Author: David White

How to redesign teaching and assessment to respond to the abrupt appearance of AI has become a heated question for Higher Education. For some institutions, how to reliably validate learning has become an existential challenge. This is less the case for Arts Education which has always focused on process and the demonstration of learning, avoiding over reliance on a final ‘product’ in assessment. I believe that many institutions and subjects could now benefit from applying Arts Education thinking.

My role at UAL includes developing strategies to respond to the challenges and opportunities of AI, this post is my thoughts on the recently published QAA draft principles on academic standards in the age of AI rather than a UAL position. It outlines where Arts Education maps well to the draft principles, and where the territory is still problematic for all.

Arts Education is (part of) the answer

Many institutions have developed AI policies, positions and principles. How successful these are is not yet clear, but I have seen a recent trend towards Principles (including our approach at UAL) as a response which can weather the noise around AI’s potential impact.

MIT recently published an interesting report on ‘AI Use in Teaching, Learning, and Research Training’ which leads with Principles. The UK Quality Assurance Agency for Higher Education also recently published ‘Assuring Academic Standards in the Age of AI: Draft sector principles for assessment validity and standards’.

What both have in common is that an Arts Education approach to teaching and assessment hiding in plain sight. Neither expresses this view directly, however, the educational practices best placed to respond to the reports look distinctly Arts Education.

Responding to the QAA

To illustrate my point, I’m going to run through some of the headlines in the QAA draft report and describe how a well-informed Arts Education approach is a valid response to much of what has been drafted.

This is not to say that we don’t have a significant amount of work to do at UAL, or that our approach to teaching and assessment is always as robust or confident as it could be. I’m drawing from the many practice-based discussions I’m party to, rather than an exhaustive knowledge of everything Teaching and Learning that happens across the institution.

The underpinning approach in the QAA draft principles

It’s reasonable to design a chunk of assessed learning banning the use of AI as an engine in the production of outputs (as distinct from AI use in ‘finessing’ work, such as grammar checks or for access purposes). It’s also reasonable to specify the use of a particular genre of AI tool where this use is the focus of the learning.

However, the QAA draft principles are attempting to outline practices which maintain learning as a process on the assumption that students are using AI on a day-to-day basis and probably in ways which will damage their learning if this risk has not been considered in the design of their learning. The draft principles take the position that detection of AI use or policing of use ‘after the fact’ (i.e. after work has been submitted) is not an effective strategy to maintain academic integrity.

The emphasis then is on the design of learning and assessment which is inherently both inclusive and discourages damaging (in educational terms) AI use. The aim here being to respond to the cause through robust design, rather than the symptom through policy and regulations which are likely to be unenforceable.

I agree with this approach, but it does raise some difficult questions.

1. Not relying on the academic integrity of students is an erosion of trust. However, this is less the case when students are not clear on where AI use starts to damage the learning process. The design of learning and assessment should, in some senses, manage this on the student’s behalf. i.e. It’s not fair to ‘trust’ students to make pedagogic judgements but critical literacy should be taught in some form to support students in identifying how/where to stay intentional. (See below for my video of my recent talk on this at MTU)
   
2. It’s genuinely difficult to imagine a way of managing ‘some but not all AI is allowed’ learning and assessment that responds to these draft principles. Even defining the edges of what genres of AI and what practices are allowed is tricky. Designing the pedagogy that can then identify those edges and keep things on track is even harder.
   
   The ‘some but not all’ approach probably works ok where specific AI skills are being taught directly but where an academic practice is much broader, such as essay writing, this middle way tends to put the burden on the student to interpret a complex and rapidly changing territory (or worse they have to second guess the interpretation of the academic who might be marking their work).
   
   Again, this is where the teaching of critical literacy is crucial. As is equipping students to develop strategies which maintain their agency and intentionality. This is no small ask within curriculum which is already likely to be full to the brim with subject specific content and assessment.
   
3. Most effective or legitimate responses to the draft principles are likely to put more pressure on teaching staff. Adding ‘meaningful moments’ of accountability which build towards a final submission requires time and focus. Similarly, shifting towards more dialogue-based forms of assessment is difficult to do within workload models where time allotted per student has been trimmed down over the years to balance the books.
   
   At UAL we have spent years developing systems and practices which allow us to undertake this form of teaching and learning at scale but it’s not always easy and it is not cheap.

The Draft Principles where an Arts Education approach works well

I’m going to group some AI specific extracts from the draft principles rather than going through them one by one in full.

From draft principle 1: Assessment provides valid evidence of learning at individual assessment and programme level

“ Assessment is designed to reasonably verify the student’s own learning and capability, rather than the output of an AI tool, against the learning outcomes, including, where AI use is permitted, the student’s direction and critical evaluation of that use.“

From draft principle 3: Summative assessment includes proportionate points of assured demonstration across the programme

“Providers identify meaningful points across the programme where independent capability must be demonstrated… That demonstration can take different forms, such as oral, practical, observed, placement-based, simulation, presentation and questioning, or structured defence of an artefact.”

From draft principle 4: Assessment drives learning, building foundational knowledge, skills and competencies over time, valuing process alongside outcome

“Students are assessed on how they reach, justify and refine their work as well as on the final product. Feedback is timely, effective and designed to be applied to improve later tasks.”

