types can be generated, and potentially manipulated by AI, underlining more than ever the critical and evaluative skills that are needed to work in the digital world.

The opportunities Generative AI offers to higher education and the graduates of the future. 

There is also growing evidence of significant cognitive implications of regular use of generative AI, where it exacerbates cognitive ‘atrophy’ or ‘offloading’, reducing the capacity for critical thinking and reflection. This is hugely significant for high education as these are exactly the graduate skills that we claim to develop in our students.  

In addition, recent work has revealed that, used for the purposes of ideation, generative AI can actually constrain creativity and ‘outside the box’ thinking, with the current configurations of AI tools being incompatible with encouraging truly original ideas (Anderson et. al., 2024).  

As generative AI tools become increasingly integrated into education University must foster environments that balance the benefits of AI with the development of higher cognitive skills enhancing, rather than diminishing, key graduate skills such as creativity, critical thinking and problem solving.

Perhaps one of the most fundamental threats from Generative AI is its impact on the trust relationships between universities and their students, and between students themselves.

The core of the issue lies is the fact that students are rapidly embracing generative AI for its perceived benefits in efficiency and work quality, while universities are struggling to keep pace with their approaches and polices. The concern here is that a vacuum is created, where both students and educators are uncertain as to what is acceptable use of generative AI and what is not.

Within this vacuum, a "trust deficit" has the potential to propagate, where students are expressing apprehension about the fairness and clarity of institutional guidelines, and educators become suspicious of the intentions of students whilst grappling with the pedagogical and ethical implications of AI in the classroom. As we navigate the AI enabled future, we must build a foundation of trust through transparency, collaboration, and clear guidance.

All our partners have acknowledged the central role that continual dialogue is playing in understanding how students are engaging with generative AI and their key concerns. Students’ AI literacies are clearly wide ranging, but many students remain extremely anxious about its (mis)use and about ‘getting it wrong’… These concerns necessarily need to feed into our strategic approaches to generative AI.

How does AI support graduates of the future- an update on project progress. 

Students need to be supported to understand the limitations of generative AI and recognise that whilst its use may be expedient and appropriate in some settings, educational settings require a level of criticality and authenticity not supported by, or delivered through, these tools.

Ensuring that our students know and understand our institutional guidance on Generative AI is a key priority for all partners, but we recognise that students are starting to challenge the efficacy and authenticity of policies and guidance as they increasingly use these tools in their own environments and perceive how they are being used in broader society. 

The opportunities Generative AI offers to higher education and the graduates of the future. 

Compounding these issues of trust, students (and staff) are also beginning to question and challenge their institution’s position to generative AI as it pertains to the protection of (their) intellectual property, environmental sustainability and broader issues of equity and fairness.

Generative AI is a powerful mirror reflecting the values, priorities, and biases of our societies, as far as they are represented across the global web. 

The uncritical pursuit of AI-driven productivity and automation poses a grave and multifaceted threat to the core principles of the Diversity, Equity, and Inclusion, principles that higher education claim to hold up. Dangers include embedding systemic bias into admissions, new and more complex digital divides, and the hyper-homogenization of knowledge.   

By virtue of the data on which they are trained, generative AI outputs are known to lack representation from women and people of colour. They also have a “language problem” performing poorly in non-English languages and are geographical challenged with knowledge created in whole continents being severely under or misrepresented.  

As a result, the exponential growth of Generative AI has the potential to reverse the progress made by high profile campaigns in global higher education which call for epistemic diversity through decolonising our higher education curricula and (re)valuing indigenous and local knowledge systems. 

However, with this challenge, comes an opportunity to reopen key debates about how to more effectively capture diverse knowledges through higher education curricula and pedagogy, addressing the destruction of other valid ways of knowing and understanding the world.  

One obvious opportunity is the challenge that generative AI is placing on text-based assessment. A move away from text-based assessment will require universities to embrace more diverse modes of assessment which, in turn, will encourage the development of other skills in students such as oracy, supported through a move to more dialogic learning and teaching. In turn, this could open up ways to (re)value the authentic voices of our students, supporting neuro-diversities and access to higher education for students with less formal qualifications.

Academic writing has been privileged over other forms of writing and viewed as more legitimate than different forms of expression, including oral traditions. However, the ability to write academically has for centuries been limited to a small group of people – academics themselves and others with access to that educational privilege and cultural capital (very often those whose parents and grandparents also enjoyed the benefits of university educations)…if generative AI leads to the diminished value of the art of academic writing, it may open  the door to, and restore, our understanding of the importance of dialogue, of dialogic and dynamic processes of teaching, learning and assessment which empower the diversity and distinctiveness of individual voices and the interrelation of those voices. 

Hughes and Linsey (2025) How AI may make us rethink the purpose of HE, QAA Blog.  

Whilst Generative AI has the potential to ‘level-up’, whether through embracing alternative forms of assessment, democratising academic writing or creating AI tutor tools which support students who feel marginalised in the academic system, knee jerk reactions need to be avoided that may reverse progress made in the inclusivity realm. Amongst others, these reactions may include a wholesale return to closed book examinations or abandoning anonymous marking. The impact of these changes across our diverse student communities needs to be carefully considered.

One of the key challenges for universities is how to embrace generative AI whilst mitigating the ethical risks which include issues of control, equity, environmental sustainability and the ethical implications in the global distribution of labour in the AI industry (see for example, Muldoon et. al, 2024).  

There are several complex copyright issues related to the development and use of generative AI tools, most pressingly that Generative AI tools are being trained on published and unpublished material with little regard for the copyright of the original content creators. This is being compounded by the fact that there is currently unclear statutory copyright protection of AI-generated outputs. 