From draft principle 7: Assessment security is supported by realistic, meaningful and transparent academic regulations and accessible policy and regulation, not by academic integrity alone

“Assessment security consists of meaningful actions to prevent and detect attempts by students to avoid accountability for their learning. Detection here means gathering evidence of learning, for example through questioning or evidence of process, rather than reliance on AI-detection software.”

How UAL education maps to the points above

Almost all assessment at UAL is the wholistic assessment of a creative portfolio with other elements such as written reflections and/or creative artefacts. Most of our teaching is geared around project-based learning where a creative brief is set and then project work is assessed against Learning Outcomes which are geared to our assessment criteria.

This approach embeds several elements which respond well to the substance of the QAA draft principles noted above:

1. Project-based learning inherently builds incrementally toward summative assessment.
   
2. In keeping with the project-based approach, teaching incorporates ‘meaningful points’ where the work-in-progress is discussed and tested. These commonly act as moments of accountability where students will demonstrate ‘independent capability’.
   
   Often these moments are dialogue based, incorporating formative feedback which builds understanding and knowledge towards the final submission. This is rooted in studio-based pedagogy where work-in-progress is developed visibly and shared within trusted groups for reflection and discussion.
   
   This is not to say that all UAL education is centred around a physical studio, it’s more that the philosophy of ‘studio’ is woven through much of our offer.
   
3. One of our five assessment criteria is ‘Process’. The portfolio constructed by the student is used not only to defend the validity of a final output but also to tell the story of how that output was arrived at. In effect, the portfolio documents the decision making of the student. It is a record of critical thinking and therefore strong evidence of learning-through-process. Other elements such as written self-reflections add more opportunity to describe ‘what was learned’ to go beyond describing ‘what was produced’.
   
   If AI has been used as process then the creative/project portfolio should contain an account of how this use influenced learning/understanding/identity. A list of AI technology used with no critical reflection is of very little value and can easily become a non-enforceable form of policing ‘after the fact’. Additional constructive approaches to acknowledging AI use are outlined by Borowik, Smith, and Birkett on the UAL AI Network blog.
   
4. Another UAL assessment criteria is Realisation. This is focused on ‘standards of production’ and as such is weakened by AI as a proxy for learning, especially in the light of the draft QAA principles. For many of our students Realisation shakes out as ‘how good is the thing I’ve created’ or ‘to what extent would I want to include this in my professional portfolio when seeking work’.
   
   Ian Truelove and I have been discussing how AI pushes many aspects of Realisation outside of ‘evidence of learning’ and reemphasises the importance of Process. One response to this would be to design learning wherein Process is awarded credit and Realisation is validated by industry professionals (which includes many of our tutors). The same people might be assessing both criteria but within distinct framings which helps to reduce the temptation for students to over focus on product over process.

Problematic aspects of the draft principles

From draft principle 2: Assessment is inclusive and accessible by design

“Assessment is designed to be inclusive and accessible from the outset, rather than adjusted after the fact.”

Clearly this is a good idea. My question though is about AI resistance or refusal. If the course of study is not optional or not obviously signalled as incorporating AI is it acceptable to mandate AI use as evidence of meeting the learning outcomes?

In other words, does inclusive assessment demand learning outcomes that can be met without the use of AI? I think this should be the case as resistance and refusal are defendable positions within the framework of most universities stated values.

From draft principle 5: All assessments have a clear and justifiable design rationale for their position on AI use.

“Because students may have the opportunity to use AI regardless of whether it is permitted, where the validity of an assessment depends on AI not being used, this is secured through the conditions of assessment rather than by instruction alone.”

From draft principle 6: Prohibition and detection are not reliable or effective approaches

“Given this evolving and contested evidence base, providers do not rely on detection software as a primary safeguard; the more durable response is robust, valid assessment design, combined with meaningful assessment security actions, rather than after-the-fact policing.”

From draft principle 7: Assessment security is supported by realistic, meaningful and transparent academic regulations and accessible policy and regulation, not by academic integrity alone

“It is insufficient, and unfair, to set rules that providers have no practical means of verifying – for instance, distinctions between permitted and prohibited AI use that cannot be evidenced, checked, or distinguished after the fact – and then rely solely on students’ individual academic integrity to ensure compliance. This approach transfers all responsibility for assessment security onto students, and risks penalising honest students who gain no advantage from AI use while others who breach the rules go undetected.”

I agree that AI detection software is not effective. I also agree that our response to AI should be to improve the quality of our teaching and assessment rather than find ever more detailed, and yet vague, methods of policing ‘after the fact’. This is an approach we have taken at UAL from day one and our guidance has been developed on that basis.

The problem here is if there is no way of detecting AI use then what is a valid ‘assessment security action’, surely it can’t be ‘make everyone sit a paper-based exam’? Also, if we shouldn’t be policing ‘after the fact’ then how are we avoiding permitted or prohibited AI use that ‘cannot be evidenced, checked, or distinguished after the fact’? This latter point somewhat invalidates the traffic light approach to permitted AI use in assessment as anything in ‘amber’ is practically impossible to verify/check (This is an area we are currently in discussion about at UAL).

Additionally, it’s indicated that we shouldn’t only rely on students’ academic integrity as this is unfair on the students that have academic integrity. This nets out as ‘don’t trust the students because this is unfair on the students you can trust’, which leaves us at: ‘the fairest approach is to not trust students or technology’. It’s difficult to know how to respond to this in day-to-day teaching and assessment practice.

A way out of this knot

Rather than get tangled up in the self-defeating nature of this I turn to the excellent ‘Trained to stop learning’ report form WonkHE. The report highlights the effectiveness of ‘points of accountability’, a pragmatic description of what the QAA principles refer to as ‘meaningful points’ . These are exactly what they sound like, points where students are required to explain their work. The WonkHE report suggests this can work well peer-to-peer as well as peer-to-tutor. A few moments of accountability (or what I’d describe as ‘Designed Friction’) reduce the temptation to overuse AI as it quickly becomes clear where any cognitive offsetting has eroded an individual’s ability to ‘speak to the work’.

However, asking every student to account for their learning through dialogue will disadvantage some and, as ever, offering a choice of how to make that account is good practice. There is a long history of accounting-through-dialogue at UAL, the arts crit being the classic form. There is also an acknowledgement that the practice can be distressing if not undertaken carefully and plenty of examples of practice which responds to this. For example, work on compassionate crits and other supportive forms of communication.

A difficult principle

From draft principle 8: Staff are supported in valid assessment design through ongoing professional development

“Teaching staff, and staff who support assessment design, are supported to develop and maintain competence in developing and delivering discipline-specific approaches to assessment that secure academic standards…”

This principle goes on to acknowledge that this is an ‘evolving area’ so not to rely on ‘one-off or definitive training’. We have a healthy staff AI Network at UAL which is full of lively discussion and practice sharing. I’ve also been running an UAL AI Principles workshop which has had around 300 attendees over six sessions. However, it’s often the case that our most anxious and/or busy staff won’t connect with this kind of support and it’s hard to sell the idea of staff training which might go out-of-date almost immediately.

Having said this, even over the last six months discussion at the Principles workshops has shifted from broad concerns to pragmatic responses. The overall response being cautious engagement combined with an acknowledgement of resistance and refusal as legitimate positions.

The enhancement principles

There are three of these in the QAA draft. For now, I’ll just highlight an extract from the third one as this is suggesting something I’ve been advocating for:

 “Overloaded portfolios and an imbalance toward summative assessment drive up both student and marker burden while narrowing what is actually being tested…This is a matter of purpose rather than volume…”

What we know only too well at UAL is that our assessment practices are time-consuming. We also know that any student who feels like they ‘have to get through the work’ is more likely to lean on AI. Add to this that the more flexible the offer is the more assessment points are required to award small chunks of credit. This can rapidly spirals up to an enormous and ever-changing amount of work for everyone involved.

‘Less but better’

I worked with Ian Truelove and colleagues in UAL Online to developed an assessment framework for the group which asked for a much smaller volume of work from students. The focus was ‘less but better’, and it appears to be working well.

Most discussion around AI and academic standards points to the downsides of our habit of upping the wordcount as students go up through the levels. Academic rigour is not proportional to the volume of output. I would argue that the emergence of AI strengthens the opposite view.

Learning as becoming

When we consider the teaching and assessment practices which respond positively to the risk AI poses to the learning process, it all tends towards a learning as becoming model of education. Assessment becomes about evidencing a change in the person; evidence that the learning has become part of who they are rather than an object located in a nearby computer/phone. This has always been central to Arts Education and it’s fascinating to see institutions with a predominantly STEM focus lean into it. For example, this from the MIT report which could easily have been published by UAL:

“If learning is (mis)understood primarily as content delivery, then highly responsive AI systems may seem capable of replacing teachers, mentors, peers, collaborators, and the interpersonal dimensions of learning altogether. But while this attitude may be increasingly common, it’s shortsighted. Education is a cultural practice through which students learn to make meaning, exercise judgment, form identities, and participate responsibly in community.”

Now there is a machine which can analyse and produce most forms of content we are challenged to reassert an education which goes beyond that. Ironically, universities have always made claims around meaning making and identity formation (becoming) but narrow skills agendas, massification and academic career pathways which favour research over teaching have enticed many an institution to shift the day-to-day experience of learning towards content consumption and reproduction. We are in no way immune to those pressures at UAL but our focus on practice-based education and subjectivity means our approach has retained a focus on becoming and experience which AI has made increasingly pertinent beyond Arts Education.

Perhaps the single most positive thing about AI is the way it is forcing institutions to develop a form of education they claimed to be offering all along.

It’s widely accepted that the process of learning requires some level of difficultly. Or, to put it another way, 100% ‘frictionless’ learning isn’t learning. Separately, the generally accepted role of technology is to help us to get more done with less effort. Bring these two principles together within education, add a solid layer of market forces and you see some strange effects. The latest one being the emergence of Cognitive Surrender, which introduces the possibility of employing AI to gain educational credit while simultaneously attenuating your ability to think. This is a problem.