For example, in the UK there is copyright protection for purely computer-generated works, but the UK Government have argued in their ‘Consultation on Copyright and Artificial Intelligence’ that “it is not clear that this protection is widely used, or that it functions properly within the broader copyright framework” (UK Government, 2024). Writers, artists and musicians, amongst other creatives, are becoming increasingly vocal about the implications of AI and copyright for the ownership of their artistry. 

Furthermore, the environmental impact of generative AI has been widely debated. Whilst its impact is undoubtedly complex and as not fully tested, we do know that is consumes a significant amount of energy and water, which is particularly impactful on a regional scale, and creates large amounts of carbon emissions. Inevitably the global’ price of knowledge creation will have significant and differentiated local impacts and no doubt will fall on the most vulnerable communities.

Whilst it remains very difficult to get accurate data on the environmental impacts of Generative AI (Crawford, 2024), there are a growing number of case-studies about how its exponential growth has placed already vulnerable communities in increasingly perilous situations.

QAA Project Blog (2024) The opportunities Generative AI offers to higher education and the graduates of the future. 

Whilst still embryonic, we are seeing a growing backlash by both staff and students against the use of generative AI.  

…we are beginning to see a backlash by both staff and students who are choosing not to use this technology, based on ethical concerns including bias, environmental impacts, and the infringement of intellectual property. 

How does AI support graduates of the future- an update on project progress. 

Currently, this opposition is relatively discipline specific, with students studying creative subjects allied to music, creative writing, design and art particularly unsettled about the implications of AI technologies for their industries and graduate prospects. However, there is no reason to suspect that this backlash will diminish in the foreseeable future.  

Whilst the big picture solutions are beyond the bounds of individual institutions and indeed the sector more broadly, universities must ensure that they support students to understand these ethical dilemmas. In many cases, students are still unaware that the content they are inputting into commercial AI platforms through prompts or uploads, may grant the tool the right to reuse and distribute this content, and/or may breach copyright and/ or their privacy.  

It is incumbent on all universities to ensure that their staff and students know and understand what Generative AI is, and what it isn’t, what it can and cannot do, and fundamentally that no matter how ‘clever’ these tools seem, they are not intelligent in the human sense. This understanding can help to define the higher-order skills that generative AI cannot deliver.  

As with the other challenges identified here, there is an equal opportunity for higher education educators to address social justice and sustainable futures in their programme syllabi, as they relate to AI, further supporting their long-standing commitment to both the protection of intellectual property and environmental sustainability. However, universities will need to tread a fine line between ensuring that students are prepared to work in an AI enabled world, whilst defending the key ethical dilemmas raised by AI more broadly.

The final challenge identified in this project is around embracing ‘big tech’. 

Whilst, it is not solely related to generative AI, it has certainly been crystallised by it. For some time, Global IT companies have been consolidating their place in the digital ecosystems of higher education with the widespread use of cloud based digital infrastructure – both hardware and software. 

For the most part, this tech has not been created explicitly for education and universities are having to acknowledge and mitigate against the challenges of using this technology for educational purposes.

Updates are being made to these technologies without due consideration to the academic cycle which can make it challenging to embed generative AI tools into learning, teaching and assessment, if features are being changed, updated, or withdrawn with little or no warning. 

How does AI support graduates of the future- an update on project progress 

One major point of tension is between the narratives deployed by the ‘big tech’ companies around the efficiency and productivity gains of generative AI tools, and Universities’ educational mission to develop the skills of criticality and creative engagement in their students. 

Whilst not innately incompatible, the high-level messaging of the tech companies can be problematic and directly contradict academic policies on academic integrity, as well as the quality of educational knowledge and the development of key higher order skills.  

Other challenges include those related to continuity and digital equity. Companies may offer deals to students directly offering enhanced capabilities and privacy protection for time limited periods or offer free access to advanced tools again for a limited time while the tools are in beta release.  

Students can use personal free accounts (including options not to login) or subscribe to access more capabilities, with two paid subscription tiers now commonly available.

In addition, some of the more powerful specialist tools and capabilities that are available such as the research assistance tools have access to pay walled journals – creating inequity between those who can pay and those that cannot.  

It is therefore essential that universities continue to monitor generative AI tools and platforms as they evolve and new capabilities are released, and assess the ongoing implications for learning, teaching and assessment.

It will be necessary to maintain universal access to licenced tools as our university communities engage with many different activities demanding enhanced assurance around data (such as creating feedback on student work, use in administration activities, use in open classroom situations).

However, these tools should not necessarily be regarded as fixed over time and it may be beneficial, within financial constraints, to consider the mix and range of tools available to students and staff, importantly including assessing specialist tools developed specifically for education and research, such as virtual learning environments integrated chatbots and research assistant tools, which may or may not be generative AI. 

In doing this though we need to be purposeful and carefully assess the impact, particularly the impact on student learning. What we do know for certain is that the range and capabilities of generative AI tools will rapidly evolve moving forward.

Anderson, B.R., Shah, J.H. and Kreminski, M., 2024, June. Homogenization effects of large language models on human creative ideation. In Proceedings of the 16th conference on creativity & cognition (pp. 413-425) https://doi.org/10.48550/arXiv.2402.01536 

Muldoon, J., Graham, M. and Cant, C. (2024) Feeding the Machine: The Hidden Human Labour Powering AI. Canongate Books. 

Sarkar, A. Xu, X. Toronto, N. Drosos, I, Poelitz, C. (2024) When Copilot Becomes Autopilot: Generative AI’s Critical Risk to Knowledge Work and a Critical Solution Sarkar_2024_GenAI_critical_v1.01.pdf