We could level our critique at the designers and owners of mainstream AI. That is certainly worth doing but we should also consider the character of what we offer, our education. Arguing that ‘overuse’ of digital technology is not the proper way to learn is not going to do all the heavy lifting here. The narrative of efficiency-as-success is now so dominant that any counter must be carefully constructed to avoid appearing archaic or pro inconvenience.

The friction used to be inherent, now we must explain it

As a colleague said to me recently: “You don’t need to explain why things are difficult if there is no easier option”. When the only way to find a book is to rifle through numerous pieces of cardboard in a draw, you don’t question it, you learnt how to do it as well as possible. When writing meant getting your line of argument planned out before putting pen to paper, you didn’t think ‘writing an essay is a bit linear’, you just attempted to produce a decent first draft.

What’s important here is that the friction was built into the process. You had to be intentional, or the quality of your work would suffer and there was the ever-present risk that you’d waste a lot of time taking unconsidered routes. As digital technology has improved the efficiency of many academic and creative processes that friction has been ironed out, increasing the opportunities to be un-intentionally productive, that is, to produce work without experiencing learning.

The expert can manage this, as they have the experience and knowledge to judge where to smooth the way and where to dig in and engage. The student or novice does not yet have this ability and requires the educational process to incorporate ‘designed friction’ to ensure learning takes place.

However, because old-school frictions were not understood as useful for learning, but simply part of a system with no other choices, it means that we haven’t rehearsed how to explain the value of friction in the process of education.

The role of Critical Literacy

In my recent keynote at the KRIK conference I suggested that Critical Information Literacy and associated pedagogy is a useful counter to the risks of Cognitive Surrender. The ‘critical’ aspect focusing on verifying information through experience (or knowledge of context), testing the useability of information rather than simply gaging ‘quality’, actively questioning meaning and intent.

Mainstream AI is a model of language and language is a limited expression, or model of, experience. Therefore, mainstream AI is a model of a model. Useful but operating at a significant distance from that which it appears to profess. It requires us to imbue automated outputs with meaning through the lens of our lived experience and understanding. This process is a form of Critical Literacy.

Criticality is not instinctive, it must be taught

During my talk I complained that many universities demand that students ‘think critically’ when using AI, as if that was a skill any novice could switch on at will rather than a literacy that is developed through experience, something which should be taught. I claimed that teaching Critical Literacy is crucial in the AI era and is now on par with the importance of subject specific teaching.

I argued that the subjective nature of Arts Education means that much of our pedagogy has always been a form of Critical Literacy, as the strength of student work is assessed by the ability to account for its genesis (process) and demonstrate intention in its production. In Arts Education an element of this is specific to each piece of work and therefore connected to the lived experience of the student or group.

Outside of Arts Education, the classic essay or ‘attempt’ should also function in this manner but in a massified system the practice of defending your argument has become more of a process of arguing ‘correctly’, something which can now be automated. In my view, a lot could be learn from Arts Education in the AI era as we have held onto forms of teaching and assessment which are now relevant across the sector.

Work to be done

This is not to say that Arts Education does not need to get better at explaining its own ‘designed frictions’ or that we are not vulnerable to un-intentional production. We have a responsibility to teach relevant digital skills and equip students with critical frameworks that inform where they build friction into their practices.

For libraries, the focus of my KRIK talk, I argued that the aim is not to attempt to be as seamless as AI, as this would defeat the purpose of what the library stands for. Instead, we should celebrate and explain well designed seams and frictions, one of which is the varied expertise and insights of library staff.

I’m delighted to share that I’m now have the additional role of Reader in Postdigital Education at the University of the Arts London.

This is a teaching career Readership which I became eligible for because of a shift towards what REF 2029 calls “New Insights, Effectively Shared”. Thanks to Trevor Keeble for taking UAL in this progressive direction and Silke Lang and Mark Ingham for designing the criteria and supporting the application process. Thanks also to Ian Truelove for being an excellent sounding board and to Susan Orr for years of support.

Reader in Postdigital Education is an achingly academic title so I feel I should explain.

The Reader part essentially means ‘UK associate-professor’. The strict interpretation of the term, I assume, is that I’m reading my field before I go out and start professing about it. In practice, I do a lot of ‘professing’ via my blog and speaking invites. I will use the new role to build on this and develop various research outputs.

The ‘Postdigital’ part is in the ‘post’ family of Postmodern and Post-punk. As in, it is not implying that the digital is finished, but rather that it does not exist in isolation. Using ‘post’ as a prefix is troublesome but in a positive way it is exactly that trouble which I want to stay with. A good way of putting it is as:

“…a term that sucks but is useful.” (Cramer, 2015, p. 13).

Postdigital describes an approach or method rather than a state which makes it a useful framing for research into emerging phenomenon.

Starting in the arts then moving to education

Fittingly, Postdigital as a concept has roots in creative practice, an early use describing the generation of music from the glitches, or failures, of digital technology (Cascone, 2000), but since then the term has evolved and become a field of research in its own right. The Journal of Postdigital Science and Education being a key location.

More specifically, in an educational context the Postdigital highlights that:

“…contemporary student practices with technology are complex entanglements between physical and digital technologies, spaces, activities, and time.” (Jandric, et al, 2018).

This framing is useful in countering the cultural/commercial habit of implying ‘digital’ is always ‘new’ and ‘future’. It uncovers ‘inevitable’ digital futures as commonly a vehicle for standard models of capital and power.

The 52group (Postdigital and education from way back)

I’ve been thinking Postdigitally since 2009 when myself, Ian Truelove, Lawrie Phipps and Richard Hall (The mysteriously named ‘52group’) met and co-opted the term for online/networked education.

Our use of the term built on the idea that the digital, especially the internet, was already woven into almost all aspects of life and learning. This was quite a bold claim just three years after Facebook has been launched for the general public but we were becoming impatient with an obsession in digital education with every new technology that came along.

It’s difficult to recall now, but at that time there was a feeling that the Web was a novel frontier which was free from the baggage and politics of our stuffy old institutions. We hadn’t yet understood the extent to which our data was becoming commercially valuable. In education there was a somewhat uncritical obsession with technological innovation and how this could be applied in teaching and learning.

Our 2009 provocation, “Preparing for the postdigital era” now reads like an early call to be more human-centred and weary of the tech gurus:

“Postdigital aims to throw off the yoke of digital dogma, where the language of a perceived digital elite drives not only development, but also skews innovation, where innovation is only seen as being that associated with the “latest” technology… …Innovation in a postdigital era is more effectively articulated as being associated with the human condition and the aspiration toward new or enhanced connectedness with others.” (52group, 2009)

What has changed since then is a broad cultural shift from a largely unquestioning acceptance of the convenience being sold via the digital to a position of scepticism and, in some cases, fear as each new ‘revolution’ is presented as inevitable. That is not to say that we don’t use the ‘revolutionary’ solutions, use continues to increase, but now with a suspicion that the overall direction of travel will benefit the providers of the technology more than us.

The Postdigital era really did arrive. It’s complex, messy and requires serious thought.

52group.(2009). Preparing for the postdigital era http://docs.google.com/View?id=aqv2zmc9bgm_51ft65rbn2 accessed August 14^th 2026.

Cascone, K. (2000). The aesthetics of failure: ‘post-digital’ tendencies in contemporary computer music. Computer Music Journal, 24(4), 12–18.

Cramer, F. (2015). What is ‘post-digital’? In D. M. Berry & M. Dieter (Eds.), Postdigital aesthetics: Art, computation and design (pp. 12–26). New York, NY: Palgrave Macmillan.

Petar Jandrić , Jeremy Knox, Tina Besley, Thomas Ryberg, Juha Suoranta & Sarah Hayes (2018) Postdigital science and education, Educational Philosophy and Theory, 50:10, 893-899, DOI: 10.1080/00131857.2018.1454000

This is an expanded version of a talk I gave at the Postdigital backlash conference in Zagreb in June 2026. My proposal is that while the backlash (refusal or resistance) to Gen AI is commonly based on notions of stolen labour underlying this is a powerful reaction to the simulation of Person. The, in some terms, successful simulation of Person in mainstream forms of Gen AI extends the hyperreality of the digital environment to a point at which it approaches saturation. This all encompassing hyperreality has no exit as the final referent to the real of ‘Person’ has now been erased. We cannot differentiate between person and non-person (phantom) where our attempted access to the real is mediated by the screen.

I also reference backlash based on other factors such as bias in outputs and environmental concerns but maintain that our innate sensitivity to the eerie is always present wherever the technology has been designed to perform a synthetic Personhood. This, I suggest, is one of the reasons that we feel shame or guilt when using Gen AI as we struggle with the dissonance of engaging with a non-person and cannot identify ‘who’ has produced the work.

I suggest that work created with Gen AI, while not fundamentally without value, lacks aura because a route cannot be traced back to an identifiable author or authors. In essence, work produced with Gen AI will always be a simulacra.

I variously discuss how this plays out in creative arts education and propose an alternative conception of Gen AI and education which avoids the abject state that many find themselves in when using the technology. Abject, because the act of production is efficient and ‘sensible’, while concurrently unsettling because we engage in person-like dialogue with that which we know not to be alive.

Much of the talk is thinking-in-process rather than precisely crafted lines of argument. Making the video has helped me to move my thinking forward in a territory where there are no clean answers, only the opportunity to navigate complexity.

Part 1 used the “Where are the people?” question from an AI Principles workshop to highlight the accountability of the designers and managers of AI, especially where that technology is presented and promoted as having a ‘mind of its own’. Part 2 is a review of activity which took place two days after I ran the workshop.

On the 2nd of July, we hosted an AI and Creative Education symposium at the University of the Arts London which reframed the “Where are the people?” question once again.

The second answer to “Where are the people?” is that they are at UAL, questioning, experimenting with and critiquing AI. The symposium demonstrated that our staff and students have waded into AI creatively and critically. Yes, there is a feeling of overwhelm, there is also a lightly-held confidence that art and design education is well placed to navigate the complexity of AI creatively, ethically and politically. UAL is already imagining a range of human centred futures which include AI, rather than accepting wholesale the singular future articulated by Silicon Valley. There are AI skills to be taught and learnt. There is also a need to sense make and collectively develop meaningful practices which respect learning as a process and amplify creative agency.

The symposium was designed by Rupert Norfolk, Darryl Clifton, Ian Truelove, Chris Rowell and myself with organisational support from Lynn Finn. It was an internal event for UAL staff, and we had around 130 attendees of which the biggest group was Course Leaders and Programme Directors. Just about everyone who attended is directly involved in teaching and learning, plus people from our Digital and Technology department.

To give you a sense of what we covered take a look at the schedule for the day. I’m not going to review the day session by session as that would be book length. Instead I will highlight a non-exhaustive set of themes I saw emerging. I’m sure others attending would highlight additional lines of thinking.

1. UAL Expertise: Apart for a keynote from Mack Marshall of Wonkhe on their excellent ‘Trained to Stop Learning‘ report all other speakers were UAL staff. This made visible the significant, and often cutting edge, AI expertise at the university.
2. Use, Resistance and Refusal: While the symposium took a critical stance towards the mythologies surrounding AI, there was a clear understanding that it is being used by almost all staff and students in some form. There was also an understanding that informed resistance and refusal are important. Many students are uncomfortable about the use, and in some cases, the existence of the technology. The institutional aim is to support students to be well informed whatever position they take.
3. Intentional Practice: It’s not difficult to use AI to produce ‘polished’ work, or to use it to produce a significant volume of material (text, image, sound, video etc). Neither of these are a central interest to an arts university where the intention of the work created is key (and a large part of what is assessed). The speakers at the symposium tended to focus on how to retain this intentionality while using AI rather than ‘how to prompt AI well’ which is extremely context specific.
4. Meaning over Myth: Most speakers started by distinguishing between the grand narrative of AI as told by Silicon Valley and the underlying principle of how the technology functions. There was a broad sense that AI is extremely interesting, powerful and full of creative potential but that the main ‘frontier’ models present risks, such as diffuse claims of intelligence and engineered modes of engagement which are anthropomorphised and often sycophantic.
5. Local and Small Scale: Quite a few sessions highlighted the potential of locally installed models which give the user more agency, in that you can choose what the model has been trained on, track energy usage and often manipulate the model directly. Using undisclosed amounts of energy and taking advantage of data which might not have been given willingly is not the only option when engaging with AI. I can foresee an interesting possible future where UAL (and/or the UK Higher Education sector) promotes local approaches and thereby gives students the opportunity to develop relevant AI skills without putting them in the abject state of feeling forced to go against their values just to ‘keep up’.
6. Assessment: There were examples of early-stage tests incorporating AI into assessment practices. These didn’t use the technology to generate feedback or grades directly but incorporated it somewhere in the ‘middle’ of the process and notably, in one case, to support dialogue based, ‘live assessment’ practices.
7. Complexity: This is clearly a complex and emergent space. On the one hand there is critique of the mainstream models and how they are framed. On the other there are calls for more clarity about how best to incorporate AI into curriculum and teaching practices. During the symposium I was encouraged to see UAL navigating this complexity and acknowledging there is a lot more to think through. AI might be ‘everywhere’ but it’s still early days.
   
   The breakneck proliferation of the technology has outpaced the development of ‘good AI practice’ and our understanding of what the technologies long-term impact might be. One reading is that almost every institution is ‘behind’ in this regard. I don’t see it that way. My view is that we have invented and distributed a technology before we agreed or discovered what it might be useful for.
   
   Many technologies become incorporated into practice over time in ways in which were not originally envisaged by the designers. However, in the case of AI much of what has been envisaged to date is linked to notions of efficiency and the simulation of human-like skills. The question of the technologies’ relationship to the production of understanding or cultural value has been less considered.
8. Demos: The symposium hosted a small demo area where staff showed work in progress with AI in various forms. The range of these demos was also indicative of the complexity of the space as some used AI directly to intentionally produce creative work while others made visible the inherent tensions in the technology.

The symposium was a useful moment in time to share practice and get a sense of where UAL thinking is at. It fuelled debate and acted as a waypoint in ongoing discourse about AI as a technology of cultural production with all its inherent ethical and political implications. There are several lines of inquiry that emerged from the day and plenty to be incorporated into practice and fed into strategy.

We avoided simplistic answers and instead revealed a solid foundation of AI expertise within UAL to be built on. In part one the answer to “Where are the people?” is “They are hiding behind the technology”. In contrast the ‘people’ of UAL, students, academics, technical and support staff, are actively and often visibly engaged with AI and the debates the technology engenders.

(Part 2 is a review of the University of the Arts London AI and Creative Education symposium.)

At a recent workshop I was running on the University of the Arts London AI Principles I paused after a couple of minutes of setting the scene and asked if anyone had any questions. The first was “Where are the people?”. I asked what they meant and they said that when they use AI (by which they meant mainstream chatbots) the answers come up, “…but where are the people based who are writing the answers?”. It took me a couple of seconds to reorient my perspective and give a quick overview of the basics of how the technology functions.

By the end of the workshop the questioner had rebuilt their concept of the technology and asked a couple of pertinent questions based on their new understanding. Principle 4, “Think of AI as a machine, not a person” was a useful point of discussion.

The question came from someone in a role which does not require specific digital expertise or specialist digital knowledge. It’s a perfectly reasonable question if you have little technical context and exactly what the providers of the main GenAI platforms want you to think; that it is a person.

I mentioned the workshop question to a technical developer friend of mine and a couple of days later, out of nowhere, he brought it back up and said that he thought it was profound. “Where are the people?”, he said slowly. Think about it, he said, directing the question at me, “Where are the people?”. This took me by surprise as I had encountered the moment as a failure of my workshop design, my habit of not laying out the basics of a subject before plunging into interpretation. My friend’s reaction made me rethink and reinterpret the question. Yes, it came from a lack of technical knowledge, but it highlighted something deeper.

Everything technological is ‘people’ and when this is obscured, it is almost always to misdirect our attention away from an asymmetry or an abuse of power. “The algorithm has decided” is a neat way to relocate a decision from the designers or managers of a system into the system. The judgement taken has been abstracted into the machine, but the intention still lies with those who control the technology. Perhaps we forget this because we are coxed into believing that we, the users, control the technology?

This mechanism is inherent in many complex systems, including long standing institutions. “The process has decided” implies that a decision has been reached entirely objectively, ironing out the subjectivity or bias of individual decision makers. Any yet, somebody designed the process or signed it off. So, we can always ask “Where are the people?”, a fundamentally political act.

The question highlights that the systems which shape our lives are invented and not discovered. Those who present AI as ‘inevitable’ are implying that the technology is naturally occurring to further obscure the intentions inherent in what has been designed. Mainstream GenAI, especially chatbots, add another layer to this judgement distancing by designing in a persona who is making the decisions. The simulated ‘person’ in the mainstream chatbots is promoted as knowing more than any human with the underlying assumption that more data equals more truth or a greater objectivity.

This is a heady, having your cake and eating it, combo. On the one hand the AI has the contemporary authority of data and on the other it has the authority of a person who ‘knows’ all this data. The AI is sold as a non-subjective subject, a construct usually reserved for deities.

A designer of these systems might point out that they are probabilistic and therefore the judgements made cannot be traced back, there is no mathematical accountability path. All the AI is doing, it could be claimed, is independently finding the truth in the data. Data so vast that the truth must be in there somewhere?

Even beyond the obvious limits of a philosophy of data-as-truth, there are layers and layers of directly constructed interventions to manage and ‘refine’ outputs. For example, the built-in sycophancy which is redolent of the chemical manipulation of tobacco to push addiction. Also, the software harnesses wrapped around the core inference models as a conscious management of the technology conforming to specific social codes and ideologies.  

Any complex system is the designed accumulation of decisions made by people. It might not be possible to trace any given decision back to an individual but there were in there somewhere. So, the question “Where are the people?” might appear naive in our immediate understanding of GenAI because it’s a machine, not a person. However, the very same question when applied not to the use of GenAI, but to its design, does become profound. “Where are the people” should be asked of any complex system because while it might be possible to automate judgement, decisions are always a result, however obscured, of design and design can always be traced back to people.

‘We don’t have all the answers or even all the questions…’ …is how I start most of my talks on AI. So given the emergent state of the technology how do you develop a set of AI Principles for all staff at a large arts university?

On the one hand you must highlight inherent risks and ethical complexities, on the other you don’t want to stifle experimentation and creativity. Any Principles should create a critical space where both resistance and engagement can co-exist and, importantly, be in dialogue with each other. This allows our staff and students to variously learn, co-opt and innovate while questioning and critiquing the ‘zero-sum’ future promoted by Silicon Valley.

These approach underpinned the development of the Principles below which provide a shared framing for approaches to AI across almost all contexts at UAL. Any plans for, or use of, AI can be held up to the Principles to assess what level of risk (technical, conceptual or social) is being taken and to indicate what areas should be given additional consideration. Any use and any critical refusal of AI can be informed by the Principles.

These high-level Principles sit alongside the guidance developed specifically for the use of AI within Teaching and Learning. As mentioned they have been designed for staff (academic, technical and professional services) but they can also be used by and with students.

As outlined in our internal version of the Principles, “They are primarily designed to respond to Generative AI, which can produce text, image and other forms of media when prompted.”

---

1)  Be intentional and stay in control of the process

• Stay cautious and curious. Question what should be trusted and why.
• Start by exploring AI use to improve the quality of your work, before considering how it might help you to work faster. Consider the balance between convenience and active decision-making.
• If the use of AI in your process is significant (for example, it is intended to influence decision making), track and cite your use as part of the work.

2) Don’t automate your judgement

• You are accountable for your use of AI. Have clear aims and standards to assess the quality of AI outputs.
• Critically evaluate (review, edit and adjust) all AI outputs. AI output often looks good at first glance but might contain errors or omissions, especially if you prompt AI to produce a lot of material with minimal input or effort.
• Be cautious when using AI in processes which have an impact on others. AI reflects biases inherent in its training data and therefore inherent in wider society.

3) Use responsibly: there will be climate, social and reputational impacts 

• Consider whether AI is necessary for what you’re trying to achieve.
• Digital work has physical impacts and AI can use a lot of energy. Intentional approaches to AI should minimise use in line with our Climate Action Plan.
• As a creative community, we value intellectual property. Authors and creators whose work is used to train AI are often not credited.
• Some aspects of AI use are in tension with our Climate, Racial and Social Justice principles. Equality should be considered in both use and access to AI tools. Consider the ethics of AI use in your context.

4) Think of AI as a machine, not a person

• Generative AI is designed to mimic human-like interaction, but it is a machine. Keep this in mind when using AI tools.
• Some Generative AI is designed to reinforce your point of view rather than enrich your thinking or insights.

5) Don’t assume AI knows everything or is impartial

• The information any AI ‘contains’ is extensive but not exhaustive.
• AI outputs are unlikely to contain niche, marginalised or emergent knowledge, and will not represent all possibilities and perspectives.

6) Only input what you have a right to use and share

• Only use Copilot Chat when logged-in with your UAL account, this will ensure data is protected and your chats are not used to train AI models.
• If uncertain, assume that data may be sensitive and avoid inputting confidential, personal, or restricted information. Use only a UAL‑approved AI tool (Copilot Chat) when working with protected information.
• Ensure any information you put into an AI tool complies with UAL’s Information Security Policy and Data Protection Policy. Guidance can also be found in the ‘AI Guidance on Information Governance and Risk’ (links to policies removed for this blog post as they or located on the UAL intranet)
• Avoid entering material that may have intellectual property (IP) restrictions, such as, copyrighted texts, images, or assets you don’t have permission to use.

7) Using AI not provided by UAL incurs increased risk and costs.

• Even when using UAL‑approved tools such as Copilot Chat, critical judgement is still required. All other UAL policies and the AI Principles continue to apply.
• Using AI that is not approved by UAL introduces significant risks and may put students’ and staff data at risk.
• Consider your flexibility to change provider as costs are likely to increase over time.

---

The Process

The process to develop the Principles was in-depth but not complex. A couple of writing workshops with members of the UAL AI Group I co-chair and consultation with relevant groups and expert individuals. Having been immersed in ‘AI and education’ for a couple of years I didn’t find it too difficult to draft a first version for discussion and editing. The headline themes from various AI talks I’ve given since 2023 acted as a reasonable starting point.

Over a couple of months of consultation we narrowed it down to seven Principles which covered a lot of ground. One of the strengths of UAL as a community is that there is an understanding that brevity done well is difficult but worth the effort. The process was also helped by the great work that had already been done by our Teaching and Learning Directorate who, like many universities, had rapidly developed guidance for staff and students in the use of AI within the curriculum.

The Language

The Principles had to work across all our staff groups, not just the academic community. Plus, I was confident that if we got this right the Principles would also be used by and with students which seems to be happening already. Respect to our internal communications department here who are experts in spotting overly complex language and were happy to suggest much improved alternatives. (For example, my first draft contained the phrase ‘attenuating agency’.).

I was also keen to avoid anthropomorphising AI through our use of language used and so stuck with phrases like ‘the output from AI’ rather than referring to AI as ‘it’. This is pretty subtle but I believe we have to constantly reassert that AI is a machine at every possible opportunity.

The ground covered

Although not presented in this way I think of the seven Principles in three sections:

Practice: Principles 1, 2 and 3.
Incorporating AI into process and practice (or why you might actively chose not to). The central point being that whatever the technology produces a person, probably you, will be responsible for the overall process.

Reality check: Principles 4 and 5.
I am certain that being aware that AI is a ‘machine with limits’ leads to better critical evaluation of outputs than imagining or assuming AI is an all-knowing guru. This might seem obvious but it’s easy to tacitly fall into this trap because of the way the technology is packaged and presented. i.e. Most of the major Gen AI providers want you to think of the technology as an omniscient person.

Safety and data: Principles 6 and 7.
These are specific to UAL in that they refer to Copilot Chat which is provided to all staff at UAL. For the most part these two Principles are simply a reminder of the IT policies which have applied to the data and the digital environment for years. The main addition here is around cost as the try-before-you-buy model of technology proliferation is ok if you can hop between platforms (as many of our students do) but risky if you build a ‘free’ technology into business-as-usual processes.

Principles not rules

Beyond a couple of redlines around academic misconduct and data security AI use is very much a live debate. We need to be critically questioning and thoughtfully experimenting. As the a recent report from HEPI pointed out we can’t expect our communities to develop critical approaches to AI if we don’t trust them to experiment with AI thoughtfully. Whether its developing ways of conceptualising and understanding what the tech might mean or it’s experimenting with what the tech can and cannot do, this must involve trust.

In that spirit, we are developing a workshop which involves using the Principles to assess your use, or critical non-use, of AI. An approach to AI which is in tension with a Principle is not necessarily something to be stopped, it is something which requires a deeper level of consideration.

Experimentation and innovation by their very nature will involve taking risks. The question then becomes identifying these risks and seeking agreement that they are reasonable. How reasonable a risk might be will depend on context and potential impact.

This is nothing new but each emergent digital technology offers more power and therefore greater risks. With AI however, the sheer scale of investment has generated a huge amount of hype and vested interests. An individual who gets caught up in this might have poorly informed expectations of the technology or they might have a false conception of how the tech functions. In this case many unforeseen negative effects are likely.

The Principles act as set of coordinates to help navigate the sociotechnical storm of AI.